CMS baseline: GUIDE RFA Appendix B, Section 6. The Proxi column identifies candidate automation, not functionality already deployed. The executed GUIDE Participation Agreement remains the final requirements check before these tasks are used in the field. Respite is outside this SOP.
In the Community connection tasks, the service seeker may be the beneficiary, the caregiver, or both. When the service is caregiver-facing, the caregiver makes their own choice, authorizes sharing of their own information, completes their own intake requirements, and evaluates whether the service met their need.
In Person required?, Yes means some person — including the service seeker or an external organization — must perform or own part of the task, Partial means software can perform substantial preparation or routine execution but a person is needed for a choice, authority act, external decision, or triggered exception, and No means Proxi can complete the routine path after approved prerequisites are present. This column does not by itself create Proxi labor. The detailed cards control clean-path versus triggered human minutes: the service seeker chooses, the external program decides eligibility/admission/service delivery, and Proxi owns preparation, connection mechanics, pursuit, and truthful tracking rather than the person's or program's decision.
Detailed community-connection procedure — tasks 1-57#
This pass expands all 57 tasks, from receipt of a need through service-start/outcome follow-up, community-inventory operations, the AAA or Tribal Aging Program route, and applicable Medicaid HCBS coordination. The detailed inventory remains below as the compact cross-check.
Source, service-seeker, and labor boundary#
Cross-pillar preparation, no-repeat handoffs, owner acceptance, and shared labor follow 25_GUIDE_Eight_Pillar_Service_Integration.md. A clear authenticated service-seeker choice may proceed directly; L2 enters only for an actual required or requested human service or another D-021 condition.
The current public requirement anchor is the CMS GUIDE Request for Applications, Appendix B 6.1, 6.3, and 6.4 (SRC-CMS-GUIDE-RFA-V1), rechecked 2026-07-12. CMS requires referral and connection to community-based services and supports, participation in or maintenance of a community referral inventory for screened HRSNs, maintenance of or access to a relevant community-resource inventory, and appropriate sharing of those resources. CMS does not prescribe Proxi search logic, ranking method, inventory freshness intervals, staff minutes, option count, direct-verification cadence, or proof standards for practical fit and dynamic availability. Those are Proxi design decisions under O-022 and O-044. The executed Participation Agreement remains missing under O-PA-001.
The service seeker may be the beneficiary, caregiver, or both. When the need is caregiver-facing, the caregiver defines the caregiver's desired result, constraints, participation, information-sharing choice, and service choice. Caregiver information does not silently become beneficiary information, and caregiver participation does not automatically establish authority over a beneficiary-facing service.
All minutes below are provisional calibration assumptions, not CMS requirements or field measurements. The clean connected-data path through these ten rows can be 0 Proxi human minutes: software can receive authenticated inputs, ask bounded clarification questions, search current sources, filter hard mismatches, and compare factual service requirements. L2 does not validate every deterministic match or clear authenticated choice. L2 enters only when a human GUIDE service is required, the person requests help, relationship-sensitive practical fit remains, or ambiguity, disagreement, distress, accessibility failure, unclear authority, a warm introduction, or useful human confirmation is present. When current information cannot be obtained electronically and manual organization outreach is needed, use one shared provisional 13-minute L1 Philippines administrative exception episode for the search/verification batch, not 13 minutes for each row or each candidate. L1 may capture administrative facts verbatim; the external program alone decides its own eligibility, acceptance, availability, pricing, waitlist placement, and intake disposition.
In every task card, L3/L4 means the separate U.S. beneficiary-location-authorized clinical workforce under D-015. Puerto Rico L2 performs nonclinical navigation only and never performs clinical or medical work. Low / typical / high estimates are per applicable episode unless a card explicitly says the estimate is per inventory batch or per organization record; conditional minutes are not universal beneficiary touches.
Search, filtering, or prequalification does not by itself prove that CMS-required referral/connection occurred or that a qualifying GUIDE service was furnished. CC-E26, O-022, and O-025 preserve the later distinctions among resource sharing, referral sent, receipt, intake, service start, and actual connection.
Community connection — receive an identified need#
When. Whenever a beneficiary-facing or caregiver-facing community need is identified through the comprehensive assessment, caregiver assessment, care plan, ongoing contact, transition, 24/7 call, outside source, or direct beneficiary/caregiver request. Repeat only for a materially distinct need or corrected/new request; do not create accidental duplicate work.
What and how much. Open one distinct community-connection work item for each service seeker and materially distinct need. Capture what was reported, who needs the service, who supplied the need, how current it is, why help is wanted now, any prior attempts/results, and whether another open work item may be the same need. Keep beneficiary needs separate from caregiver needs even when one event reveals both. Preserve multiple real needs rather than merging them for convenience.
Data. Use service-seeker identity, need statement in the speaker's words, source and observed/reported time, beneficiary/caregiver relationship and authority facts when relevant, related assessment/care-plan findings, prior attempts, current location, and any urgent/clinical signal. An assessment finding proves that a need was observed at assessment time, not that it remains current or that the service seeker wants a referral. A caregiver report proves the caregiver reported it, not that the beneficiary agrees. A duplicate-looking record does not prove duplicate need.
Potential Proxi work. Ingest the source, preserve original wording, connect related evidence, identify potential duplicates, ask the authenticated service seeker whether the need remains current, and create a concise work summary. AI may summarize but cannot decide currentness, merge distinct needs, infer authority, select a service, or close the work from nonresponse.
Human role. No human is required on the authenticated, current, unconflicted path. L2 handles conflict between beneficiary/caregiver reports, sensitive context, uncertain service seeker, or a need whose currentness cannot be established digitally. L1 may pursue a missing administrative source but does not determine need priority or authority. Separate U.S. clinical/safety staff enters only on the task-specific trigger handled in the fifth row below.
Provisional clean-path Proxi human minutes by role. L1 0; L2 0; L3 0; L4 0. Conditional L2 clarification is an exception or part of an already-scheduled navigator contact, not a universal new touch. These are calibration assumptions.
Completion evidence. Distinct work item identifies the service seeker, current or pending-current need, source, time, prior attempts, and related open work; duplicates are linked rather than discarded; and any urgent/clinical signal is visible for immediate routing.
What does not prove completion. A raw assessment code; an AI summary; caregiver statement presented as beneficiary agreement; a copied care-plan need with no currentness check; a duplicate flag; or a closed prior referral assumed to resolve a new request.
Edge cases/open decisions. Existing: X-001, X-002, X-003, X-004, X-005, X-009, X-015, CC-E16, CC-E19, CC-E20; O-004, O-011, O-022, O-PA-001. Minimum tests: beneficiary request; caregiver requests their own support; caregiver reports a beneficiary need; stale assessment need; two similar but distinct needs; exact duplicate; corrected source; move during intake; immediate safety statement.
Community connection — define the desired result#
When. After the need is received and before inventory search, unless the source already contains a current authenticated desired result. Reopen when the service seeker says the desired result changed or when later constraints show the original result was misunderstood.
What and how much. Produce one service-seeker-authenticated outcome statement for each distinct need: the problem to solve, what practical improvement would count as useful, who should receive the benefit, any timing concern stated by the service seeker, and what they do not want. Keep the outcome service-neutral so search is not prematurely restricted to a familiar program or provider.
Data. Use the original need, authenticated beneficiary/caregiver response, communication/accommodation needs, care-plan goal where relevant, and any legal-representative scope. A care-plan goal may inform the question but does not prove the current desired community-service result. A selected category or provider name does not prove the underlying outcome. AI-rephrased text is not authenticated until the service seeker confirms or corrects it.
Potential Proxi work. Ask bounded text, voice, or chat questions; provide plain-language examples; preserve the original answer; propose a concise outcome statement; detect internal conflicts; and allow correction. Proxi cannot choose the outcome, convert silence into acceptance, or make a clinical/safety priority decision.
Human role. No human is needed when the authenticated service seeker can define and confirm the desired result digitally or by bounded AI conversation. L2 enters for ambiguity that changes the service search, distress, disagreement, communication barriers, or a preference-sensitive conversation the person wants. A separate U.S. authorized clinician enters only if defining the result requires a clinical judgment.
Provisional clean-path Proxi human minutes by role. L1 0; L2 0; L3 0; L4 0. Conditional L2 time is bundled with an existing navigator contact where possible. These are calibration assumptions.
Completion evidence. Service seeker confirms a specific, understandable outcome statement tied to the need; original wording is preserved; unclear or conflicting parts remain explicit; and the search can use the outcome without assuming a provider or service choice.
What does not prove completion. Generic category such as "transportation"; an assessment label; navigator interpretation without service-seeker confirmation; provider name; AI-generated goal; or a caregiver outcome substituted for a beneficiary-facing need without applicable authority.
Edge cases/open decisions. Existing: X-002, X-003, X-004, X-009, X-015, CC-E19; O-011, O-041, O-044. Minimum tests: clear digital answer; vague request; caregiver's own goal; beneficiary/caregiver goals differ; representative scope partial; language/accessibility need; service seeker changes goal; nonresponse; clinical goal embedded in a social request.
Community connection — confirm who should participate#
When. After identifying the service seeker and before collecting or sharing person-specific constraints; repeat whenever the participant list, caregiver relationship, representative authority, disclosure preference, or service recipient changes.
What and how much. Establish one participation record for the connection episode: who is the service seeker; who receives the service; who may join option review, choice, intake, and follow-up; whose information belongs to whom; and what role each person has. For a caregiver-facing service, preserve the caregiver's independent right to choose and share the caregiver's own information. For a beneficiary-facing service, do not treat caregiver presence as authority to choose or disclose.
Data. Use authenticated identities, beneficiary participation preferences, caregiver relationship, representative documents/scope/effective period, service recipient, caregiver's own consent for caregiver information, and any conflicts/revocations. Emergency contact status does not prove representative authority. Caregiver status does not prove permission to receive beneficiary information. A representative's authority for one domain does not prove authority for every service decision.
Potential Proxi work. Present known participation facts, ask each relevant person to confirm or update them, distinguish beneficiary/caregiver records, flag scope/expiry/conflict, and limit later views/messages to confirmed participants. AI cannot decide whose choice controls, infer authority, or resolve disagreement.
Human role. No human is needed for a clean confirmed participant list. Software enforces the approved participation, permission, representative-authority, and disclosure rules. L2 authenticates preference-sensitive participation, obtains missing relationship facts, and handles beneficiary/caregiver disagreement without deciding legal authority. Healthcare Legal Counsel receives only a prepared legal-authority question that the approved rule cannot resolve. Separate U.S. clinical staff handles decision-specific clinical capacity only when triggered. L1 may collect documents but cannot decide their legal effect.
Provisional clean-path Proxi human minutes by role. L1 0; L2 0; L3 0; L4 0. These are calibration assumptions.
Completion evidence. Current participant list, service seeker/recipient distinction, roles, information ownership, participation scope, and authority/permission facts are explicit; unresolved authority blocks only the affected choice/disclosure rather than the whole search.
What does not prove completion. Caregiver listed in the chart; portal proxy; emergency contact; attendance at a prior visit; family relationship; beneficiary silence; or representative document with no scope/effective-period check.
Edge cases/open decisions. Existing: X-001, X-002, X-004, X-005, X-015, CC-E19, CC-E35; O-011, O-PA-001. Minimum tests: beneficiary alone; caregiver seeking own service; caregiver assists beneficiary; two caregivers disagree; valid representative with limited scope; expired authority; revocation; capable beneficiary objects; service recipient changes.
Community connection — gather service constraints and preferences#
When. After the service seeker and participants are established and before search/filtering. Refresh any constraint likely to change—location, schedule, transport, technology, cost, caregiver availability—before relying on it for a consequential match.
What and how much. Gather every search-relevant fact for the need: service location and acceptable travel radius; in-home/virtual/site preference; language and cultural preferences; accessibility; schedule; transportation; technology/device access; cost ceiling and funding preference; service population; caregiver availability; required modality; and any other practical requirement. Distinguish hard constraints that make an option unusable from preferences that influence ranking but do not automatically exclude it. Record unknown rather than guessing.
Data. Use authenticated service-seeker responses, current residence/service location, communication/accommodation needs, benefit information when available, caregiver schedule only when caregiver involvement is permitted, and source/date for every prefilled fact. Address does not prove transportation. Insurance/Medicaid status does not prove program eligibility or cost. Language field does not prove preferred service language. Missing information is unknown, not unrestricted.
Potential Proxi work. Conduct the structured intake, prefill source-labeled facts, ask only missing questions, separate hard constraints from preferences, detect conflicts/staleness, and prepare search filters. Proxi cannot decide that a stated preference is unimportant, invent a cost tolerance, or treat missing data as permission to include/exclude.
Human role. No human is needed for a complete, authenticated intake. L2 enters when relationship context changes the practical constraint, when the service seeker is unsure and wants help, or when participants disagree. L1 may verify an administrative fact but cannot choose which preference to relax. Separate U.S. clinical staff enters only for a true clinical suitability/safety question.
Provisional clean-path Proxi human minutes by role. L1 0; L2 0; L3 0; L4 0. These are calibration assumptions.
Completion evidence. Search-ready constraint set with source/currentness, hard-versus-preference designation confirmed by the service seeker, explicit unknowns, and participant-specific facts kept separate.
What does not prove completion. ZIP code alone; a generic language field; prior transportation use; Medicaid eligibility; caregiver schedule assumed from residence; AI inference from demographics; or all blank fields defaulted to "no preference."
Edge cases/open decisions. Existing: X-002, X-003, X-004, X-005, X-009, X-015, CC-E20, CC-E27; O-011, O-022, O-044. Minimum tests: all constraints known; missing transport; hard cost cap versus soft preference; temporary location; accessibility need; beneficiary/caregiver disagreement; changed schedule; no technology; culturally specific preference; no option satisfies all hard constraints.
Community connection — redirect urgent or clinical needs#
When. Before routine inventory search and again whenever any new statement or source introduces possible immediate danger, abuse/neglect/exploitation, medical symptoms, medication concern, behavioral-health crisis, or another issue requiring qualified clinical/safety action.
What and how much. Screen the exact reported concern for explicit approved stop/transfer conditions. If one is present or ambiguity itself requires professional assessment, pause only the affected routine community-search work and connect the person/facts to the approved human clinical, safety, protective-service, or emergency route. Keep the community need open for later continuation unless an authorized person determines otherwise. Do not ask L2 or L1 to make the clinical urgency judgment.
Data. Use the speaker's verbatim report, time, current location, person at risk, immediate circumstances, identity/callback information available, prior relevant plan facts, and actual handoff/answer evidence. An AI cue is a candidate signal sufficient to summon human review, not proof of urgency, diagnosis, causality, or appropriate destination. Completing an emergency handoff does not prove the community need is resolved.
Potential Proxi work. Surface explicit concern language, display the approved route, connect the correct human, preserve the caller/source words, alert on failed handoff, and keep the routine work from proceeding unsafely. Proxi cannot determine urgency, diagnose, select treatment, decide abuse, or close the safety/clinical issue.
Human role. L1/L2 may recognize a stop cue and initiate the approved connection but do not triage. Separate U.S. beneficiary-location-authorized L3/L4, emergency services, 988, APS, or other approved authority determines and acts within scope. L2 resumes relationship/navigation work after the safety/clinical disposition allows it.
Provisional clean-path Proxi human minutes by role. For the ordinary nonurgent community need, L1 0, L2 0, L3 0, L4 0. Triggered clinical/safety labor is event-based and belongs to the relevant clinical/safety pool, not a universal row minute. These are calibration assumptions.
Completion evidence. Either no reported/known stop condition after required bounded questions, with source/time retained, or an actual verified human handoff to the appropriate route with failed-handoff recovery. The community-search work is paused/resumed truthfully and remains distinct from the safety/clinical outcome.
What does not prove completion. AI risk score; keyword alone; L1/L2 opinion that the issue is safe; voicemail to a clinician; 911/APS number displayed but not connected; emergency handoff treated as referral connection; or silence interpreted as no risk.
Edge cases/open decisions. Existing: X-003, X-004, X-008, X-009, X-010, X-014, X-015, CC-E13, CC-E19; O-003, O-004, O-032. Minimum tests: routine meal need; fall with injury; suspected exploitation; suicidal statement; vague "not safe"; medication symptom; caller location unknown; clinical route fails; emergency resolved but meal need remains.
Community connection — search the available inventory#
When. After a current desired result and search constraints exist and the fifth-row screen has not identified an unresolved urgent/clinical stop. Repeat when constraints change, inventory facts stale, serious options fail, or no-match search scope materially expands.
What and how much. Search the maintained/accessed community inventory across the full applicable service area and need category. Apply current hard constraints and retain every serious candidate rather than only the first result. Search beneficiary-facing and caregiver-facing needs separately where service populations differ. If no serious candidate exists, return a truthful no-match with searched scope, not a fabricated recommendation.
Data. Use desired result, hard constraints, preferences, current service location, inventory entries, service categories, populations served, coverage geography, source/date, freshness status, and known exclusions. Inventory inclusion proves candidate existence only, not current operation, availability, eligibility, affordability, intake acceptance, quality, clinical suitability, or connection. Search-engine ranking does not prove relevance or endorsement.
Potential Proxi work. Perform the complete search, normalize entries, combine approved inventory sources, exclude known hard mismatches, preserve source/currentness, disclose search coverage and exclusions, and return candidates/no-match. Search/filtering is definitely software on the clean path. AI may map need language to service categories but cannot choose a service or turn uncertain categories into fact.
Human role. No person is required for the initial search. L2 does not review every result. L1 enters only when the inventory is incomplete or inaccessible or a local manual source must be checked. L2 enters later for genuinely relationship-sensitive fit, choice, or referral coordination.
Provisional clean-path Proxi human minutes by role. L1 0; L2 0; L3 0; L4 0. Any phone/paper-only search begins the one shared provisional 13-minute L1 exception episode for the batch. These are calibration assumptions.
Completion evidence. Source-labeled candidate set or truthful no-match; search scope/categories/geography disclosed; hard filters traceable to confirmed constraints; stale/unknown data visible; and no unexplained/sponsored default ranking.
What does not prove completion. One web search result; first result only; a directory dump; sponsored listing; stale inventory hit; service category inferred by AI without evidence; or a candidate list that hides no-match areas and excluded sources.
Edge cases/open decisions. Existing: X-003, X-004, X-005, X-009, X-010, X-015, CC-E18, CC-E20, CC-E27; O-001, O-022, O-044. Minimum tests: several current matches; no match; all matches stale; wrong service area; caregiver-facing versus beneficiary-facing service; phone-only inventory; duplicate organizations; sponsored result; constrained language/accessibility; changed location.
Community connection — build a relevant option list#
When. After the inventory search returns candidates and before dynamic organization verification or option explanation. Rebuild when a hard constraint, inventory fact, availability fact, or desired result changes.
What and how much. Produce one comparison list containing every serious candidate that survives confirmed hard constraints, with each candidate's matched need, material advantages, hard/soft constraint result, source date, unknowns, and reasons other candidates were excluded. Rank only by disclosed service-seeker priorities and factual fit; do not silently prefer a partner, sponsor, high-margin service, or familiar organization. A truthful no-match is an acceptable result.
Data. Use search result set, desired result, hard constraints/preferences, service purpose/population/geography, source/currentness, and any disclosed commercial relationship. Factual filter success proves only candidate-level match on known data. It does not prove beneficiary/caregiver preference, community-support fit, external eligibility, availability, cost, quality, clinical appropriateness, or eventual connection. An LLM explanation is not evidence of fit unless grounded in cited facts.
Potential Proxi work. Remove hard mismatches, calculate factual comparison, preserve all serious candidates, generate plain-language reasons/unknowns, disclose inventory scope and commercial relationships, and provide no-match. Software should perform this entire clean route; AI may explain but cannot insert unsupported criteria or select for the service seeker.
Human role. No universal navigator validation. L2 reviews only when a relationship-sensitive fact cannot be represented deterministically, participants disagree, the service seeker asks for human help, or the option carries a genuine sensitive barrier. A clinician reviews only clinical suitability, not routine community-service matching.
Provisional clean-path Proxi human minutes by role. L1 0; L2 0; L3 0; L4 0. Conditional L2 option discussion belongs to later explanation/choice rows, not this filtering row. These are calibration assumptions.
Completion evidence. Complete serious-option comparison or truthful no-match; each inclusion/exclusion tied to current facts; unknowns and source dates visible; ranking rationale disclosed; and no candidate chosen on behalf of the service seeker.
What does not prove completion. Short list with no excluded-candidate reasons; top-ranked search result; navigator signature on software output; sponsored option without disclosure; fabricated alternatives to avoid no-match; or list presentation treated as service-seeker choice.
Edge cases/open decisions. Existing: X-003, X-004, X-005, X-009, X-010, X-015, CC-E16, CC-E27; O-022, O-044. Minimum tests: exact hard match; several equally suitable candidates; one candidate violates cost ceiling; unknown accessibility; no match; sponsored partner; duplicate listing; stale top option; caregiver/beneficiary constraints differ; AI explanation cites wrong fact.
Community connection — confirm community-support fit#
When. After the relevant option list is built and before the service seeker is asked to choose. Perform for each serious option using the identified need, service recipient, service purpose, setting, and confirmed constraints. Repeat when material service facts or need/setting change.
What and how much. Confirm whether each serious option is, factually, the right kind of community support for the identified beneficiary/caregiver need and setting. Separate three questions: (1) factual/practical fit that software can compare; (2) relationship-sensitive fit that the service seeker/L2 may need to discuss; and (3) clinical suitability that only an authorized clinician can decide when relevant. Do not use this row to decide external-program eligibility or availability.
Data. Use desired result, service recipient, service purpose/population, delivery setting, hard constraints/preferences, safe-community-living relevance where applicable, source/currentness, and unresolved ambiguity. Category match does not prove practical fit. Practical fit does not prove external eligibility, current availability, quality, clinical safety, service-seeker choice, or connection. A navigator's familiarity is not current organization evidence.
Potential Proxi work. Compare current service facts against the need/setting/constraints, explain matched and mismatched elements, route unknowns to verification, and identify the narrow subjective question that may need L2 or service-seeker input. Proxi should confirm deterministic fit when facts are sufficient; it cannot decide clinical suitability or preference-sensitive tradeoffs.
Human role. L2 validates only relationship-sensitive practical fit—such as whether a setting, routine, cultural context, or caregiver dynamic changes the person's likely willingness/use—when software facts are insufficient. The service seeker ultimately chooses later. Separate U.S. L3/L4 enters only for genuine clinical suitability/safety, not ordinary service fit. L1 verifies factual organization details only.
Provisional clean-path Proxi human minutes by role. L1 0; L2 0; L3 0; L4 0 when facts establish practical fit. If a soft/relationship-sensitive issue remains, use a provisional 3 L2 minutes for the option cluster, not per option, preferably within the later option discussion. These are calibration assumptions.
Completion evidence. Each serious option has a factual fit/mismatch/unknown explanation; subjective and clinical questions are explicitly separated and routed; no eligibility/availability/choice claim is made; and a no-fit result remains truthful.
What does not prove completion. Service category label; navigator familiarity; dementia marketing; inventory listing; external eligibility prediction; geographic proximity alone; clinician review of every option; or software score presented as the person's choice.
Edge cases/open decisions. Existing: X-003, X-004, X-005, X-009, X-010, X-015, CC-E18, CC-E27; O-003, O-022, O-044. Proposed open decision after deduplication: definition/evidence for deterministic practical fit versus relationship-sensitive fit versus clinical suitability. Minimum tests: clear meal-delivery match; adult-day program with schedule mismatch; cultural preference ambiguity; accessibility unknown; clinically restricted activity; caregiver versus beneficiary service; no fit; multiple equal fits.
Community connection — verify current service availability#
When. After serious options survive fit review and before option explanation/choice. Reverify any time the availability source becomes stale, the service seeker delays action materially, an intake fails, or the organization reports a change.
What and how much. For each serious option, establish whether the organization is currently operating, serves the service seeker's geography and population, accepts new inquiries/intakes for the relevant service, and has a working contact/intake route. Record current hours and any temporary closure or limited capacity that affects pursuit. Keep "operating," "accepting inquiries," "accepting applications," and "has an immediate opening" separate.
Data. Use official organization/service source, maintained inventory, direct response when needed, observation/verification date, geography/population, service-specific capacity statement, contact route, and source confidence. Website presence does not prove operation. A directory listing does not prove capacity. Accepting inquiries does not prove eligibility, intake acceptance, placement, service start, or ongoing availability. A staff statement is time-bound to what was actually asked and answered.
Potential Proxi work. Recheck current organization sources, compare changes, verify endpoint operation, distinguish availability meanings, timestamp results, and prepare specific unanswered questions. If an API/current source supplies the facts, software completes the task. For phone/paper-only services, L1 can ask the approved factual questions and record the answer verbatim.
Human role. No navigator is needed for clean source-backed availability verification. L1 performs manual administrative outreach when no connected/current source exists. L2 enters only if a service-seeker relationship issue affects whether to continue; the external organization alone states its capacity or availability.
Provisional clean-path Proxi human minutes by role. L1 0; L2 0; L3 0; L4 0. Manual outreach is part of the one shared provisional 13-minute L1 exception episode for the verification batch, not a separate charge per candidate. These are calibration assumptions.
Completion evidence. Dated, service-specific availability statement and source for every serious option; verified geography/population and working intake route; distinctions among inquiry/application/opening retained; unresolved facts visible.
What does not prove completion. Website exists; phone rings; directory lists program; organization generally operates; old successful referral; "call for availability" represented as available; or availability treated as eligibility/service start.
Edge cases/open decisions. Existing: X-003, X-004, X-005, X-006, X-013, X-015, CC-E18, CC-E20, CC-E27; O-001, O-004, O-022, O-044. Minimum tests: current opening; accepting waitlist only; seasonal closure; website stale; phone disconnected; serves adjacent county but not current ZIP; correct organization/wrong service; no response; availability changes after verification.
Community connection — verify eligibility, cost, waitlist, and intake needs#
When. After service availability is established for serious options and before the option comparison is presented for choice. Refresh dynamic facts if the service seeker delays, circumstances change, or the organization updates rules/capacity.
What and how much. For every serious option, gather the basic published or directly stated eligibility criteria, application versus enrollment distinction, expected cost/fees, funding/subsidy possibilities, current waitlist status/estimated delay, required documents, intake steps, and any material commitment. Compare the service seeker's confirmed facts only to identify likely match, likely mismatch, and unanswered questions. Never state that the person is eligible, approved, enrolled, funded, or placed unless the external program issues that result.
Data. Use organization-issued criteria/current response, source/date, service-seeker facts with provenance, benefit/funding information, fee schedule, waitlist statement, required-document list, and intake channel. Published eligibility rules do not prove individualized eligibility. Apparent fact match is pre-screening only. Cost estimate does not prove final charge, coverage, subsidy, or affordability. Waitlist length does not guarantee timing. Document readiness does not prove intake acceptance.
Potential Proxi work. Retrieve current criteria/cost/waitlist/intake facts, compare confirmed data without deciding, identify missing documents/questions, produce a factual comparison, and preserve source/currentness. L1 may call a phone-only organization for factual clarification. AI cannot approve eligibility, promise cost/funding, estimate placement without source, or convert likely match into acceptance.
Human role. No L2 is required to collect or compare routine facts. L1 verifies dynamic organization facts when they are not electronically available. L2 later explains material tradeoffs and supports the service seeker's choice; it does not decide external eligibility. The external program performs eligibility, intake, waitlist, pricing, funding, and admission decisions. The Medicaid Program Specialist receives only a bounded unresolved Medicaid-program question; a separate U.S. clinician receives only a clinical-content question; and Healthcare Legal Counsel receives only a legal-authority question that the approved rule cannot resolve.
Provisional clean-path Proxi human minutes by role. L1 0; L2 0; L3 0; L4 0. Phone/paper-only verification uses the one shared provisional 13-minute L1 exception episode for the entire batch. These are calibration assumptions.
Completion evidence. Current source-backed criteria, cost/funding facts, waitlist statement, required documents, intake steps, and explicit unknowns for each serious option; likely-match language remains clearly nondecisional; external-program decision boundary is visible.
What does not prove completion. Automated eligibility score; Medicaid/Medicare status alone; likely match; cost estimate; empty waitlist last month; completed documents; staff saying "should qualify"; application submitted; or navigator confidence.
Edge cases/open decisions. Existing: X-003, X-004, X-005, X-006, X-009, X-010, X-015, CC-E18, CC-E19, CC-E20, CC-E27; O-001, O-004, O-022, O-044. Proposed open decision: which dynamic facts require direct organization verification and how long each remains usable. Minimum tests: clearly likely match; likely mismatch; missing income fact; variable/sliding cost; grant funding unknown; open waitlist with no estimate; rules change after collection; staff gives tentative answer; phone-only intake; external denial after strong pre-screen.
Community-connection detailed tasks 11-20#
Tasks 11-20 preserve one referral/intake episode rather than manufacturing a new human touch for each row. The prepared accessible clean route may explain the current source-backed comparison and capture a clear authenticated choice with zero L2 minutes where the final approved policy permits it. When a human service is required or requested, or ambiguity, disagreement, distress, accessibility failure, a warm introduction, or relationship work remains, use one 6-minute typical L2 explanation/choice cluster plus a provisional 1-minute authorized V1 release. Phone, paper, missing-acknowledgement, and manual intake exceptions share one 13-minute L1 Philippines episode across the batch. Referral transmission, recipient receipt, intake disposition, external acceptance, and service start remain separate facts.
Community connection — explain the options#
When. After tasks 1-10 produce a current, source-labeled serious-option set, and again only when material service facts or the service seeker's priorities change before choice.
What and how much. Present every serious option and the truthful no-option alternative. Explain the desired-result match, material benefits and burdens, hard constraints, current availability, likely-but-not-decided eligibility, cost/funding facts, waitlist, intake steps, material unknowns, inventory limits, and any commercial relationship. Do not recommend an option or hide no-match.
Data/provenance. Use the authenticated desired result and priorities, current option comparison, source/verification dates, constraints, availability, organization-issued criteria/cost/waitlist/intake facts, accessibility needs, and disclosed commercial relationships. Presentation does not prove understanding, agreement, choice, permission, eligibility, or acceptance.
Potential Proxi work. Generate and render the factual comparison in approved text, video, or voice formats; answer bounded questions from cited facts; preserve unknowns; and stop when a question requires human authority. AI may explain but cannot steer, suppress alternatives, or decide clinical suitability.
Human role. The prepared comparison is presented first. A service seeker may receive the approved explanation, ask bounded source-backed questions, and make a clear authenticated choice directly on the approved clean path. L2 enters only for the human-entry conditions above and then spends time on voluntariness, unresolved tradeoffs, distress/disagreement, accessibility recovery, or relationship work—not on rebuilding the option list. Interpreter/accessibility support enters when required. A separate U.S. L3/L4 clinician receives a prepared exact question only for genuine clinical suitability.
Provisional human minutes (low / typical / high). L1 Philippines 0 / 0 / 0; L2 Puerto Rico nonclinical navigator 0 / 6 / 12 for the combined explanation-and-choice cluster; separate U.S. L3/L4 clinical 0 / 0 / 10 only when triggered. The low route is the prepared authenticated direct path. Do not add the typical 6 minutes again in the next row.
Completion evidence. The service seeker receives an accessible current comparison, can identify the material options/no-option result, and has questions/corrections recorded without a choice being made for them.
What does not prove completion. Directory sent, top-ranked result, video viewed, chatbot session, navigator recommendation, inaccessible format, or explanation treated as choice.
Edge cases/open decisions/minimum tests. Existing: CC-E06, CC-E07, CC-E19, CC-E20, CC-E27; O-022, O-041, O-044. Tests: several options; no-match; changed price; waitlist; interpreter required; sponsored partner; clinical question; priorities change mid-explanation.
Community connection — obtain the service seeker's choice#
When. In the same clustered event after option questions are addressed, and again when the service seeker changes the choice before release or requests a new option after failure.
What and how much. Record one, multiple intentionally nonduplicative, no-option, deferred, refused, or new-search choice in the service seeker's own terms. Keep beneficiary and caregiver choices separate when they seek different services.
Data/provenance. Use authenticated actor identity, the exact option set shown, explanation record, beneficiary/caregiver/representative role and scope, original choice wording, conditions, and time. Clicking is an event, not automatic proof of identity, voluntariness, comprehension, permission, or external acceptance.
Potential Proxi work. Present bounded controls, preserve original wording, identify apparent duplicates/parallel pursuit, show the next step, and allow correction or withdrawal. Proxi cannot infer a choice from ranking, silence, or caregiver preference.
Human role. The service seeker or decision-specific authorized representative chooses. A clear authenticated choice proceeds directly under the approved rules. L2 supports the choice inside the prior row's clustered human event only when the human-entry conditions apply and preserves disagreement without blessing the choice. Healthcare Legal Counsel receives only a legal-authority question that the approved rule cannot resolve; a separate U.S. clinician receives only a clinical-capacity or safety question within scope.
Provisional human minutes (low / typical / high). L1 0 / 0 / 0; L2 0 / 0 / 3 incremental beyond the prior row (high means a later changed choice requires a short reconfirmation); separate U.S. L3/L4 0 / 0 / 10 only for a triggered capacity/clinical question.
Completion evidence. Authenticated choice/no-choice/defer/parallel-pursuit result, actor and scope, deliberate-parallel rationale, and downstream work limited to the selected route.
What does not prove completion. Explanation, default ranking, prechecked box, silence, staff selection, caregiver choice substituted for beneficiary choice, or application preparation.
Edge cases/open decisions/minimum tests. Existing: X-002, CC-E05, CC-E06, CC-E16, CC-E19, CC-E20; O-011, O-022, O-041, O-044. Tests: choose one/none/defer; justified parallel choices; duplicate; caregiver-own service; disagreement; limited representative; withdrawal; option changes before release.
Community connection — prepare the referral or intake request#
When. After choice, applicable permission, and current destination requirements are established; repeat for corrected/missing material or a new destination without overwriting a released packet.
What and how much. Build one destination-specific minimum-necessary packet. Mark fields the service seeker or representative must personally provide or attest, especially personal, financial, household, and program-specific facts. Missing and not-applicable remain different.
Data/provenance. Use destination requirements, scoped permission, confirmed source-linked facts, exact documents, accessibility/channel needs, and personal-attestation fields. Prefill and document presence do not prove currentness, validity, eligibility, approval, or receipt.
Potential Proxi work. Populate allowed fields, attach permitted documents, flag stale/conflicting/missing items, ask targeted questions, and produce an accessible review. AI may propose candidate fields but cannot attest, sign, invent, or decide eligibility.
Human role. The service seeker or decision-specific authorized representative supplies and attests personal facts. Software enforces the approved disclosure rule. L1 pursues permitted missing administrative material. L2 enters only for relationship-sensitive clarification. Healthcare Legal Counsel receives only a legal-authority question that the approved rule cannot resolve.
Provisional human minutes (low / typical / high). L1 0 / 0 / 5 within the shared 13-minute batch exception; L2 0 / 0 / 2; separate U.S. L3/L4 0 / 0 / 0.
Completion evidence. Destination-specific current packet; source-linked facts; scope matches permission; proper actor completes attestation fields; unresolved items remain visible before release.
What does not prove completion. Generic template, AI-filled form, old application, every box populated, navigator attestation of personal facts, expired attachment, or preparation without permission.
Edge cases/open decisions/minimum tests. Existing: X-003, X-004, X-005, CC-E02, CC-E04, CC-E18, CC-E20, CC-E27, CC-E35; D-007, O-001, O-011, O-022. Tests: electronic clean; paper form; missing income; stale document; caregiver-own/mixed packet; duplicate; changed destination; correction after preparation.
Community connection — send the referral or make the introduction#
When. After the exact packet is ready, permission is active, endpoint is verified, and the V1 authorized human releases the exact content. Perform a live introduction only when the service seeker requests it or relationship context makes it valuable.
What and how much. Execute one selected route: transmit the packet, provide a permitted self-referral route, or make the agreed warm introduction. Record destination/service, content, channel, release actor, time, immediate result, and intentional parallel routes.
Data/provenance. Use exact selected option, endpoint, packet, permission, release event, modality, recipient, execution result, and duplicate-prevention facts. Release approval is not transmission; transmission is not receipt; introduction is not intake; self-referral instructions are not action.
Potential Proxi work. Verify prerequisites, present a one-click release, transmit, support a three-way call, record results, detect duplicate/failure, and open follow-up. Proxi cannot self-release PHI or claim downstream success.
Human role. The role authorized under the still-open community release decision approves the exact recipient/content/purpose/channel. L2 performs a requested warm introduction; it is not needed for routine transmission. L1 supports approved manual logistics.
Provisional human minutes (low / typical / high). L1 0 / 0 / 5 within the shared exception; L2 0 / 0 / 3 for a requested warm introduction; separate U.S. L3/L4 0 / 0 / 0. Additionally reserve 1 / 1 / 2 minutes for the as-yet-unassigned authorized V1 releaser rather than falsely assigning that authority to L1, L2, or a clinician.
Completion evidence. Selected route executed; exact permission and human release linked to recipient/content; objective send/introduction result; receipt remains separately open unless independently evidenced.
What does not prove completion. Packet preparation, release click, server/fax success alone, voicemail, self-referral instructions, warm introduction, or acknowledgement of another packet.
Edge cases/open decisions/minimum tests. Existing: CC-E01, CC-E02, CC-E04, CC-E16, CC-E18, CC-E20, CC-E26, CC-E35; D-007, O-001, O-004, O-011, O-022. Tests: electronic; fax/paper; self-referral; warm introduction; wrong endpoint; revocation; duplicate retry; parallel alternatives; failure; recipient change in flight.
Community connection — help complete intake#
When. After transmission or self-referral selection when outstanding intake steps exist, while the service seeker wants help and until Proxi-supportable steps finish, the organization takes over, or a barrier/decision routes elsewhere.
What and how much. Maintain one checklist per destination for appointments, forms, documents, attestations, interviews, technology, transportation, deadlines, and owners. Schedule within confirmed constraints and track each step without assuming Proxi can attest or decide intake.
Data/provenance. Use current organization instructions, referral receipt, choice, due dates, owners, objective step evidence, missing documents, scheduled events, responses, and barriers. Checklist completion is administrative readiness, not eligibility, intake completion, acceptance, or start.
Potential Proxi work. Build/update the checklist, prefill confirmed facts, schedule, remind, collect valid signatures, retrieve permitted documents, and flag overdue steps. AI may explain approved instructions but cannot sign, attest, or interpret clinical/financial disputes.
Human role. L1 handles phone/paper scheduling, unconnected offices, dynamic administrative requirements, and document pursuit. The proper service seeker/representative supplies attestations. L2 is reserved for relationship-sensitive refusal/confusion; the organization conducts and decides intake.
Provisional human minutes (low / typical / high). L1 0 / 0 / 13 shared across tasks 14-20 rather than per item; L2 0 / 0 / 3; separate U.S. L3/L4 0 / 0 / 0.
Completion evidence. Every Proxi-supportable step has an owner/status and objective completed, pending, or barrier result; attestations remain attributed; organization decisions remain open until issued.
What does not prove completion. Checklist, prefill, scheduled appointment, reminder, upload, interview attendance, or navigator assertion without organization evidence.
Edge cases/open decisions/minimum tests. Existing: CC-E07, CC-E08, CC-E18, CC-E19, CC-E20, CC-E26, CC-E27; O-001, O-004, O-011, O-022. Tests: online; phone-only; missing document; refusal to attest; changed form; no-show; inaccessible portal; no callback.
Community connection — address access barriers#
When. Whenever cost, transport, language, accessibility, technology, paperwork, schedule, geography, caregiver availability, cognition, or organizational process blocks option use, intake, attendance, communication, or service access.
What and how much. Treat each barrier separately. Present current factual remedies, owner/funder, and burdens; obtain the service seeker's choice; execute only the selected remedy; and verify whether it worked. Reassess when one remedy fails or creates another barrier.
Data/provenance. Use the service-seeker report, failed-step evidence, barrier source/time, remedy source/currentness, cost/funding responsibility, accommodation facts, preference, and objective result. A requested ride, interpreter, or funding lead is not a completed remedy.
Potential Proxi work. Detect stalled steps, match current remedies, show costs/unknowns, schedule approved logistics, translate approved content, remind, and track results. Proxi cannot choose a burden, invent funding, decide clinical safety, or close from a plan alone.
Human role. L1 implements approved administrative logistics and verifies dynamic facts. L2 handles distress, refusal, disagreement, or relationship-sensitive tradeoffs. Any interpreter or accessibility provider remains an outside support provider. Separate U.S. clinical staff enters only for an actual clinical-content question.
Provisional human minutes (low / typical / high). L1 0 / 5 / 13 within the shared batch episode; L2 0 / 3 / 8 for the barrier cluster; separate U.S. L3/L4 0 / 0 / 10 only for a triggered clinical/safety issue.
Completion evidence. Each barrier has source, selected remedy, owner/funder, performed action, and objective result; unresolved barriers and their effects stay open.
What does not prove completion. Barrier note, resource list, ride/interpreter request, funding application, rescheduling, or one resolved barrier hiding another.
Edge cases/open decisions/minimum tests. Existing: CC-E06, CC-E07, CC-E08, CC-E18, CC-E19, CC-E20, CC-E26, CC-E27; O-003, O-004, O-022. Tests: each barrier type; multiple barriers; declined remedy; clinical safety issue; remedy failure.
Community connection — confirm referral receipt#
When. After each transmission or self-contact attempt, using the channel's expected response, with follow-up beginning when no matched receipt exists.
What and how much. Establish whether the exact referral reached the correct organization and service, or whether the service seeker made the intended contact. Record received, rejected, wrong recipient/service, no response, or disputed receipt. Keep receipt separate from intake disposition.
Data/provenance. Use exact send/contact event, destination/service, channel/time, packet identity, acknowledgement source, and mismatch/rejection. Delivery success is not receipt; front-desk receipt is not intake review; a call attempt is not contact.
Potential Proxi work. Match electronic acknowledgements, detect mismatch/duplicates, issue approved reminders, and open manual pursuit. L1 may obtain administrative confirmation verbatim. AI cannot infer receipt or close intake.
Human role. No human is needed on matched electronic receipt. L1 pursues missing or disputed receipt. L2 handles service-seeker confusion or refusal. Software applies the approved wrong-recipient containment and incident rule; Healthcare Legal Counsel receives only a legal-authority question exposed by the incident that the approved rule cannot resolve.
Provisional human minutes (low / typical / high). L1 0 / 0 / 5 within the shared 13-minute episode; L2 0 / 0 / 2; separate U.S. L3/L4 0 / 0 / 0.
Completion evidence. Exact organization/service receipt or truthful nonreceipt/rejection/mismatch tied to the exact referral; no-response stays open; intake remains separately pending.
What does not prove completion. Send success, fax delivery, email acceptance, voicemail, reminder, another packet's receipt, scheduled intake, or service-seeker belief.
Edge cases/open decisions/minimum tests. Existing: CC-E02, CC-E03, CC-E04, CC-E16, CC-E18, CC-E20, CC-E26; O-004, O-022. Tests: matched receipt; delivery only; rejection; wrong service/person; no response; duplicate/late acknowledgement; self-referral; changed destination.
Community connection — confirm intake disposition#
When. After receipt/direct contact and whenever new organization or service-seeker evidence changes status, until a current disposition or truthful stalled/no-response state with owner exists.
What and how much. Preserve the externally sourced state: scheduled, completed/pending decision, accepted, waitlisted, denied, withdrawn, more information required, wrong program, stalled/no response, or verbatim other. Keep completed intake, acceptance, and service start distinct.
Data/provenance. Use organization response, service-seeker report, appointment/intake evidence, source/time, reason, action, and conflicts. Organization authority covers only its own process; service-seeker report does not substitute for its eligibility decision.
Potential Proxi work. Retrieve/normalize status, preserve both sources, detect overdue response, request clarification, and present conflicts. AI cannot decide eligibility, reinterpret denial, estimate an unsourced wait, or convert silence into closure.
Human role. L1 pursues missing administrative status. L2 handles the person's next choice, disagreement, or sensitive barrier. The external organization alone decides eligibility/intake status.
Provisional human minutes (low / typical / high). L1 0 / 0 / 5 within the shared episode; L2 0 / 0 / 3; separate U.S. L3/L4 0 / 0 / 0.
Completion evidence. Current destination-specific status, source/time, reason/action, and conflicts; no-response has an owner; service start remains separately unknown absent its own evidence.
What does not prove completion. Receipt, appointment, form completion, tentative staff comment, expectation, waitlist estimate, claim/authorization, or acceptance assumed to mean start.
Edge cases/open decisions/minimum tests. Existing: CC-E05, CC-E06, CC-E08, CC-E18, CC-E19, CC-E20, CC-E26, CC-E27; O-004, O-022, O-025. Tests: every listed disposition; no response; source conflict; accepted but no start.
Community connection — respond to denial, waitlist, or failed intake#
When. When the organization denies, waitlists, rejects, stalls, requests remediable information, or otherwise fails to advance intake; repeat if its reason changes or a late decision arrives.
What and how much. Preserve the exact result and distinguish eligibility decision, capacity/waitlist, missing information, administrative failure, withdrawal, and no response. Present only source-backed next routes: correct/resubmit, provide items, use an actual reconsideration route, accept waitlist with alternate/safeguard, pursue another option, pause, or stop. The service seeker chooses.
Data/provenance. Use exact disposition/reason/source/time, packet facts, criteria, missing-item request, published reconsideration route, waitlist facts, alternatives, preference, and triggered safety implications. One denial is not universal ineligibility; no response is not denial.
Potential Proxi work. Prepare corrections, retrieve alternatives, track waitlist/reconsideration, and summarize the stated reason. AI cannot invent appeal rights, override the organization, choose the next option, or promise reversal.
Human role. L2 explains the confirmed result and supports the next choice. L1 performs factual clarification and approved resubmission logistics. The Medicaid Program Specialist receives only a bounded Medicaid-program question; a separate U.S. clinician receives only a clinical-content question; and Healthcare Legal Counsel receives only a legal-authority question that the approved rule cannot resolve.
Provisional human minutes (low / typical / high). L1 0 / 0 / 5 within the shared episode; L2 0 / 4 / 10 for the triggered explanation/choice cluster; separate U.S. L3/L4 0 / 0 / 10 only for a triggered clinical/safety question.
Completion evidence. Exact result retained; accurate next options presented; service seeker chooses or declines; remediable work has an owner; waitlist/alternate/safeguard remains explicit; parent need is not falsely closed.
What does not prove completion. Denial note, waitlist placement, corrected submission, reconsideration request, alternate list, nonresponse, or truthful non-success labeled successful connection.
Edge cases/open decisions/minimum tests. Existing: CC-E05, CC-E06, CC-E07, CC-E18, CC-E19, CC-E20, CC-E26, CC-E27, CC-E34; O-003, O-004, O-022, O-025. Tests: eligibility denial; wrong program; remediable document; finite/indefinite wait; no response; accept wait; alternate; stop; urgency changes; late acceptance.
Community-connection and inventory detailed tasks 21-30#
Tasks 21-26 continue an individual connection episode. Tasks 27-30 are portfolio inventory operations; their minutes are per service-area review batch or organization/service record, not per beneficiary. An inventory candidate is never evidence of current availability, individualized eligibility, intake acceptance, or service delivery.
Referral-closure boundary#
Referral transmission, receipt, intake, acceptance, service start, and usefulness are separate milestones. A successful-start outcome requires evidence that the correct service actually began; a referral, appointment, authorization, claim, or intake acceptance does not establish that result. When the desired result includes practical usability or benefit, complete a proportionate first-use follow-up or state why that follow-up is not applicable.
If the service does not start, preserve the truthful denial, waitlist, nonresponse, barrier, withdrawal, or no-match result and return the next choice to the service seeker. The broader need ends only when the intended outcome is achieved, the person declines further help, the need is confirmed no longer current, or another named person or organization explicitly accepts continuing ownership. Attempt exhaustion and waitlist placement do not themselves close the need.
Community connection — confirm service start#
When. After an external acceptance, scheduled start, or expected-start date, and again when sources conflict or the expected start passes without evidence.
What and how much. Determine whether the service seeker actually received the selected service at least once or entered the organization's active service-delivery process. Record actual start date, service/location/modality, source, initial frequency if known, any immediate barrier, and whether ongoing delivery remains pending. Keep start separate from referral, receipt, intake, approval, authorization, billing, and usefulness.
Data/provenance. Use organization-issued start/attendance evidence, authenticated service-seeker report, dated encounter/delivery evidence when available, selected service identity, expected and actual dates, and source conflicts. An authorization, scheduled intake, accepted application, claim, invoice, transportation event, or staff expectation does not alone prove start.
Potential Proxi work. Monitor start-date evidence, match service/destination/person, ask a bounded confirmation question, detect overdue/conflicting status, and prepare manual outreach. AI cannot infer start from administrative progress or decide whether the service is clinically effective.
Human role. No human is required for matched objective start evidence. L1 verifies administrative status when disconnected. L2 confirms lived status when the service seeker's report conflicts or a relationship barrier appears. The external organization performs the service.
Provisional human minutes (low / typical / high). L1 0 / 0 / 5; L2 0 / 2 / 5; separate U.S. L3/L4 0 / 0 / 0.
Completion evidence. Dated person/service-specific start evidence, source, modality/location, and any remaining ongoing-delivery obligation; or a truthful not-started state routed to further action.
What does not prove completion. Referral, receipt, intake, acceptance, authorization, appointment, ride, claim, invoice, waitlist placement, or navigator expectation.
Edge cases/open decisions/minimum tests. Existing: CC-E05, CC-E06, CC-E07, CC-E08, CC-E18, CC-E20, CC-E26, CC-E27; O-004, O-022, O-025. Tests: verified start; accepted/no start; missed first date; organization says started/person disagrees; claim only; one visit then stop; wrong service; late start.
Community connection — check whether the service meets the need#
When. At a proportionate first-use follow-up or when the service seeker reports dissatisfaction, inability to use the service, a new barrier, or changed need. This is proposed value beyond the public minimum; cadence is not a CMS-prescribed interval.
What and how much. Ask the person receiving the service whether it is usable, acceptable, and addressing the authenticated desired result; identify what works, what does not, any burden or unmet need, and whether they want to continue, modify support, or pursue another route. Do not convert a satisfaction score into clinical effectiveness.
Data/provenance. Use the original desired result, actual start/use facts, service-seeker response in their own words, timing, utilization evidence if available, barriers, and any new clinical/safety signal. Caregiver report remains caregiver report; absence of complaint does not prove benefit.
Potential Proxi work. Deliver a bounded accessible follow-up, compare response to the desired result without deciding success, summarize source-faithfully, detect a stop cue, and prepare next options. AI cannot judge clinical effectiveness, dismiss dissatisfaction, or choose continuation.
Human role. L2 enters for a relationship-sensitive discussion, distress, conflicting reports, or requested help deciding the next nonclinical step. Separate U.S. L3/L4 enters only when symptoms, safety, treatment, or clinical appropriateness is raised.
Provisional human minutes (low / typical / high). L1 0 / 0 / 0; L2 0 / 3 / 8; separate U.S. L3/L4 0 / 0 / 10 only on trigger.
Completion evidence. Dated service-seeker outcome report tied to the desired result, source/conflicts, chosen next step, and any barrier/safety/alternate-service action.
What does not prove completion. Survey sent, survey score alone, no complaint, continued billing, organization outcome report without service-seeker input, or AI conclusion that the need is met.
Edge cases/open decisions/minimum tests. Existing: X-008, CC-E07, CC-E19, CC-E20, CC-E26, CC-E34; O-003, O-004, O-022. Proposed decision: which service types merit follow-up and at what event-based cadence. Tests: clearly helpful; partially helpful; unusable; stopped; beneficiary/caregiver reports conflict; new safety concern; no response; goal changed.
Community connection — pursue another option when needed#
When. When the chosen route is unavailable, unaffordable, inappropriate, declined, unsuccessful, closed, or no longer wanted and the service seeker still wants assistance.
What and how much. Open a linked alternate pursuit without forcing the person to repeat current facts. Carry forward source-linked desired result, constraints, permissions that remain applicable, failed-option reason, remedies tried, and lessons; refresh dynamic facts and repeat the necessary search/fit/choice/referral steps. Do not carry forward stale availability or destination-specific permission.
Data/provenance. Use the prior episode, exact failure/outcome, service-seeker continuation choice, current constraints, reusable versus destination-specific facts, alternate sources, and any changed urgency. One organization's failure does not prove no alternative exists.
Potential Proxi work. Reuse confirmed current inputs, exclude only source-proven mismatches, refresh inventory, show alternate options, and link prior barriers. AI cannot choose the replacement, relax hard constraints, or hide prior failure.
Human role. L1 verifies manual alternate facts. L2 supports changed tradeoffs and the next choice. Separate U.S. clinical staff enters only when alternate clinical suitability or safety must be decided.
Provisional human minutes (low / typical / high). L1 0 / 5 / 13 for an alternate-search administrative cluster; L2 0 / 3 / 8; separate U.S. L3/L4 0 / 0 / 10 only on trigger. Reused earlier work is not billed again as new human labor.
Completion evidence. Service seeker confirms continued pursuit; reusable facts and required refreshes are explicit; a new option/no-match result and next action are linked; the failed route remains truthfully preserved.
What does not prove completion. Generic alternate list, silent rerouting, stale copied packet, staff-selected replacement, failed option marked deleted, or new referral assumed connected.
Edge cases/open decisions/minimum tests. Existing: CC-E05, CC-E06, CC-E07, CC-E16, CC-E18, CC-E19, CC-E20, CC-E26, CC-E27; O-004, O-022, O-044. Tests: unavailable; unaffordable; person declines; changed constraint; no alternate; justified parallel route; stale permission; clinical concern; late acceptance from first option.
Community connection — explain the outcome#
When. After every material referral/intake/start outcome and before a person is left with a new action, wait, alternate, or truthful ending condition.
What and how much. Give one current plain-language summary: what was attempted; what each organization actually reported; the present state; what is not yet known; outstanding documents/barriers; who owns each next action; dates; and when Proxi will follow up. Distinguish successful start, pending, truthful non-success, and service-seeker decision.
Data/provenance. Use only confirmed episode events with source/time, current disposition, exact external reason, service-seeker choice, next owners/dates, and accessible communication needs. A generated summary does not prove it was received or understood.
Potential Proxi work. Assemble the status summary, render approved accessible language, deliver through the selected channel, record questions/corrections, and detect inconsistency. AI cannot soften a denial into pending, label connection without start evidence, or choose the next step.
Human role. Software may deliver a clean factual update. L2 enters when the outcome is sensitive, disappointing, disputed, confusing, or requires a new preference-sensitive decision. A separate U.S. clinician explains only actual clinical implications.
Provisional human minutes (low / typical / high). L1 0 / 0 / 0; L2 0 / 2 / 5; separate U.S. L3/L4 0 / 0 / 8 only on trigger.
Completion evidence. Current accessible summary delivered; source-backed state/unknowns/owners/dates; questions/corrections retained; and a next action or truthful endpoint that does not overstate connection.
What does not prove completion. Summary generated, message sent without delivery, jargon-heavy status, denial omitted, pending represented as complete, or no accountable next owner.
Edge cases/open decisions/minimum tests. Existing: X-007, CC-E05, CC-E06, CC-E19, CC-E20, CC-E26, CC-E34; O-004, O-022. Tests: successful start; waitlist; denial; no response; conflicting sources; inaccessible message; changed owner; sensitive clinical implication.
Community connection — update the care plan#
When. When a selected service, confirmed start, material barrier, responsibility, outcome, or continuing gap changes beneficiary care-plan content; caregiver-only service details enter beneficiary-facing content only when relevant and permitted.
What and how much. Add the service recipient, need/goal link, selected service, responsible organization/contact, cost/payer facts, participant responsibilities, actual connection state, unresolved barrier/gap, and follow-up. Preserve beneficiary-facing and caregiver-owned facts and permissions separately.
Data/provenance. Use confirmed episode facts, exact service recipient, choice, organization, start/outcome, responsibility, source/time, and permission. Referral sent or intake accepted does not justify recording a service as active. Clinical implications are not inferred from nonclinical service facts.
Potential Proxi work. Draft the relevant care-plan delta, identify conflicting/stale facts, show source links, and route only affected elements. AI cannot publish a service as active, decide material clinical impact, or expose caregiver-owned information.
Human role. L2 validates nonclinical coordination content, keeps caregiver-owned information separately attributed, and handles the beneficiary/caregiver relationship conversation. Software enforces the approved recipient-specific disclosure rule. A separate U.S. L3/L4 clinician reviews only triggered clinical implications. Healthcare Legal Counsel receives only a legal-authority question that the approved rule cannot resolve.
Provisional human minutes (low / typical / high). L1 0 / 0 / 0; L2 1 / 3 / 7; separate U.S. L3/L4 0 / 0 / 8 only on trigger.
Completion evidence. Reviewed current care-plan content accurately states service recipient, confirmed state, responsibilities, gaps, and follow-up with appropriate information boundaries.
What does not prove completion. Draft generated, referral copied as active service, caregiver service exposed without basis, old status retained, or clinical implication added by AI/L2.
Edge cases/open decisions/minimum tests. Existing: X-002, X-004, X-005, CC-E19, CC-E20, CC-E26, CC-E35; O-010, O-011, O-022, O-042. Tests: beneficiary service start; caregiver-own service; pending referral; failed service; changed provider; clinical implication; permission revoked; conflicting status.
Community connection — continue unresolved follow-up#
When. Whenever a need remains unresolved because receipt, intake, acceptance, start, barrier, alternate route, or continuing owner is pending; stop only on an evidenced ending condition.
What and how much. Maintain one next action, owner, due/expected date, last confirmed state, attempted channels, and fallback for every open obligation. Continue until the service seeker declines further help, the need is confirmed no longer current, service start/other intended outcome is evidenced, or another person/organization explicitly accepts continuing ownership. Retry count and timing remain provisional under O-004.
Data/provenance. Use exact open obligation, current source-backed status, attempts/responses, expected timing, service-seeker preference, owner acceptance, and changed risk. Nonresponse is neither refusal nor resolution; assignment is not owner acceptance; waitlist is not connection.
Potential Proxi work. Monitor due events, issue approved reminders, batch manual questions, escalate failed channels, preserve late responses, and show the next action. AI cannot choose when pursuit is futile, infer refusal, or close from retry exhaustion.
Human role. L1 performs batched administrative pursuit. L2 handles ongoing relationship, refusal, changed preference, dispute, or an unresolved service risk. Separate U.S. clinical staff enters only for a triggered clinical/safety issue; Puerto Rico never performs that work.
Provisional human minutes (low / typical / high). L1 0 / 5 / 13 per unresolved follow-up cluster; L2 0 / 3 / 8; separate U.S. L3/L4 0 / 0 / 10 only on trigger.
Completion evidence. Every obligation has a current evidenced state and owner; the episode ends only with confirmed intended outcome, authenticated decline, confirmed need resolution, or explicit accepted continuing ownership plus disclosed residual risk.
What does not prove completion. Reminder sent, attempts exhausted, waitlist, staff assignment, voicemail, referral closed locally, nonresponse, or safety handoff that leaves the community need open.
Edge cases/open decisions/minimum tests. Existing: X-006, X-007, X-014, X-015, CC-E03, CC-E06, CC-E19, CC-E20, CC-E26, CC-E34, CC-E36; O-004, O-022, O-025. Tests: prompt response; repeated no response; late response; service seeker declines; need resolves; owner accepts; owner assignment without acceptance; urgency changes; terminal late evidence.
Community inventory — identify needed service categories#
When. On a proposed monthly automated demand scan and quarterly nonclinical portfolio review, and when repeated no-match or failed referrals reveal a material category/geography/language/accessibility gap. CMS does not prescribe this cadence.
What and how much. Aggregate de-identified, nonduplicated service needs, no-match results, barriers, failed categories, service areas, languages, accessibility needs, and caregiver/beneficiary service populations. Produce a coverage-priority list without exposing individual information or treating demand frequency alone as importance.
Data/provenance. Use assessment/referral need categories, service seeker type, geography at an approved aggregation level, current inventory coverage, no-match/failure evidence, source period, deduplication, and suppression/privacy rules. A frequent category does not prove adequate demand measurement; one no-match does not prove a market gap.
Potential Proxi work. Aggregate, normalize categories, detect thin coverage, show source counts/time windows, preserve low-volume signals, and prepare a review queue. AI may propose category mapping but cannot set priorities or suppress rare high-impact needs.
Human role. The Puerto Rico GUIDE navigator (L2), performing the referral-coordination task, sets nonclinical inventory priorities after reviewing the evidence. L1 may resolve administrative data-quality exceptions. No clinical review is routine.
Provisional human minutes (low / typical / high). Per service-area review batch, L1 0 / 10 / 20; L2 5 / 10 / 20; separate U.S. L3/L4 0 / 0 / 0. These are portfolio minutes, not beneficiary minutes.
Completion evidence. Dated review period, categories and service-seeker populations, demand/coverage evidence, known blind spots, ranked human-approved priorities, and accountable next inventory action.
What does not prove completion. Raw assessment counts, AI category ranking, one anecdote, service directory size, zero recorded demand interpreted as no need, or PHI-rich case list.
Edge cases/open decisions/minimum tests. Existing: X-004, X-005, X-009, CC-E26, CC-E27; O-001, O-022, O-044. Proposed decision: aggregation/privacy threshold and inventory review cadence. Tests: common gap; rare critical gap; duplicate needs; caregiver versus beneficiary category; missing geography; stale period; no recorded demand; language/accessibility gap.
Community inventory — join or maintain a referral inventory system#
When. Before serving the geography and at proposed quarterly operational review, annual vendor/partnership renewal, material system change, data-source failure, or owner/freshness breach. CMS requires participation in or maintenance of an inventory system but does not prescribe Proxi's review intervals.
What and how much. Confirm the system covers assessed HRSN/service categories and service areas; has accountable owners, source/feed access, searchable fields, update/error processes, staff access, and usable referral endpoints; and records gaps. Joining a vendor is not enough if content is unusable or unmaintained.
Data/provenance. Use inventory/vendor identity, covered categories/geographies, source agreements, access tests, last updates, ownership, error/correction records, referral linkage, downtime, and gap evidence. Contract signature does not prove coverage, current data, operational access, or successful referrals.
Potential Proxi work. Test feeds/search/access, compare expected versus actual coverage, flag stale/unowned fields, compile error/failure evidence, and prepare renewal/correction work. AI cannot decide contractual sufficiency or certify CMS compliance.
Human role. The GUIDE Participant accountable authority performs the exact retained organizational act of selecting and accepting the inventory system. L1 supports access and data checks. L2 validates nonclinical referral usability. Software enforces the approved routing, privacy, and disclosure rules. Healthcare Legal Counsel receives only a legal-authority question that the approved rule cannot resolve; no L3/L4 clinical labor is routine.
Provisional human minutes (low / typical / high). Per system-review event, L1 0 / 10 / 20; L2 0 / 5 / 15; separate U.S. L3/L4 0 / 0 / 0; the retained GUIDE Participant selection/acceptance act and any triggered Healthcare Legal Counsel legal-authority question remain outside the ladder at 15 / 30 / 60.
Completion evidence. Current system/access/owner record, tested coverage, update/error process, gap list and remediation owner, the GUIDE Participant accountable-authority disposition when a retained organizational act is required, and any bounded Healthcare Legal Counsel disposition on a legal-authority question that the approved rule cannot resolve.
What does not prove completion. Vendor contract, login screen, directory count, old data feed, demo search, staff statement, or one successful referral.
Edge cases/open decisions/minimum tests. Existing: X-005, X-013, CC-E18, CC-E27; O-022, O-PA-001. Tests: usable maintained system; inaccessible feed; no owner; partial geography; missing category; duplicate sources; vendor outage; termination/migration; contract exists/data stale.
Community inventory — discover candidate organizations#
When. When a priority category/geography lacks adequate options, a need returns no match, an organization closes, a new service source becomes available, or a scheduled inventory enrichment batch begins.
What and how much. Find every serious candidate organization/service in the searched scope using authoritative/local sources. Record organization identity separately from service/program identity, source and observed date, apparent geography/population/category, and unanswered verification questions. Preserve low-information candidates as hypotheses rather than dropping them.
Data/provenance. Use official directories, government/aging-network sources, organization pages, partner knowledge, referral-system sources, prior response history, and source dates. Discovery proves candidate existence only, not operation, eligibility, availability, quality, acceptance, or endorsement.
Potential Proxi work. Search sources in parallel, deduplicate identities without discarding distinct services, extract candidate facts, retain provenance, flag conflicts/staleness, and queue direct verification. AI entity/category mapping remains candidate data until grounded.
Human role. No human is needed for routine connected discovery. L1 checks local or manual sources and resolves administrative identity conflicts. L2 contributes local referral knowledge or resolves material relationship-sensitive ambiguity; clinical staff is not required.
Provisional human minutes (low / typical / high). Per discovery batch, L1 0 / 10 / 25; L2 0 / 5 / 15; separate U.S. L3/L4 0 / 0 / 0.
Completion evidence. Scope/date/source-labeled candidate set including low-information hypotheses, organization/service identity distinctions, duplicates linked, conflicts/unknowns retained, and verification queue.
What does not prove completion. Search-engine top results, sponsored list, one directory, organization name without service, deduplication that deletes distinct programs, or candidate presented as available.
Edge cases/open decisions/minimum tests. Existing: X-001, X-004, X-005, X-009, CC-E18, CC-E20, CC-E27; O-001, O-022, O-044. Tests: many sources; same organization/multiple services; aliases; closed organization; rural/no result; Tribal route; low-information lead; sponsored entry; conflicting geography.
Community inventory — capture useful service details#
When. On candidate creation, when a current source supplies new information, before consequential matching if a required field is stale/unknown, and after direct verification or correction.
What and how much. Maintain one source-linked record per organization/service for description, population, geography, language, accessibility, contact/intake routes, hours, cost/funding, published eligibility, documents, waitlist/capacity meanings, transportation, and verification date. Keep observed, directly confirmed, conflicted, stale, and unknown facts distinct.
Data/provenance. Use field-level source, exact observed/confirmed date, service identity, respondent role when direct, and correction history. Website text, third-party directory, and direct staff response retain separate provenance. A populated field does not prove truth/currentness; blank is unknown.
Potential Proxi work. Extract and normalize fields, compare sources, preserve original wording, detect change/conflict, and prepare targeted verification. AI cannot merge conflicting facts into a confident value or infer eligibility/availability from marketing language.
Human role. No person is needed for clean current extraction. L1 verifies material administrative facts or resolves record identity. L2 handles consequential relationship-sensitive ambiguity or local referral context. Separate U.S. clinical staff does not review routine community-resource data.
Provisional human minutes (low / typical / high). Per organization/service record, L1 0 / 3 / 8; L2 0 / 0 / 5; separate U.S. L3/L4 0 / 0 / 0.
Completion evidence. Service-specific record with field-level provenance/currentness, explicit unknown/conflicts, organization-versus-service identity, required verification questions, and correction history.
What does not prove completion. Scraped page, all fields populated, organization-level facts copied to every service, marketing claims, stale direct response, or AI-resolved conflict.
Edge cases/open decisions/minimum tests. Existing: X-001, X-003, X-004, X-005, X-009, CC-E18, CC-E20, CC-E27; O-001, O-022, O-044. Tests: clean official source; third-party conflict; service-level differences; no price; ambiguous service area; changed phone; direct correction; duplicate source; blank versus not applicable.
Community-inventory detailed tasks 31-36#
These are portfolio operations except the final resource-sharing task. Portfolio minutes are charged once per organization/service record, service-area batch, or scheduled review event as stated; they are never multiplied by the beneficiary population. The clean path is software-led. Human work is reserved for facts that cannot be verified electronically, consequential conflicts, relationship ownership, or approval decisions.
Community inventory — verify an organization directly#
When. When a material fact required for matching or referral is absent, stale, conflicted, or unavailable from a current authoritative source; after repeated referral failure; and before relying on a high-consequence or frequently changing fact. Do not call merely to reconfirm every populated field.
What and how much. Ask a short service-specific checklist covering only unresolved material facts: operating status; exact program; population/geography; inquiry, application, and opening status; contact/intake route; cost/funding; documents; accessibility/language; and meaning/date of any capacity statement. Attribute every answer to the organization respondent and time. The organization supplies facts about its service; it does not decide whether Proxi should recommend it or whether a particular person is eligible unless it completes its own intake.
Data/provenance. Use the organization/service identity, prior field-level sources, unresolved questions, verified endpoint, respondent name or role when available, exact answer, contact time, and correction history. A front-desk answer is evidence only for what that person actually stated; it does not prove individualized eligibility, acceptance, quality, service start, or continuing availability.
Potential Proxi work. Generate the smallest verification checklist, place or support the call, transcribe source-faithfully, compare answers with prior sources, flag contradictions, and update only supported fields. AI cannot convert tentative language into fact, infer eligibility, or silently choose one conflicting source.
Human role. L1 performs administrative verification and records the outside organization's answers verbatim. L2 enters only for a consequential relationship issue or disputed service meaning that cannot be resolved administratively. The outside organization remains the source of its own service facts. Separate U.S. clinical staff is not routine.
Provisional human minutes (low / typical / high). Per organization/service verification event, L1 0 / 5 / 12; L2 0 / 0 / 5; separate U.S. L3/L4 clinical 0 / 0 / 0. Several questions for the same organization should be batched into one contact.
Completion evidence. Dated service-specific response, verified respondent/endpoint, exact questions and answers, supported field updates, retained conflicts/unknowns, and the next recheck basis.
What does not prove completion. Dial attempt, voicemail, generic “we are open,” old staff answer, organization-level answer copied to every program, AI summary without source text, or a direct answer treated as beneficiary eligibility or service start.
Edge cases/open decisions/minimum tests. Existing: X-001, X-004, X-005, X-006, X-009, CC-E18, CC-E27; O-001, O-004, O-022, O-044. Tests: complete current electronic source requires no call; phone-only service; no answer; respondent unsure; two respondents conflict; organization has multiple programs; changed price; tentative waitlist statement; wrong organization; later correction.
Community inventory — confirm service-area relevance#
When. On candidate creation, before presenting a consequential option when geography is unclear, and whenever service area, beneficiary community, delivery modality, or program boundaries change.
What and how much. Determine the service-specific coverage rule using the smallest useful geography: address/ZIP/county/municipality, catchment, travel radius, in-home route, virtual availability, Tribal eligibility/geography, or other published boundary. Preserve conditional and partial coverage rather than flattening it to yes/no. This task concerns the inventory service area, not individualized program eligibility.
Data/provenance. Use the exact organization/service, official coverage statement or direct response, geography unit, delivery modality, exceptions, source/date, and GUIDE service area. A nearby office, statewide organization, virtual label, or prior referral does not prove service to the current community.
Potential Proxi work. Normalize supported geographies, compare them with the GUIDE footprint, show partial/unknown areas, and prepare a targeted verification question. Proxi cannot infer coverage from distance, marketing territory, or organization headquarters.
Human role. No human is needed for an exact current boundary. L1 verifies ambiguous administrative boundaries with the outside organization. L2 handles local or relationship context only when source facts do not answer the referral question. No clinical role is routine.
Provisional human minutes (low / typical / high). Per ambiguous organization/service record, L1 0 / 3 / 8; L2 0 / 0 / 4; separate U.S. L3/L4 clinical 0 / 0 / 0.
Completion evidence. Service-specific dated coverage statement, geography/modality, exceptions, comparison with the GUIDE footprint, and explicit partial or unknown areas.
What does not prove completion. Headquarters address, map distance, statewide brand, national helpline, virtual-service label, one prior beneficiary served, or an organization record with no program-specific boundary.
Edge cases/open decisions/minimum tests. Existing: X-004, X-005, CC-E20, CC-E27; O-022, O-044. Tests: exact ZIP list; county boundary; border community; in-home radius; virtual only in selected states; Tribal route; temporary location; partial service area; boundary changed; no published geography.
Community inventory — correct changed or inaccurate information#
When. Whenever an authoritative or directly confirmed source contradicts the current inventory, a referral exposes an error, an organization submits a correction, or a previously unknown fact becomes known.
What and how much. Preserve the old field and source, add the new source and effective/observed time, determine which future searches and open referrals are affected, and correct only the supported organization/service fields. Mark closed or unusable options promptly without deleting history. Do not retroactively rewrite what users were shown.
Data/provenance. Use old/new values, sources, observed/effective times, exact service record, reporter/respondent, impacted option lists/referrals, and conflict status. A changed web page may be a correction, a new service variant, or an error; it is not automatically authoritative.
Potential Proxi work. Detect differences, link the correction to affected records, suppress known-invalid future presentation, queue material open-case review, and retain history. AI cannot decide a source conflict or declare a program closed from ambiguous language.
Human role. L1 verifies administrative changes with the outside organization and applies source-supported corrections. L2 handles consequential referral-relationship changes and person-facing follow-up. Separate U.S. clinical staff enters only if an open beneficiary case raises a distinct clinical or safety question.
Provisional human minutes (low / typical / high). Per corrected organization/service record, L1 0 / 3 / 8; L2 0 / 0 / 5; separate U.S. L3/L4 clinical 0 / 0 / 0 for portfolio correction. Beneficiary-case follow-up is charged to that case, not this inventory row.
Completion evidence. Immutable prior value, supported current value or explicit conflict, source/effective time, impacted-record review, suppression/correction of future presentation, and correction notice where an open referral was materially affected.
What does not prove completion. Overwriting the field, deleting the organization, updating only one duplicate listing, treating “temporarily unavailable” as permanently closed, or correcting the inventory without reviewing affected open referrals.
Edge cases/open decisions/minimum tests. Existing: X-004, X-005, X-011, CC-E20, CC-E26, CC-E27, CC-E36; O-022, O-044. Tests: phone correction; official page correction; two sources disagree; temporary closure; new price; same organization/different program; duplicate listing; open referral affected; late correction after episode closure.
Community inventory — recheck entries regularly#
When. On a risk-based schedule and before high-consequence reliance when freshness has expired. CMS does not specify a cadence. Frequency should reflect use, volatility, referral failures, source reliability, and consequence of error rather than one universal interval.
What and how much. Revisit the due portfolio slice, refresh electronically available fields first, batch unresolved direct-verification questions by organization, and classify each field as confirmed current, changed, conflicted, unavailable, or still unknown. High-use/change-prone entries receive more attention; low-use entries are retained with transparent staleness rather than dropped.
Data/provenance. Use last field verification, source reliability, use/referral/failure history, change history, due basis, current sources, and recheck results. A recent page fetch does not prove every extracted field is current, and no observed change is not direct confirmation.
Potential Proxi work. Schedule risk-based rechecks, refresh connected sources, compare deltas, batch manual work, mark stale facts, and show overdue coverage. AI may prioritize hypotheses but cannot certify freshness or delete low-information entries.
Human role. L1 performs batched manual checks. L2 reviews consequential referral conflicts, repeated failures, and unresolved local knowledge. The GUIDE Participant accountable authority approves the retained cadence policy, which software then enforces. Clinical staff is not routine.
Provisional human minutes (low / typical / high). Per scheduled service-area recheck batch, L1 5 / 15 / 35; L2 0 / 5 / 15; separate U.S. L3/L4 clinical 0 / 0 / 0. These are portfolio minutes and must not be charged per beneficiary or per unchanged entry.
Completion evidence. Dated portfolio denominator, fields/sources actually rechecked, due/overdue items, changed/conflicted/unknown results, batched manual outcomes, and next review basis.
What does not prove completion. Job ran, page downloaded, every record date mass-updated, one successful call, unchanged website, or stale entries silently removed from the denominator.
Edge cases/open decisions/minimum tests. Existing: X-005, X-006, X-013, X-014, CC-E18, CC-E27; O-004, O-022, O-044. Proposed decision under O-022: field-specific freshness and risk-based recheck policy. Tests: stable official feed; volatile waitlist; high-use phone-only service; outage; overdue batch; low-use rural record; conflicting update; no accountable owner.
Community inventory — identify coverage gaps#
When. At the portfolio review, after truthful no-match results, after repeated referral failures, and when new population, geography, language, accessibility, cost, or service-category needs appear.
What and how much. Compare current usable inventory—not raw listing count—with de-identified observed needs and hard constraints. Identify no-option and weak-option gaps by category, service area, population, modality, language, accessibility, affordability, and capacity. Preserve low-volume/high-impact gaps and distinguish missing data from confirmed absence.
Data/provenance. Use current inventory usability, demand/no-match/failure evidence, service-seeker type, coverage geography, source period, known data blind spots, and suppression/privacy rules. Few recorded needs may reflect under-screening; many listings may represent one unusable service.
Potential Proxi work. Compare supply with demand, collapse true duplicates without dropping distinct programs, map confirmed/possible gaps, show uncertainty, and prepare outreach priorities. AI cannot decide community priority, infer absence from missing data, or suppress rare needs.
Human role. The Puerto Rico GUIDE navigator (L2), performing the referral-coordination task, validates the practical gap and sets relationship-development priorities. L1 resolves administrative data-quality questions. The GUIDE Participant accountable authority performs any retained portfolio escalation decision. No clinical review is routine.
Provisional human minutes (low / typical / high). Per service-area coverage review, L1 0 / 5 / 15; L2 5 / 10 / 20; separate U.S. L3/L4 clinical 0 / 0 / 0. Do not charge this portfolio analysis to individual referral episodes.
Completion evidence. Dated need/inventory denominators, confirmed versus possible gaps, affected categories/populations/geographies, data limitations, human-approved priorities, and accountable next action.
What does not prove completion. Directory count, no-match anecdote, AI heat map, zero recorded requests, raw ZIP coverage, or a single organization that is inaccessible, unaffordable, or not accepting people.
Edge cases/open decisions/minimum tests. Existing: X-003, X-004, X-005, X-009, CC-E26, CC-E27; O-022, O-044. Tests: confirmed rural gap; language gap; accessibility gap; unaffordable-only options; many duplicate listings; unknown capacity; rare high-impact need; under-screened geography; temporary versus persistent gap.
AAA or Tribal Aging Program detailed tasks 37-45#
Tasks 37-42 and 45 are Participant-level route, relationship, and operating-process work. Their labor is per decision, organization, agreement, process-design, or maintenance event—not per beneficiary. Task 43 is an individual referral episode, and task 44 may be either an individual failure or a portfolio pattern. A written agreement is one optional way to satisfy §6.1; it does not itself prove that an individual was assisted or connected.
AAA or Tribal Aging Program route — choose how the Participant will meet §6.1#
When. Before service launch in each covered geography and whenever internal capacity, aging-network coverage, agreement status, Participant scope, or CMS/Participation Agreement requirements materially change.
What and how much. Choose and document one supported operating route: direct Participant referral and connection, a written local AAA/Tribal Aging Program agreement requiring assistance, or both. Compare geographic/population coverage, actual operating capacity, relationship ownership, referral evidence, failure recovery, and continuity. Do not treat “and/or” as permission to have no functioning route.
Data/provenance. Use current CMS §6.1, the executed Participation Agreement when obtained, Participant service area/population, internal staffing/capacity, official AAA/Title VI mapping, existing agreements, referral performance, risks, and any applicable Healthcare Legal Counsel disposition on a legal-authority question that the approved rule could not resolve. A vendor relationship or informal contact does not prove the selected route satisfies §6.1.
Potential Proxi work. Assemble the decision packet, map coverage/capacity, compare route responsibilities, expose missing evidence, and preserve the approved decision and review triggers. AI cannot select the Participant's organizational route under §6.1 or certify sufficiency.
Human role. The GUIDE Participant accountable authority performs the exact retained organizational act of choosing the §6.1 route. Software applies the approved requirement and route rules. L2 contributes referral-operating evidence but does not make the retained organizational decision. Healthcare Legal Counsel receives only a legal-authority question that the approved rule cannot resolve. No L1 or clinical labor is routine.
Provisional human minutes (low / typical / high). Per Participant route-decision event, L1 0 / 0 / 0; L2 0 / 10 / 25; separate U.S. L3/L4 clinical 0 / 0 / 0; the retained GUIDE Participant decision and any triggered Healthcare Legal Counsel legal-authority question remain outside the ladder at 20 / 40 / 90. These are portfolio minutes.
Completion evidence. Dated approved route by geography/population, decision owner, supporting capacity/coverage facts, agreement requirement where selected, operational next steps, review triggers, and explicit gaps/conditions.
What does not prove completion. RFA citation, directory link, unsigned draft, informal introduction, prior agreement outside scope, internal intention, or leadership email with no operating route.
Edge cases/open decisions/minimum tests. Existing: X-004, X-005, X-010, CC-E27; O-001, O-022, O-PA-001. Tests: direct only with capacity; agreement only with full coverage; both; no local counterpart; Tribal population; partial geography; agreement expired; internal capacity loss; PA changes interpretation.
AAA or Tribal Aging Program route — identify the correct organization#
When. After choosing an agreement or partnership route, before outreach or agreement drafting, and when service geography, population, jurisdiction, or official aging-network assignments change.
What and how much. Map each covered community/population to the official local Area Agency on Aging or applicable Title VI Tribal Aging Program. Preserve overlapping, disputed, or conditional jurisdictions and identify the counterpart with authority to discuss operations. Do not select an organization merely because its office is closest.
Data/provenance. Use official aging-network/Title VI sources, exact geography/population, program identity, service/jurisdiction statement, current contact, source/date, and any official confirmation. A general aging organization, ADRC, contractor, state unit, or nearby AAA is not automatically the §6.1 counterpart.
Potential Proxi work. Search official sources, map service areas, distinguish organizations/programs, retain uncertainty, and prepare a confirmation request. AI cannot resolve jurisdictional conflict or confer authority.
Human role. L1 gathers and verifies official administrative facts. L2 confirms local referral-operating fit and resolves ambiguous mapping with the outside organizations. Healthcare Legal Counsel receives only a legal-authority question that the approved rule cannot resolve. No clinical role is routine.
Provisional human minutes (low / typical / high). Per service-area mapping event, L1 0 / 5 / 12; L2 0 / 3 / 8; separate U.S. L3/L4 clinical 0 / 0 / 0; an outside-organization response and any triggered Healthcare Legal Counsel legal-authority question remain outside the ladder at 5 / 10 / 20.
Completion evidence. Official source-linked organization/program mapping, covered geography/population, verified operating and authorized contacts, unresolved overlaps, and review date.
What does not prove completion. Google result, nearest office, statewide aging agency, ADRC listing, contractor relationship, old agreement party, or organization self-description without jurisdiction confirmation.
Edge cases/open decisions/minimum tests. Existing: X-001, X-004, X-005, CC-E20, CC-E27; O-022. Tests: one clear AAA; overlapping counties; Tribal Title VI route; state boundary; beneficiary population outside Tribal eligibility; organization merger; changed contact; two organizations dispute jurisdiction.
AAA or Tribal Aging Program route — establish the relationship#
When. After identifying the correct organization and before relying on it as a §6.1 route; repeat when counterpart ownership changes or the relationship becomes inactive.
What and how much. Contact the organization, explain the Participant and proposed GUIDE need, learn its actual capabilities and limits, identify operational and authorized decision contacts, agree on the next design step, and record unanswered questions. Relationship establishment is discovery and mutual engagement, not agreement execution.
Data/provenance. Use organization/program mapping, Participant overview, proposed population/geography, contact attempts, meeting participants/roles, capability statements, constraints, authority questions, decisions, and next owners/dates. Attendance does not prove organizational commitment or signatory authority.
Potential Proxi work. Prepare the brief, outreach, agenda, capability questions, notes, decision/action summary, and follow-up tracking. AI cannot represent tentative discussion as commitment or decide partnership terms.
Human role. L2 owns the nonclinical referral relationship with the outside AAA or Tribal Aging Program counterpart and supplies practical referral context. L1 schedules and prepares. The GUIDE Participant accountable authority performs only a retained organizational commitment or approval act. Software enforces approved disclosure rules; Healthcare Legal Counsel receives only a legal-authority question that the approved rule cannot resolve.
Provisional human minutes (low / typical / high). Per relationship-establishment event, L1 0 / 5 / 10; L2 0 / 5 / 15; separate U.S. L3/L4 clinical 0 / 0 / 0; retained GUIDE Participant acts, outside-counterpart time, and any triggered Healthcare Legal Counsel legal-authority question remain outside the ladder at 20 / 35 / 75.
Completion evidence. Verified counterpart identities/roles, dated discussion, source-faithful capability/constraint record, mutual next step, assigned owners/dates, and explicit uncommitted terms.
What does not prove completion. Outreach sent, voicemail, meeting scheduled, generic capability deck, friendly conversation, staff attendance, or one person's interest treated as organizational authorization.
Edge cases/open decisions/minimum tests. Existing: X-001, X-006, X-014, CC-E27; O-004, O-022. Tests: responsive counterpart; no response; wrong contact; no authority; capability mismatch; partial geography; organization interested but cannot commit; leadership changes; protected-information question arises.
AAA or Tribal Aging Program route — define the assistance to be provided#
When. Once both organizations understand the proposed route and before drafting final agreement language or process instructions; revisit when services, responsibilities, evidence, or failures show ambiguity.
What and how much. Define the assistance the aging-network partner will provide: populations/geographies, needs/services, referral entry, information required, beneficiary/caregiver contact, navigation/intake help, status return, follow-up, inability/decline handling, escalation, and evidence of actual assistance. Assign each responsibility without promising eligibility, service availability, admission, or delivery the partner has not authorized.
Data/provenance. Use CMS §6.1, agreed capabilities/limits, Participant scope, service seeker/data-owner boundaries, operational capacity, current contacts, downstream organization dependencies, proposed status/evidence terms, and each party's approved statements. A responsibility draft is not acceptance.
Potential Proxi work. Build a responsibility matrix, expose gaps/overlaps, draft precise operational language, distinguish handoff from assistance/connection, and track unresolved terms. AI cannot assign responsibility or approve legal/operational sufficiency.
Human role. L2 performs the nonclinical referral-coordination work with the outside AAA or Tribal Aging Program operational counterpart and defines the proposed operating responsibilities. The GUIDE Participant accountable authority performs only the retained organizational approval act. Software checks approved requirements and disclosure conditions. Healthcare Legal Counsel receives only a legal-authority question that the approved rule cannot resolve. L1 prepares supporting facts only.
Provisional human minutes (low / typical / high). Per arrangement-design event, L1 0 / 5 / 10; L2 5 / 15 / 30; separate U.S. L3/L4 clinical 0 / 0 / 0; retained GUIDE Participant acts, outside-counterpart time, and any triggered Healthcare Legal Counsel legal-authority question remain outside the ladder at 30 / 60 / 120.
Completion evidence. Mutually approved responsibility matrix covering scope, entry, information, assistance, status, failure, evidence, and owners; unresolved terms remain explicit and block finalization where material.
What does not prove completion. Generic “coordinate services,” one-sided draft, list of partner services, referral email address, agreement boilerplate, or partner acceptance without operational responsibilities.
Edge cases/open decisions/minimum tests. Existing: X-002, X-004, X-010, X-015, CC-E26, CC-E27; O-011, O-022. Proposed decision under O-022: minimum agreement-route assistance and returned-status evidence. Tests: full assistance; partner only supplies list; partner contacts caregiver only; no status return; eligibility denial; downstream service unavailable; scope changes; party disputes ownership.
AAA or Tribal Aging Program route — execute a written agreement when using that option#
When. After operational terms are approved and before representing the written-agreement route as active; renew or amend before expiry or material scope/party/responsibility changes.
What and how much. Convert approved terms into a written agreement naming the correct parties, covered geography/population, required assistance, responsibilities, information/privacy terms, contacts, operating process, status/evidence, effective/expiry/termination terms, change process, and signatures. No individual referral may rely on this route while the agreement is absent, expired, unsigned, or outside scope.
Data/provenance. Use the approved responsibility matrix, legal entity/signatory authority, current CMS and executed PA requirements, privacy/security terms, dates, attachments, signature events, and superseded agreements. A draft or one signature is not execution.
Potential Proxi work. Populate approved terms, compare draft to the operational matrix, flag missing/inconsistent provisions, route review/signature, preserve the executed artifact, and alert on renewal/change. AI cannot approve legal terms, verify signatory authority by itself, or activate the route.
Human role. The exact authorized GUIDE Participant signatory and authorized outside AAA or Tribal Aging Program signatory execute the agreement as retained and outside authority acts. L2 confirms that the written referral process matches the approved operating design but has no signature authority by default. Software checks approved required terms and routing conditions. Healthcare Legal Counsel receives only a legal-authority question that the approved rule cannot resolve. L1 and clinical roles have no signature authority by default.
Provisional human minutes (low / typical / high). Per agreement execution event, L1 0 / 0 / 0; L2 0 / 0 / 0; separate U.S. L3/L4 clinical 0 / 0 / 0; authorized signatory acts and any triggered Healthcare Legal Counsel legal-authority question remain outside the ladder at 30 / 90 / 240.
Completion evidence. Fully executed current agreement, verified parties/signatories, scope, effective/expiry dates, required assistance and operational attachments, repository/access location, renewal owner, and link to the working process.
What does not prove completion. Draft, template, unsigned PDF, email assent, one signature, expired agreement, agreement for another geography, or contract with no required assistance.
Edge cases/open decisions/minimum tests. Existing: X-004, X-005, X-011, X-015, CC-E27, CC-E35; O-022, O-PA-001. Tests: clean execution; wrong legal entity; unauthorized signer; missing attachment; future effective date; expiry; amendment; termination; scope outside geography; PA requirement conflict.
AAA or Tribal Aging Program route — establish the working referral process#
When. After the arrangement is operationally approved and, for the agreement option, executed; before the first individual referral; retest after material process, contact, channel, or responsibility change.
What and how much. Turn the arrangement into staff-usable instructions: who may refer; how the service seeker chooses; information/permission required; exact channel/endpoint; packet; receipt/status meanings; partner contact/assistance expectations; downstream intake/start evidence; retry/failure route; and escalation contacts. Run at least one non-PHI test or approved simulation before activation.
Data/provenance. Use the current arrangement/agreement, operating contacts, endpoint/channel capabilities, content requirements, release authority, response/status definitions, SLAs if approved, test results, and training acknowledgement. A process document does not prove the endpoint or return path works.
Potential Proxi work. Draft instructions/checklists/materials, validate fields and endpoints, simulate handoff/return statuses, identify contradictions, and publish only the approved current procedure. AI cannot approve the release matrix or declare readiness from a document alone.
Human role. L2 and the outside AAA or Tribal Aging Program operational counterpart define and validate the nonclinical referral workflow. The GUIDE Participant accountable authority performs only a retained workflow-approval act. L1 tests administrative mechanics. Software enforces the approved disclosure, release, routing, and failure-recovery conditions. Healthcare Legal Counsel receives only a legal-authority question that the approved rule cannot resolve. No clinical role is routine.
Provisional human minutes (low / typical / high). Per process-design/test event, L1 5 / 15 / 30; L2 10 / 25 / 45; separate U.S. L3/L4 clinical 0 / 0 / 0; retained GUIDE Participant acts, outside-counterpart time, and any triggered Healthcare Legal Counsel legal-authority question remain outside the ladder at 20 / 45 / 90.
Completion evidence. Approved current instructions, roles, exact packet/release/endpoint, atomic status meanings, tested handoff and return path, failure/escalation route, owners, and staff-access/training evidence.
What does not prove completion. Agreement signature, SOP draft, email address, successful test send without matched return, staff training without endpoint test, or handoff state represented as service connection.
Edge cases/open decisions/minimum tests. Existing: X-002, X-007, X-013, X-014, CC-E02, CC-E03, CC-E26, CC-E27, CC-E35; O-001, O-004, O-011, O-022. Tests: electronic/manual route; wrong endpoint; revocation; partner rejects packet; no returned status; downtime; duplicate send; process version changes; referral outside scope.
AAA or Tribal Aging Program route — send and follow an individual referral#
When. After the service seeker selects the aging-network route, applicable permission and current route requirements are met, and the exact packet is authorized for release. Follow until assistance/connection evidence or a truthful route-specific non-success result exists.
What and how much. Prepare and release the minimum permitted referral, confirm matched receipt, learn whether the AAA/Tribal program accepted the handoff, track the assistance it actually provided, and continue through downstream intake/acceptance/start when the intended result requires it. Preserve referral sent, received, partner disposition, assistance, downstream intake, acceptance, service start, cessation, outcome, and closure as distinct evidence.
Data/provenance. Use service seeker/recipient, choice, permission/content owner, exact packet/version, active arrangement scope, endpoint, release actor/time, receipt, partner responses/actions, downstream organization/service events, barriers, and service-seeker outcome. An agreement, sent packet, receipt, or accepted handoff does not prove assistance or service connection. When the caregiver is the service seeker, the caregiver's service facts and outcome remain caregiver-owned and do not automatically enter the beneficiary record, care plan, or PCP-shared material without a separate permitted basis.
Potential Proxi work. Build the route-specific packet, support the human release, transmit, match acknowledgements, remind/pursue, normalize source-faithful statuses, expose missing evidence, and prepare alternate routes. AI cannot release PHI, infer downstream success, or close from retry exhaustion.
Human role. L2 owns the nonclinical relationship and route follow-through; L1 handles administrative pursuit and manual mechanics; the outside AAA or Tribal Aging Program coordinator performs and reports its assistance; outside downstream organizations decide intake, eligibility, and delivery. Separate U.S. clinical staff enters only for a triggered clinical or safety question.
Provisional human minutes (low / typical / high). Per individual agreement-route episode, L1 0 / 5 / 13; L2 2 / 6 / 12; separate U.S. L3/L4 clinical 0 / 0 / 10 only on trigger; authorized V1 releaser 1 / 1 / 2 until the release role is settled. Shared pursuit time must not be charged again at every status row.
Completion evidence. Exact referral/release and receipt, partner disposition, source-backed assistance performed, downstream intake/acceptance/start evidence when applicable, service-seeker outcome, open obligations, and route-specific closure or truthful non-success result.
What does not prove completion. Current agreement, packet prepared, sent, fax success, receipt, partner says “accepted,” resource list, scheduled intake, authorization, claim, or narrative “connected” without start/assistance evidence.
Edge cases/open decisions/minimum tests. Existing: X-002, X-006, X-007, X-015, CC-E03, CC-E05, CC-E06, CC-E18, CC-E19, CC-E20, CC-E26, CC-E27, CC-E34, CC-E35; O-001, O-004, O-011, O-022, O-042. Tests: referral/receipt/assistance/start; accepted then stalled; partner only sends list; beneficiary declines; caregiver-own referral/outcome remains outside beneficiary/PCP material absent a separate permitted basis; agreement expires mid-route; downstream denial; no response; wrong recipient; clinical trigger; late evidence after closure.
AAA or Tribal Aging Program route — resolve operating failures#
When. When an individual route fails or repeated cases show rejection, nonresponse, unavailable help, unclear ownership, incomplete status return, scope mismatch, endpoint failure, or inconsistent performance.
What and how much. Protect the affected individual first with a current next action/alternate route; then classify the operational failure, preserve case evidence, determine whether it is isolated or patterned, contact the correct counterpart, agree on a correction or truthful limitation, test the correction, and review other open cases potentially affected. Do not hide failures to preserve the partnership.
Data/provenance. Use affected referral(s), agreement/process version, exact failure/status, attempts, counterpart responses, impact, service-seeker choice/risk, pattern denominator, corrective proposal/acceptance, and retest result. One failure does not prove route-wide failure; repeated missing status is not success.
Potential Proxi work. Detect patterns, group affected cases without losing individual obligations, prepare the failure packet, track corrective actions, identify open-case exposure, and verify retest evidence. AI cannot decide contractual breach, responsibility, clinical risk, or individual closure.
Human role. L1 verifies administrative facts and affected statuses. L2 handles each service seeker's current path and performs the nonclinical process-repair work with the outside AAA or Tribal Aging Program operational counterpart. The GUIDE Participant accountable authority performs only a retained organizational correction or agreement act. Healthcare Legal Counsel receives only a legal-authority question that the approved rule cannot resolve; separate U.S. clinical staff enters only for a triggered individual clinical or safety issue.
Provisional human minutes (low / typical / high). Per failure incident or grouped pattern review, L1 0 / 5 / 13; L2 3 / 8 / 15; separate U.S. L3/L4 clinical 0 / 0 / 10 only on trigger; retained GUIDE Participant acts, outside-counterpart time, and any triggered Healthcare Legal Counsel legal-authority question remain outside the ladder at 10 / 20 / 45. Portfolio pattern work is not multiplied per beneficiary, but each affected person's follow-up remains separate.
Completion evidence. Every affected case has a truthful current path; failure type/root evidence; counterpart disposition; accepted correction or documented limitation; retest result; impacted-case review; and unresolved risks/owners.
What does not prove completion. Escalation email, partner apology, new contact name, corrective-plan draft, one successful later referral, exhausted calls, or closing affected cases as connected.
Edge cases/open decisions/minimum tests. Existing: X-006, X-007, X-014, X-015, CC-E05, CC-E06, CC-E20, CC-E26, CC-E27, CC-E34; O-004, O-022. Tests: isolated rejection; repeated nonresponse; expired process; endpoint outage; partner capacity loss; unclear ownership; correction works/fails; open cases affected; urgent individual need during route failure.
AAA or Tribal Aging Program route — maintain the arrangement#
When. On a proposed scheduled relationship review and whenever contacts, services, geography, population, responsibilities, law/requirements, agreement status, endpoints, performance, or failure patterns materially change. CMS does not specify Proxi's review cadence.
What and how much. Review current parties/contacts, agreement and process scope/dates, actual referral and assistance outcomes, unresolved failures, coverage/capacity changes, evidence-return quality, and renewal/amendment needs. Confirm the route remains usable before continuing to rely on it; suspend unsupported scope rather than presenting it as active.
Data/provenance. Use executed agreement/version, process/test evidence, contact verification, referral/receipt/assistance/start/outcome metrics, failure logs, changes, renewal dates, and counterpart statements. Relationship longevity or absence of complaints does not prove current capability.
Potential Proxi work. Produce the review packet, alert dates/changes, verify contacts/endpoints, summarize performance without collapsing statuses, track renewals/corrections, and identify affected scope. AI cannot renew, amend, or certify sufficiency.
Human role. L2 and the outside AAA or Tribal Aging Program counterpart review practical referral usability and continued operations. The GUIDE Participant accountable authority performs any retained continuation or agreement-change approval. L1 performs contact and process checks. Software enforces the approved agreement, disclosure, routing, and review rules. Healthcare Legal Counsel receives only a legal-authority question that the approved rule cannot resolve. No clinical role is routine.
Provisional human minutes (low / typical / high). Per arrangement-review event, L1 5 / 10 / 20; L2 0 / 5 / 15; separate U.S. L3/L4 clinical 0 / 0 / 0; retained GUIDE Participant acts, outside-counterpart time, and any triggered Healthcare Legal Counsel legal-authority question remain outside the ladder at 10 / 25 / 60. These are portfolio minutes.
Completion evidence. Dated review denominator, current contacts/scope/agreement/process status, disaggregated performance/failure findings, continued/suspended/amended disposition, owners/dates, and affected-case or route corrections.
What does not prove completion. Calendar meeting, contract auto-renewal, no complaints, one recent referral, refreshed contact list, or agreement still signed while operations no longer work.
Edge cases/open decisions/minimum tests. Existing: X-005, X-006, X-013, X-014, CC-E26, CC-E27; O-004, O-022, O-PA-001. Proposed decision under O-022: arrangement review cadence, performance denominator, and suspension threshold. Tests: stable route; contact changes; service loss; partial geography; agreement expiry; process failure pattern; auto-renewal; amendment; temporary suspension; termination.
Medicaid HCBS detailed tasks 46-57#
These beneficiary-specific tasks apply only when the beneficiary is both eligible for and actually receiving HCBS through a state Medicaid program. Dual eligibility, Medicaid enrollment, an HCBS application, or past HCBS receipt does not by itself trigger §6.2. Tasks 49-54 should usually be performed as one prepared case-manager coordination interaction: the low/typical/high minutes are task contributions inside that event and must not be added as separate calls by default. L1 manual identity/status pursuit across tasks 47, 49, 51, and 54 is one shared 0 / 5 / 13-minute episode unless separate events are genuinely required.
Medicaid HCBS — identify when §6.2 applies#
When. At alignment/intake when Medicaid/HCBS facts are available, after an assessment or outside source indicates HCBS, and whenever Medicaid eligibility, HCBS eligibility, actual receipt, or program status changes.
What and how much. Establish three separate facts: Medicaid status, HCBS eligibility, and whether the beneficiary is currently receiving HCBS through a state Medicaid program. Record program/waiver and effective period when known. If receipt remains unknown or conflicted, hold the applicability determination and pursue evidence; do not trigger or suppress §6.2 from dual eligibility alone.
Data/provenance. Use beneficiary/authorized-representative report, current Medicaid/plan/program documents, payer/program data, case-manager response, service authorization and actual receipt evidence, effective/end dates, and conflicts. An eligibility card, waiver application, waiting-list status, prior authorization, claim, or dual status does not necessarily prove current HCBS receipt.
Potential Proxi work. Retrieve source-linked coverage/program facts, ask targeted questions, separate statuses, detect expiry/conflict, and prepare verification. AI cannot decide eligibility, interpret state-program rules conclusively, or infer receipt.
Human role. Software completes a current unconflicted path. L1 verifies administrative sources. L2 confirms the beneficiary's reported current service picture and resolves coordination context. The outside Medicaid program or case manager supplies authoritative program facts. No clinical role is routine.
Provisional human minutes (low / typical / high). Per applicability event, L1 0 / 0 / 5 within the shared administrative episode; L2 0 / 2 / 5; separate U.S. L3/L4 clinical 0 / 0 / 0.
Completion evidence. Separately sourced Medicaid, HCBS eligibility, and current-receipt facts with effective period; §6.2 applicable/not-applicable/unknown result; unresolved conflicts and verification owner.
What does not prove completion. Dual eligibility, Medicaid card, old waiver record, application, waitlist, service authorization, generic LTSS note, claim alone, or silence treated as not receiving.
Edge cases/open decisions/minimum tests. Existing: X-003, X-004, X-005, CC-E20, CC-E28; O-023. Tests: eligible and receiving; eligible not receiving; dual only; application/waitlist; ended waiver; new enrollment; conflicting sources; service authorized but not started; managed-care HCBS; status changes mid-episode.
Medicaid HCBS — identify the waiver or HCBS program and case manager#
When. After §6.2 applicability is established or while actual receipt is being verified, before beneficiary-specific coordination is attempted, and whenever program, plan, case-management entity, or manager changes.
What and how much. Identify the exact state HCBS/LTSS program or waiver, responsible Medicaid entity, assigned case manager or accepted coordination contact, organization, contact route, and effective period. Preserve multiple program/managers and disputed responsibility rather than choosing one without evidence.
Data/provenance. Use program/plan documents, official directories, beneficiary/representative information, current authorization/service records, verified organization response, manager identity/role, endpoint, and assignment dates. A payer customer-service line or prior case manager does not prove current assignment.
Potential Proxi work. Search records/directories, normalize program/organization identities, verify endpoints, detect changes, prepare a narrow verification request, and preserve attempts. AI cannot assign case responsibility or infer the correct manager from claim patterns.
Human role. L1 performs administrative identity and endpoint pursuit. L2 confirms the relationship and coordination counterpart when ambiguity remains. The outside Medicaid entity identifies or accepts its current responsible contact.
Provisional human minutes (low / typical / high). Per identity event, L1 0 / 5 / 13 within the shared administrative episode; L2 0 / 0 / 3; separate U.S. L3/L4 clinical 0 / 0 / 0.
Completion evidence. Exact program/waiver, Medicaid entity, current manager/accepted contact, verified endpoint, effective period, sources, and unresolved multi-manager responsibilities.
What does not prove completion. Old business card, generic plan number, provider directory, prior claim, beneficiary remembering a first name, organization contact without assignment confirmation, or one manager assumed responsible for every HCBS service.
Edge cases/open decisions/minimum tests. Existing: X-001, X-003, X-004, X-005, X-006, CC-E20, CC-E28, CC-E29; O-004, O-023. Tests: one verified manager; manager changed; no manager assigned; several waivers/managers; wrong program; entity refuses confirmation; no response; temporary coverage; late manager response.
Medicaid HCBS — confirm participation and information-sharing preferences#
When. After identifying the coordination counterpart and before beneficiary-specific information is released; repeat when the beneficiary's participation, caregiver/representative involvement, recipient, purpose, content, or authority changes.
What and how much. Confirm how the beneficiary wants to participate, who may join, what information may be shared for the coordination purpose, whether the beneficiary wants a caregiver/representative included, and any channel/accessibility preference. Separate participation preference from legal authority and from the case manager's own information-sharing requirements.
Data/provenance. Use authenticated identity, beneficiary preference, decision-specific representative authority, caregiver role, proposed recipient/purpose/content, permission or other approved basis, accessibility, revocation, actor, and time. HCBS receipt, caregiver involvement, or a care-management relationship does not automatically prove permission for every disclosure.
Potential Proxi work. Display recipient/content/purpose, collect bounded decisions, separate participant roles, block conflicted scope, and prepare accessible confirmation. AI cannot infer permission, decide authority, or expand scope.
Human role. Software enforces the approved participation and disclosure rules and first presents the exact participation choices and recipient/content/purpose scope; an authenticated beneficiary or decision-specific authorized representative may choose directly on the approved clean path. L2 conducts a human conversation only when required or requested or when ambiguity, disagreement, distress, accessibility failure, or relationship context remains. Healthcare Legal Counsel receives only a prepared legal-authority question that the approved rule cannot resolve. L1 may gather documents but cannot decide their effect. Separate U.S. clinical staff enters only for a decision-specific clinical-capacity function when actually required.
Provisional human minutes (low / typical / high). Per coordination episode, L1 0 / 0 / 0; L2 0 / 4 / 8; separate U.S. L3/L4 clinical 0 / 0 / 10 only for a triggered approved capacity function. The low route is the prepared authenticated direct path. A triggered Healthcare Legal Counsel legal-authority question is outside these role minutes.
Completion evidence. Authenticated participation and inclusion choices, exact recipient/purpose/content scope, authority/basis and revocation status, accessibility/channel needs, and blocked unresolved elements.
What does not prove completion. Medicaid enrollment, case-manager relationship, caregiver listed in chart, portal proxy, emergency contact, generic consent, prior participation, or beneficiary silence.
Edge cases/open decisions/minimum tests. Existing: X-002, X-004, X-005, CC-E19, CC-E35; O-001, O-011, O-023. Tests: beneficiary participates directly; caregiver included/not included; limited representative; expired authority; narrowed scope; revocation before/after release; accessible format; beneficiary and caregiver disagree.
Medicaid HCBS — contact and attempt coordination#
When. Once applicability, counterpart, beneficiary participation, permission/release prerequisites, and a coordination purpose are established; retry/follow according to approved policy when no matched response exists.
What and how much. Make a genuine beneficiary-specific contact attempt to the verified waiver/HCBS case manager, identify the GUIDE Participant and purpose, request the coordination interaction, and preserve sent, delivered, received, answered, declined, wrong-contact, and no-response states separately. An attempt may satisfy only the attempt fact; it does not prove joint coordination or service review.
Data/provenance. Use exact beneficiary/case, program/manager, verified endpoint, permission/content, release actor, channel/time, attempt identity, delivery/receipt, response, and next action. A sent message, voicemail, or technical success is not contact with the case manager.
Potential Proxi work. Build the request, support human release, transmit, match acknowledgements, schedule, remind, detect wrong/nonresponse, and preserve attempt history. AI cannot self-release PHI, claim coordination, or convert retry exhaustion into completion.
Human role. Software builds and, after the exact release by the authorized V1 releaser, transmits the approved coordination request. L1 handles administrative pursuit and manual scheduling within the shared episode. L2 owns only the substantive live relationship exchange after acceptance, or a requested or required warm introduction; L2 is not required to create or chase the request. The outside Medicaid case manager decides whether and how to engage. No clinical role is routine.
Provisional human minutes (low / typical / high). Inside the tasks 49-54 coordination cluster, L1 0 / 5 / 13 shared across the cluster; L2 2 / 4 / 8; separate U.S. L3/L4 clinical 0 / 0 / 0; authorized V1 releaser 1 / 1 / 2 until assigned.
Completion evidence. Exact request/release, verified endpoint, objective attempt/delivery/receipt/response state, scheduled or completed contact when achieved, and truthful no-response/decline with continuing owner.
What does not prove completion. Draft, release approval, send success, voicemail, office receipt, meeting invitation, attempts exhausted, or an administrative response labeled joint coordination.
Edge cases/open decisions/minimum tests. Existing: X-006, X-007, X-014, X-015, CC-E03, CC-E29, CC-E34, CC-E35; O-001, O-004, O-023. Tests: live contact; secure message response; voicemail only; wrong manager; refusal; no response; changed endpoint; revoked permission; late response; duplicate attempt.
Medicaid HCBS — explain GUIDE#
When. During the first substantive exchange with the waiver/HCBS case manager and when a new manager or material GUIDE service change makes the prior explanation insufficient.
What and how much. Provide an approved concise explanation of GUIDE plus the beneficiary-specific GUIDE services relevant to coordination, role boundaries, current contacts, and the reason for comparison. Answer operational questions from approved current material; route legal, billing, clinical, or unresolved policy questions to the proper owner.
Data/provenance. Use the approved GUIDE explanation, current beneficiary-specific GUIDE service picture, Participant/care-team contacts, source/version/date, recipient identity, exact content released, questions, answers, and unresolved items. Sending a brochure does not prove understanding or substantive exchange.
Potential Proxi work. Assemble the approved packet, tailor only confirmed beneficiary-specific facts, render/deliver through the selected channel, capture questions, and draft cited responses. AI cannot make commitments, interpret disputed policy, or invent service responsibilities.
Human role. L2 owns the nonclinical relationship exchange; software presents the approved standard content and enforces the approved routing and disclosure rules. The Medicaid Program Specialist receives only a bounded unresolved Medicaid-program question. Healthcare Legal Counsel receives only a legal-authority question that the approved rule cannot resolve. Separate U.S. clinical staff answers only actual clinical-content questions within authority.
Provisional human minutes (low / typical / high). Inside the shared coordination interaction, L1 0 / 0 / 0; L2 2 / 4 / 7; separate U.S. L3/L4 clinical 0 / 0 / 8 only on trigger. Do not schedule a separate call solely to read standard content if it can be included in the same exchange.
Completion evidence. Dated recipient-specific GUIDE explanation, exact approved content/version, relevant beneficiary service summary, questions/answers, and routed unresolved items.
What does not prove completion. Brochure sent, link delivered, generic script generated, unexplained acronym list, case manager attendance, or silence treated as comprehension/agreement.
Edge cases/open decisions/minimum tests. Existing: X-004, X-007, X-009, CC-E10, CC-E29; O-003, O-023, O-PA-001. Tests: manager already understands; new manager; inaccessible material; asks billing question; asks clinical question; beneficiary-specific service summary stale; content delivery fails; manager disputes GUIDE role.
Medicaid HCBS — obtain the Medicaid service picture#
When. In the substantive case-manager exchange after GUIDE is explained, and again when the Medicaid service picture is incomplete, conflicted, or materially changed.
What and how much. Obtain the services the beneficiary currently receives through Medicaid/HCBS, responsible program/organization, service type, known scope/frequency, authorization/effective period, actual delivery status, case-manager role, open changes/gaps, and source limitations. Keep authorization, scheduled service, and actual delivery distinct.
Data/provenance. Use case-manager statements, current service plan/authorization, provider/agency records when permitted, beneficiary/representative reports, dates, frequency/scope, and conflicts. Authorization does not prove service start or continuing delivery; beneficiary report does not substitute for program responsibility facts.
Potential Proxi work. Prepare known facts/questions, ingest documents/responses, normalize source-faithfully, compare sources, flag gaps/conflicts, and keep unknowns visible. AI cannot infer service receipt, eligibility, or current scope.
Human role. The outside Medicaid case manager supplies and confirms the Medicaid program picture. L2 clarifies and preserves the exchange. L1 retrieves permitted administrative records within the shared episode. Separate U.S. clinical staff is not routine.
Provisional human minutes (low / typical / high). Inside the shared coordination interaction, L1 0 / 0 / 0 incremental beyond the shared administrative episode; L2 2 / 5 / 10; separate U.S. L3/L4 clinical 0 / 0 / 0.
Completion evidence. Dated source-linked Medicaid service inventory with program/provider, scope/frequency, authorization and actual-delivery states, effective period, responsible contact, conflicts/unknowns, and case-manager confirmation or limitation.
What does not prove completion. Claims list, authorization, old service plan, beneficiary memory alone, provider directory, scheduled visit, case manager saying “covered,” or every field populated by AI.
Edge cases/open decisions/minimum tests. Existing: X-003, X-004, X-005, CC-E17, CC-E28, CC-E29; O-023. Tests: complete confirmed picture; authorization/no start; beneficiary reports service/case manager omits; multiple providers; service paused; expired authorization; no case-manager access; partial document; service changes mid-review.
Medicaid HCBS — review GUIDE and Medicaid services together#
When. Once both current service pictures are available in the case-manager exchange, and again after a material service or need change.
What and how much. Review the two service pictures side by side by beneficiary need, program, provider, actual delivery, scope/frequency, payer/funder when known, and responsible owner. Confirm each side's own facts; do not make a joint clinical, eligibility, or billing decision inside the comparison.
Data/provenance. Use current GUIDE and Medicaid inventories, source/effective dates, beneficiary needs, actual delivery, responsible organizations, case-manager corrections, and unresolved conflicts. Similar labels do not prove duplicate services; different labels do not prove distinct functions.
Potential Proxi work. Generate a source-linked comparison, align service descriptions without erasing differences, expose missing facts and possible overlap/gaps, and capture corrections. AI cannot decide duplication, program responsibility, or clinical equivalence.
Human role. L2 and the outside Medicaid case manager perform the nonclinical coordination review. L1 does not substitute for the substantive exchange. Separate U.S. clinical staff enters only if a clinical-service equivalence or safety question is triggered.
Provisional human minutes (low / typical / high). Inside the shared coordination interaction, L1 0 / 0 / 0; L2 2 / 4 / 8; separate U.S. L3/L4 clinical 0 / 0 / 10 only on trigger.
Completion evidence. Dated side-by-side comparison reviewed with the case manager, source-linked corrections, aligned needs/services, actual-delivery distinctions, and explicit unknowns/possible findings for the next task.
What does not prove completion. Two lists displayed, automated similarity score, document delivery, manager receipt, same category label, claim overlap, or GUIDE staff review without the case manager.
Edge cases/open decisions/minimum tests. Existing: X-004, X-005, CC-E17, CC-E29, CC-E30; O-023. Tests: complementary services; apparent same service/different scope; actual duplicate candidate; missing frequency; authorization/no delivery; case manager corrects GUIDE assumption; no joint review due nonresponse; clinical equivalence question.
Medicaid HCBS — identify gaps and duplication#
When. During the joint review after both service pictures are aligned, and whenever a need remains unserved, delivery changes, overlapping responsibility appears, or a party disputes the comparison.
What and how much. Classify each finding as confirmed/possible gap, complementary support, overlapping responsibility, possible duplicate delivery/payment, contradictory information, or unclear ownership. Preserve each program's position and create an owned follow-up; do not terminate useful support or make payment/clinical decisions from a text match.
Data/provenance. Use the reviewed comparison, beneficiary need, exact services/scope/frequency/actual delivery, payer/provider, responsible parties, source dates, each party's statement, and beneficiary experience. Similar service names or overlapping dates do not alone prove duplication.
Potential Proxi work. Highlight unmatched needs and overlaps, show source evidence/uncertainty, prepare focused questions, and route each finding to the correct domain. AI cannot declare duplication, assign responsibility, or choose which service stops.
Human role. L2 and the outside Medicaid case manager identify and discuss nonclinical coordination findings. Software applies the approved duplication rules and preserves unresolved responsibility. The existing Billing role receives only a bounded GUIDE payment or claim-integrity question, and the Medicaid Program Specialist receives only a bounded unresolved Medicaid-program question. Separate U.S. beneficiary-location-authorized clinical staff addresses only clinical scope or safety questions.
Provisional human minutes (low / typical / high). Inside the shared coordination interaction, L1 0 / 0 / 0; L2 1 / 2 / 4; separate U.S. L3/L4 clinical 0 / 0 / 10 only on trigger. Any bounded Billing-role or Medicaid Program Specialist exception time is outside these role minutes.
Completion evidence. Finding-by-finding classification with sources, each party's position, affected need/service, accountable follow-up, and explicit clinical/billing escalation where applicable.
What does not prove completion. Similarity score, same service category, simultaneous claims, gap flag, case-manager opinion treated as billing authority, or one program unilaterally removing service.
Edge cases/open decisions/minimum tests. Existing: X-004, X-010, X-015, CC-E17, CC-E30; O-003, O-023. Tests: clear gap; complementary services; partial overlap; possible duplicate payment; same provider/different scope; conflicting frequencies; disputed responsibility; clinical overlap; unresolved unknown.
Medicaid HCBS — attempt to coordinate the response#
When. For every gap, duplication concern, contradiction, or unclear responsibility identified in the joint review; continue until the required attempt and a truthful current outcome/owner are recorded.
What and how much. Seek a specific response from each responsible program/organization: proposed responsibility, next action, service continuation/change request, missing information, and follow-up. Record confirmed, declined, disputed, unanswered, or pending states. Do not promise a service, stop an existing service, or convert the coordination attempt into agreement.
Data/provenance. Use each finding, parties/authority, proposed action, beneficiary preference, exact outreach/response, dates, service continuity/safety facts, and unresolved risks. Case-manager agreement covers only their authority; office silence is not agreement; assignment is not accepted responsibility.
Potential Proxi work. Draft the responsibility/action matrix, prepare communications, track attempts/responses, surface disputes/nonresponse, preserve late events, and maintain remaining GUIDE work. AI cannot assign external responsibility or close from attempt exhaustion.
Human role. L2 conducts nonclinical relationship and referral coordination with the outside Medicaid case manager and relevant outside provider staff. L1 pursues administrative responses inside the shared episode. Software enforces approved routing and disclosure rules. The existing Billing role receives only a bounded GUIDE payment or claim-integrity question, the Medicaid Program Specialist receives only a bounded unresolved Medicaid-program question, and separate U.S. clinical staff resolves only a clinical-content question. External organizations authorize and deliver their own services.
Provisional human minutes (low / typical / high). Inside the tasks 49-54 cluster, L1 0 / 0 / 0 incremental beyond the shared 0 / 5 / 13 episode; L2 3 / 5 / 10; separate U.S. L3/L4 clinical 0 / 0 / 10 only on trigger.
Completion evidence. Each finding has documented outreach, source-backed confirmed/declined/disputed/unanswered result, current responsibility/owner where accepted, beneficiary impact, next follow-up, and remaining GUIDE obligation.
What does not prove completion. Draft matrix, outreach sent, meeting held, one party's assumption, silence, attempts exhausted, case-manager agreement outside scope, or removal of a service without authorized action.
Edge cases/open decisions/minimum tests. Existing: X-006, X-014, X-015, CC-E17, CC-E29, CC-E30, CC-E34; O-004, O-023. Tests: accepted responsibility; declined; dispute; no response; multiple providers; beneficiary declines proposed change; service continues during dispute; clinical/billing escalation; late agreement; GUIDE retains gap work.
Medicaid HCBS — explain responsibilities to the beneficiary#
When. After the coordination exchange produces confirmed, pending, disputed, or unanswered responsibilities and before the beneficiary is expected to act; update when the responsibility picture changes.
What and how much. Explain which program/organization has actually accepted each responsibility, what action is expected, what remains uncertain or disputed, what the beneficiary must do if anything, who to contact, and what GUIDE will continue pursuing. Do not present a proposed matrix as agreement or make the beneficiary resolve inter-program conflict.
Data/provenance. Use only the latest source-backed responsibility outcomes, exact open questions, owners/contact routes, next dates/actions, beneficiary participation/accessibility preferences, and clinical/safety instructions only from authorized sources.
Potential Proxi work. Produce an accessible plain-language summary, deliver it, capture corrections/questions, and show open versus confirmed responsibilities. AI cannot convert ambiguity into certainty or give clinical advice.
Human role. L2 owns the nonclinical explanation and preference-sensitive questions. Separate U.S. clinical staff explains only authorized clinical implications. L1 handles delivery exceptions.
Provisional human minutes (low / typical / high). Per episode, L1 0 / 0 / 3; L2 2 / 4 / 8; separate U.S. L3/L4 clinical 0 / 0 / 8 only on trigger.
Completion evidence. Accessible delivered summary distinguishing confirmed/pending/disputed/unanswered responsibilities, contacts/actions/dates, beneficiary questions/corrections, and GUIDE's remaining owner.
What does not prove completion. Summary generated, message sent without delivery, proposed matrix, jargon-heavy explanation, uncertain responsibility stated as fact, or beneficiary told merely to call Medicaid.
Edge cases/open decisions/minimum tests. Existing: X-007, CC-E17, CC-E19, CC-E29, CC-E30; O-023. Tests: clear responsibilities; partial dispute; no response; inaccessible format; beneficiary disagrees; representative present; clinical implication; responsibility changes after explanation.
Medicaid HCBS — reflect the coordinated services in the care plan#
When. After confirmed Medicaid/HCBS services, responsibilities, gaps, or follow-up materially affect beneficiary care-plan content; update when those facts change. Do not wait for all disputes to disappear if the plan must truthfully show an unresolved gap.
What and how much. Add only confirmed current GUIDE and Medicaid services, actual delivery state, responsible program/organization/contact, known scope/frequency, gap/overlap finding, accepted or unresolved responsibility, and follow-up. Keep operational facts separate from clinical implications and preserve source/conflict status.
Data/provenance. Use the reviewed service comparison, coordination outcomes, actual delivery evidence, beneficiary preference, source/effective dates, owner acceptance, open disputes, and information-sharing boundary. Authorization, proposed responsibility, or attempted coordination does not prove an active service or accepted owner.
Potential Proxi work. Draft a source-linked care-plan delta, preserve conflicts/unknowns, compare against current content, and route affected elements. AI cannot publish, infer clinical impact, or represent pending responsibility as settled.
Human role. L2 validates nonclinical coordination content. Software enforces the approved recipient-specific disclosure and content rules. Separate U.S. L3/L4 clinical staff reviews only triggered clinical implications. Healthcare Legal Counsel receives only a legal-authority question that the approved rule cannot resolve.
Provisional human minutes (low / typical / high). Per care-plan update event, L1 0 / 0 / 0; L2 1 / 3 / 7; separate U.S. L3/L4 clinical 0 / 0 / 8 only on trigger.
Completion evidence. Reviewed care-plan delta accurately states service/delivery/responsibility/gap/follow-up with sources, effective dates, unresolved status, any required clinical review, and the software-enforced recipient/content/purpose/authority/restriction outcome.
What does not prove completion. Draft, copied authorization, attempted coordination, proposed owner, stale service plan, AI clinical implication, or unresolved gap omitted to make the plan appear complete.
Edge cases/open decisions/minimum tests. Existing: X-004, X-005, X-011, CC-E17, CC-E20, CC-E30, CC-E35; O-010, O-023. Tests: confirmed active HCBS; authorization/no start; unresolved gap; disputed duplicate; responsibility accepted/declined; clinical implication; care-plan conflict; late service cessation.
Medicaid HCBS — re-coordinate when circumstances change#
When. When Medicaid/HCBS eligibility or receipt, program/manager, service scope/delivery, GUIDE services, beneficiary need/location, responsibility, or material clinical/safety context changes.
What and how much. Identify the changed fact and affected prior comparison/responsibilities, preserve old/new effective times, refresh only what is material, reconfirm participation/permission as needed, repeat contact/review/coordination, communicate the revised picture, and update the care plan. Do not overwrite the prior episode or assume the old agreement still applies.
Data/provenance. Use change source/time, old/new facts, affected services/owners, current case manager/program, beneficiary preference, prior coordination outcome, open obligations, and new responses. A detected change is not proof that every prior service or responsibility ended.
Potential Proxi work. Detect and compare changes, open a linked new coordination episode, carry forward only current supported facts, prepare refreshed packets/comparison, track responses, and preserve history. AI cannot decide material responsibility, eligibility, or clinical consequence.
Human role. L1 verifies administrative changes within one shared exception episode. L2 re-engages the beneficiary and outside Medicaid case manager for the substantive nonclinical coordination. Separate U.S. clinical staff enters only for triggered clinical or safety work.
Provisional human minutes (low / typical / high). Per material-change episode, L1 0 / 5 / 13; L2 3 / 8 / 18; separate U.S. L3/L4 clinical 0 / 0 / 10 only on trigger. Reused current facts are not charged again as new human collection.
Completion evidence. Source-linked change, linked prior/new episode, refreshed applicability/counterpart/comparison/responsibility results, beneficiary communication, care-plan effect, and explicit remaining obligations.
What does not prove completion. Change alert, copied old packet, manager reassignment notice, new authorization, one outreach attempt, old case overwritten, or care-plan update without renewed coordination where required.
Edge cases/open decisions/minimum tests. Existing: X-004, X-005, X-011, CC-E20, CC-E28, CC-E29, CC-E30, CC-E35, CC-E36; O-004, O-023. Tests: manager changes; HCBS ends/starts; service paused; beneficiary moves; GUIDE service changes; need resolves/emerges; responsibility dispute reopens; late correction; clinical trigger; exact duplicate change event.
Cross-pillar referral and service feeds#
This SOP participates in cross-pillar episodes under 25_GUIDE_Eight_Pillar_Service_Integration.md. Referral and Services receives attribution only for a person-centered option or choice and the promised verified referral, receipt, intake, start, usefulness, responsibility, or truthful non-success milestone actually reached. A list, send, receipt, registration, authorization, or owner assignment is not service start.
| Direction | Named feeds | Local handling and result | Review |
|---|---|---|---|
| Inbound to Referral and Services | Comprehensive Assessment; Care Plan; Ongoing Monitoring and Support; 24/7 Access; Care Coordination; Medication Management; Caregiver Education and Support | Accept the service seeker's current need and desired result, constraints, prior options and attempts, participation or permission limits, urgent disposition, and exact promised endpoint. Reuse them; do not restart intake. | |
| Outbound from Referral and Services | Comprehensive Assessment when a baseline need changes; Care Plan; Ongoing Monitoring and Support; 24/7 context; Care Coordination; Medication Management; Caregiver Education and Support | Return each actual milestone and barrier separately, plus the person's choice or usefulness and the next owner. |
When an Ongoing, 24/7, Care Plan, or caregiver interaction already establishes the current need and choice, reuse that work and count its human minutes once. Only later referral pursuit and external or service milestones add new work.
Open all 57 task proceduresDetailed task inventory
| Task | What the task entails | GUIDE anchor | What Proxi can do | Person required? | Person or role | Review |
|---|---|---|---|---|---|---|
| Community connection — receive an identified need | Take a beneficiary-facing or caregiver-facing community-service or support need identified through the assessment, care plan, ongoing contact, a transition, or a beneficiary or caregiver request. | RFA App. B §§6.1, 6.3 | Gather the need, source, service seeker's preferences already known, and prior attempts into one work item. | Partial | Software completes an authenticated current clean path; L2 enters only for currentness conflict, sensitive context, or unclear service seeker. | |
| Community connection — define the desired result | Ask what problem the service seeker wants help solving and what a useful result would look like. | RFA App. B §§6.1, 6.3 | Ask standard clarification questions and summarize the requested outcome. | Yes | The service seeker/authorized representative defines and may confirm digitally; L2 enters only for ambiguity, distress, disagreement, or requested human help. | |
| Community connection — confirm who should participate | Confirm whether the beneficiary, caregiver, legal representative, or another authorized person should be included in the referral and follow-up, while preserving the caregiver's own role when the caregiver is the service seeker. | RFA App. B §§6.1, 6.3 | Present known participation preferences, apply the approved participation/disclosure rule, and collect updates. | Partial | Relevant people confirm on the clean digital path; L2 handles preference conflict, and Healthcare Legal Counsel receives only a legal-authority question that the approved rule cannot resolve. | |
| Community connection — gather service constraints and preferences | Ask the service seeker about location, language, cultural preferences, accessibility, schedule, transportation, technology, cost limits, and other practical requirements. | RFA App. B §§6.1, 6.3, 6.4 | Conduct the structured intake and organize requirements for searching. | Partial | The service seeker supplies/affirms facts; software completes the clean intake; L2 enters for uncertainty, relationship context, or disagreement. | |
| Community connection — redirect urgent or clinical needs | Recognize when the request includes immediate safety, abuse, neglect, medical, behavioral-health, or other needs that require a different qualified response before routine community referral continues. | RFA App. B §§6.1, 6.3; care-team scope requirements | Surface concerning statements and connect the person to the approved safety/clinical route without deciding the issue. | Partial | No clinical labor on the ordinary nonurgent path; L1/L2 may initiate a stop/connection, while the separate U.S. authorized clinician or safety authority decides and acts. | |
| Community connection — search the available inventory | Find community-based services and supports that appear relevant to the identified beneficiary or caregiver need and the service seeker's location. | RFA App. B §§6.3, 6.4 | Search the maintained inventory using the need and practical requirements. | No | No person on the clean path; L1 checks inaccessible/manual sources and L2 enters only for a later relationship-sensitive question. | |
| Community connection — build a relevant option list | Remove obvious mismatches and assemble realistic options for the service seeker rather than presenting an unfiltered directory. | RFA App. B §§6.1, 6.3, 6.4 | Compare service descriptions and known requirements against the service seeker's preferences and explain the match and uncertainty. | No | Software completes the factual clean path; L2 enters only for unresolved relationship-sensitive fit and never selects for the service seeker. | |
| Community connection — confirm community-support fit | Confirm that a serious option addresses a common need of a person with dementia or caregiver and, as applicable, supports the beneficiary's safe community living. | RFA App. B §§6.1, 6.4 and RFA footnote 50 | Compare the service purpose and setting with the identified need and show any mismatch or uncertainty. | Partial | Software confirms source-backed factual fit; L2 resolves relationship-sensitive fit only when triggered; separate U.S. clinical staff decides genuine clinical suitability. | |
| Community connection — verify current service availability | Confirm that each serious option is operating, serves the service seeker's location and population, and has current contact and intake information. | RFA App. B §§6.1, 6.4 | Check current public information and prepare organization-specific verification questions. | Partial | Software verifies connected current facts; L1 pursues manual administrative facts; the service seeker later chooses and the outside organization states its own availability. | |
| Community connection — verify eligibility, cost, waitlist, and intake needs | Learn the basic eligibility rules, cost or funding options, wait time, required documents, and intake process before the service seeker chooses. | RFA App. B §§6.1, 6.4 | Gather published details, compare them with the service seeker's confirmed facts, and list unanswered questions. | Partial | Software/L1 gathers and compares factual requirements; L2 supports later tradeoffs; only the external program determines eligibility, cost, waitlist, acceptance, and admission. | |
| Community connection — explain the options | Present the available choices, material constraints, known costs, and uncertainties in understandable language. | RFA App. B §§6.1, 6.3, 6.4 | Produce a plain-language comparison tailored to the service seeker's stated priorities. | Yes | Puerto Rico GUIDE navigator (L2); beneficiary or caregiver seeking the service; interpreter participates when needed. | |
| Community connection — obtain the service seeker's choice | Ask which service, if any, the beneficiary or caregiver wants to pursue without selecting on that person's behalf. | RFA App. B §§6.1, 6.3; beneficiary freedom of choice | Record the choice and the reasons or preferences the service seeker wants the team to retain. | Yes | Beneficiary, caregiver, or decision-specific authorized representative chooses within scope; Puerto Rico GUIDE navigator (L2) supports when the human-entry condition applies. | |
| Community connection — confirm permission to share information | Determine whose information will be sent, what may be sent, to which organization, and whether the beneficiary, caregiver, or representative must authorize it. | RFA App. B §§6.1, 6.3; applicable privacy requirements | Software enforces the approved recipient/content/purpose/authority/restriction rule, shows the proposed disclosure, and collects the required confirmation or authorization. | Partial | The proper information owner or decision-specific authorized representative decides on the clean digital path; L2 handles relationship questions, and Healthcare Legal Counsel receives only a legal-authority question that the approved rule cannot resolve. | |
| Community connection — prepare the referral or intake request | Assemble the service sought, relevant needs and preferences, contact information, and documents the selected organization requires for the beneficiary or caregiver seeking the service. | RFA App. B §§6.1, 6.3 | Populate the referral or intake materials from confirmed information and identify missing items. | Partial | Software prepares source-linked content; the proper person supplies/attests personal or financial facts; L1/L2 handles only triggered missing-item or relationship exceptions. | |
| Community connection — send the referral or make the introduction | Execute the service seeker's selected route after permission, endpoint, and V1 release prerequisites: transmit the permitted packet, support self-referral, or make a requested warm introduction. | RFA App. B §§6.1, 6.3 | Verify prerequisites, execute the selected route, record the result, and open pursuit without claiming receipt or intake. | Partial | The authorized V1 releaser performs the exact release; software handles routine transmission; L2 performs only a requested warm introduction and owns relationship follow-through, not the service decision. | |
| Community connection — help complete intake | Help schedule an intake, complete nonclinical administrative steps, gather documents, and understand what the service organization needs next. | RFA App. B §§6.1, 6.3 | Provide a checklist, prefill allowed fields, prepare calls, and remind the service seeker of outstanding items. | Partial | Software completes the clean path; L1 handles disconnected/manual exceptions; the service seeker attests personal facts and the organization completes/decides intake. | |
| Community connection — address access barriers | Help resolve transportation, language, accessibility, technology, paperwork, scheduling, caregiver-availability, or other practical barriers to intake. | RFA App. B §§6.1, 6.3 | Identify factual remedies, execute the selected administrative action, and track whether it worked. | Partial | No human when no barrier exists; L1 implements administrative logistics, L2 supports relationship-sensitive tradeoffs, the Medicaid Program Specialist receives only a bounded Medicaid question, and a separate U.S. clinician receives only a clinical-content question. | |
| Community connection — confirm referral receipt | Verify that the organization received the referral or that the beneficiary or caregiver seeking the service successfully made contact. | RFA App. B §§6.1, 6.3 | Check electronic receipt or organization response and initiate an approved follow-up when neither is available. | Partial | Software completes matched receipt; L1 pursues missing/disputed administrative status; L2 enters only for relationship or choice exceptions. | |
| Community connection — confirm intake disposition | Determine whether intake is scheduled, completed, waitlisted, denied, declined, or stalled and why. | RFA App. B §§6.1, 6.3 | Gather available status information and present unresolved cases for outreach. | Partial | Software/L1 retrieves source status; L2 supports the person's next choice only when triggered; the organization alone determines its intake disposition. | |
| Community connection — respond to denial, waitlist, or failed intake | Explain the barrier, correct remediable errors, appeal or supply more information when the service seeker wants and a process exists, or search for another option. | RFA App. B §§6.1, 6.3 | Summarize the stated reason, list missing items or alternate services, and prepare the next contact. | Yes | Puerto Rico GUIDE navigator (L2) performs relationship/referral coordination; the beneficiary or caregiver seeking the service decides whether to continue; the outside service organization decides its eligibility result. | |
| Community connection — confirm service start | Determine whether the beneficiary or caregiver actually began receiving the service rather than treating a sent referral or accepted intake as connection. | RFA App. B §§6.1, 6.3 | Look for person/service-specific start evidence and prepare outreach when absent. | Partial | Software matches objective evidence; L1 pursues administrative confirmation; L2 enters for conflicting lived status or a relationship barrier. | |
| Community connection — check whether the service meets the need | Ask the person receiving the service whether it is usable, acceptable, and addressing the identified need and whether new problems have emerged. | RFA App. B §§6.1, 6.3 | Deliver bounded follow-up and compare the source-faithful response with the desired result without deciding effectiveness. | Yes | The service seeker evaluates the service; L2 enters for relationship-sensitive discussion, and separate U.S. clinical staff handles triggered clinical/safety questions. | |
| Community connection — pursue another option when needed | Search and connect to an alternative when the service is unavailable, unaffordable, inappropriate, declined, or unsuccessful and the service seeker still wants help. | RFA App. B §§6.1, 6.3, 6.4 | Carry forward current confirmed preferences/barriers and refresh dynamic facts without forcing repetition. | Partial | Software/L1 prepares alternatives; L2 supports changed tradeoffs when needed; the service seeker chooses the next route. | |
| Community connection — explain the outcome | Tell the beneficiary or caregiver seeking the service what was done, the service status, outstanding requirements, and who will do the next action. | RFA App. B §§6.1, 6.3 | Prepare and deliver a plain-language source-backed summary. | Partial | Software may deliver a clean factual update; L2 enters for sensitive, disputed, confusing, or preference-sensitive outcomes. | |
| Community connection — update the care plan | When relevant, add the selected beneficiary- or caregiver-facing service, service recipient, responsible organization, cost or payer information, participant responsibilities, and actual connection outcome to the care plan. | RFA App. B §§6.1, 6.3; RFA care-plan requirements | Draft the care-plan change from confirmed information and identify gaps. | Partial | Puerto Rico GUIDE navigator (L2) updates nonclinical coordination content; the beneficiary and caregiver participate in their respective choices; a separate U.S. clinician reviews only triggered clinical implications. | |
| Community connection — continue unresolved follow-up | Keep pursuing an unresolved need until the service seeker declines, the need is confirmed resolved, the intended outcome is evidenced, or another person/organization explicitly accepts continuing ownership. | RFA App. B §§6.1, 6.3 | Track next actions, overdue work, approved reminders, attempts, owner acceptance, and last confirmed result. | Partial | Software tracks; L1 performs administrative pursuit; L2 owns relationship exceptions and truthful ending discussions, not silent closure from retry exhaustion. | |
| Community inventory — identify needed service categories | Use needs found in comprehensive assessments and ongoing care to identify which types of local services the inventory must cover. | RFA App. B §§6.3, 6.4 | Aggregate de-identified need categories and show missing or thinly covered categories by service area. | Partial | Puerto Rico GUIDE navigator (L2), performing the referral-coordination task, sets nonclinical inventory priorities; the GUIDE Participant accountable authority performs only a retained portfolio decision. | |
| Community inventory — join or maintain a referral inventory system | Ensure the organization participates in or operates a community referral inventory that covers health-related social needs assessed in GUIDE. | RFA App. B §6.3 | Maintain searchable entries, intake fields, service categories, and links to external inventory sources. | Partial | The GUIDE Participant accountable authority performs the retained system-selection act; software maintains the routine inventory, L1 handles administrative exceptions, and L2 owns referral relationships. | |
| Community inventory — discover candidate organizations | Find organizations providing relevant supports across the GUIDE service area, including meals, adult day services, personal care, environmental modifications, contractors, food support, transportation, in-home assistance, exercise, and socialization. | RFA App. B §§6.1, 6.4 and RFA footnote 50 | Search authoritative directories and organization sources and prepare candidate records. | Partial | Software performs routine discovery; L1 verifies local administrative facts, and L2 contributes referral-relationship knowledge only when needed. | |
| Community inventory — capture useful service details | Record service description, population, geography, contact routes, language, accessibility, cost, funding, eligibility, intake process, documents, waitlist, transportation, and hours when available. | RFA App. B §§6.3, 6.4 | Extract and organize published details while retaining the source and date observed. | Partial | Software performs routine extraction; L1 confirms administrative information that cannot be obtained from current sources, and the outside organization remains the source of its own facts. | |
| Community inventory — verify an organization directly | Contact the organization to confirm that key referral information is current and learn material details not reliably published. | RFA App. B §6.4 | Prepare a verification checklist, show the last known information, and incorporate confirmed corrections. | Yes | L1 performs administrative verification; L2 enters only for consequential referral-relationship work; the outside organization supplies its own current facts. | |
| Community inventory — confirm service-area relevance | Determine whether the service actually accepts people from the beneficiary communities served by the GUIDE Participant. | RFA App. B §§6.3, 6.4 | Compare stated coverage areas with the GUIDE service area and identify ambiguity. | Partial | L1 confirms unclear administrative boundaries with the outside organization; L2 enters only when referral-relationship context remains. | |
| Community inventory — correct changed or inaccurate information | Update contact, eligibility, price, waitlist, service, or availability information and stop presenting an option that is closed or known to be unsuitable. | RFA App. B §§6.3, 6.4 | Detect differences from current organization sources and prepare corrections. | Partial | L1 verifies administrative changes with the outside organization; L2 handles consequential referral-relationship follow-up. | |
| Community inventory — recheck entries regularly | Reconfirm high-use and change-prone services often enough to avoid repeated referrals to closed or materially changed programs. | RFA App. B §§6.3, 6.4 | Schedule rechecks, revisit organization sources, identify stale entries, and prepare direct-verification work. | Partial | Software schedules routine rechecks; L1 performs batched administrative verification, and L2 handles consequential referral-relationship failures. | |
| Community inventory — identify coverage gaps | Find locations, languages, accessibility needs, price ranges, or service categories for which the inventory has no realistic option. | RFA App. B §§6.3, 6.4 | Analyze the usable inventory against observed beneficiary needs and map the gaps. | Partial | L2 validates the practical referral gap and pursues local relationships; the GUIDE Participant accountable authority performs only a retained portfolio escalation decision. | |
| Community inventory — share relevant resources | Give a beneficiary or caregiver the specific resources appropriate to their own need rather than merely maintaining the list internally. | RFA App. B §§6.3, 6.4 | Produce a service-seeker-specific resource list with current known details and next steps. | Yes | Puerto Rico GUIDE navigator (L2) explains the options and supports the beneficiary's or caregiver's own choice when a human service is required or requested. | |
| AAA or Tribal Aging Program route — choose how the Participant will meet §6.1 | Decide whether the Participant will directly refer and connect people, use a written AAA or Tribal Aging Program agreement, or use both approaches. | RFA App. B §6.1 | Summarize the service-area needs, internal capacity, and available aging-network organizations for leadership. | Yes | The GUIDE Participant accountable authority performs the retained route-selection act; software applies the approved route rules, L2 supplies referral-operating evidence, and Healthcare Legal Counsel receives only a legal-authority question that the approved rule cannot resolve. | |
| AAA or Tribal Aging Program route — identify the correct organization | Identify the local Area Agency on Aging or the applicable Tribal Aging Program funded through Title VI of the Older Americans Act for the served geography and population. | RFA App. B §6.1 | Search official aging-network sources and map organizations to the GUIDE service area. | Partial | Software maps current official sources; L1 confirms unclear administrative jurisdiction or contact facts with the outside organization; L2 enters only for referral-relationship context, and Healthcare Legal Counsel receives only a legal-authority question that the approved rule cannot resolve. | |
| AAA or Tribal Aging Program route — establish the relationship | Contact the organization, explain the GUIDE service need, learn its capabilities, and identify the people who can authorize and operate an arrangement. | RFA App. B §6.1 | Prepare an organization brief, outreach message, agenda, and record of open questions. | Yes | Puerto Rico GUIDE navigator (L2) owns the nonclinical referral relationship with the outside AAA or Tribal Aging Program representative; the GUIDE Participant accountable authority performs only a retained commitment or approval act. | |
| AAA or Tribal Aging Program route — define the assistance to be provided | Specify how the organization will help beneficiaries and caregivers coordinate community-based services and supports, including referral routes and follow-up responsibilities. | RFA App. B §6.1 | Draft an operational responsibility list from the agreed service approach. | Yes | Puerto Rico GUIDE navigator (L2) coordinates with the outside AAA or Tribal Aging Program operational counterpart; the GUIDE Participant accountable authority performs the retained approval act, software enforces approved disclosure and routing rules, and Healthcare Legal Counsel receives only a legal-authority question that the approved rule cannot resolve. | |
| AAA or Tribal Aging Program route — execute a written agreement when using that option | Put the required assistance, parties, responsibilities, contacts, and operating terms into a written agreement and obtain authorized signatures. | RFA App. B §6.1 | Prepare the document from approved terms, identify missing provisions, and route it for review and signature. | Yes | The exact authorized GUIDE Participant signatory and outside AAA or Tribal Aging Program signatory execute the agreement; software checks approved terms and routing conditions, and Healthcare Legal Counsel receives only a legal-authority question that the approved rule cannot resolve. | |
| AAA or Tribal Aging Program route — establish the working referral process | Agree how referrals are sent, what information is needed, who contacts the beneficiary, how status returns, and whom to contact when the process fails. | RFA App. B §6.1 | Turn the agreed process into staff instructions and referral materials. | Yes | Puerto Rico GUIDE navigator (L2) and the outside AAA or Tribal Aging Program operational counterpart define and validate the nonclinical workflow; software enforces approved disclosure, release, routing, and recovery rules, and the GUIDE Participant accountable authority performs only the retained workflow-approval act. | |
| AAA or Tribal Aging Program route — send and follow an individual referral | Refer a beneficiary or caregiver through the agreed route and verify that the organization begins the assistance it agreed to provide. | RFA App. B §6.1 | Prepare and send the permitted referral, track receipt and status, and identify missing follow-up. | Partial | The authorized V1 releaser performs the exact release; software enforces approved disclosure and routing rules; L2 owns relationship follow-through; the outside AAA or Tribal Aging Program coordinator performs and reports its assistance; the beneficiary or representative participates. | |
| AAA or Tribal Aging Program route — resolve operating failures | Address rejected referrals, unavailable services, unclear ownership, nonresponse, or repeated breakdowns with the aging-network partner. | RFA App. B §6.1 | Show affected cases and the recurring failure pattern and prepare focused follow-up. | Yes | Puerto Rico GUIDE navigator (L2) owns nonclinical process repair with the outside AAA or Tribal Aging Program operational counterpart; L1 verifies administrative facts, the GUIDE Participant accountable authority performs only a retained organizational correction or agreement act, and Healthcare Legal Counsel receives only a legal-authority question that the approved rule cannot resolve. | |
| AAA or Tribal Aging Program route — maintain the arrangement | Keep contacts and procedures current and revisit the working relationship when services, geography, responsibilities, or operating conditions change. | RFA App. B §6.1 | Produce a review list from referrals, failures, changed contacts, and known service changes. | Yes | Puerto Rico GUIDE navigator (L2) and the outside AAA or Tribal Aging Program counterpart review referral operations; the GUIDE Participant accountable authority performs any retained continuation or agreement-change approval, software enforces approved agreement, disclosure, and routing rules, and Healthcare Legal Counsel receives only a legal-authority question that the approved rule cannot resolve. | |
| Medicaid HCBS — identify when §6.2 applies | Determine whether the beneficiary is eligible for and actually receiving HCBS through a state Medicaid program; do not assume the requirement applies solely because the person is dually eligible. | RFA App. B §6.2 | Gather known coverage and service information and ask targeted questions when receiving status is unclear. | Partial | Software completes the current unconflicted path; L1 verifies administrative sources; L2 resolves relationship or coordination context; the outside Medicaid program or case manager supplies authoritative program facts. | |
| Medicaid HCBS — identify the waiver or HCBS program and case manager | Find the specific state Medicaid HCBS/LTSS program, responsible case manager, organization, and current contact route. | RFA App. B §6.2 | Search available records and directories and prepare a verification request. | Partial | L1 performs administrative identity and endpoint pursuit; L2 confirms the relationship or coordination counterpart when ambiguity remains; the outside Medicaid entity identifies or accepts its current responsible contact. | |
| Medicaid HCBS — confirm participation and information-sharing preferences | Ask how the beneficiary wants to participate in the coordination and whether an authorized caregiver or representative should be included. | RFA App. B §6.2; person-centered delivery | Present known preferences and collect updates and permissions needed for contact. | Yes | Software enforces the approved participation and disclosure rules; the beneficiary or decision-specific authorized representative chooses on the clean path; L2 handles relationship questions or disagreement, and Healthcare Legal Counsel receives only a legal-authority question that the approved rule cannot resolve. | |
| Medicaid HCBS — contact and attempt coordination | Reach the waiver or HCBS case manager and make a genuine attempt to coordinate GUIDE community supports with Medicaid HCBS. | RFA App. B §6.2 | Prepare and send the coordination request, track attempts, and identify failed contact information. | Partial | L1 prepares and pursues the approved administrative request; L2 owns the substantive live relationship exchange or requested warm introduction; software enforces approved disclosure and routing rules; the outside Medicaid case manager decides whether and how to engage. | |
| Medicaid HCBS — explain GUIDE | Give the case manager the information needed to understand the GUIDE Model and the services the beneficiary is receiving through GUIDE. | RFA App. B §6.2 | Prepare a standard GUIDE explanation plus the beneficiary-specific service summary. | Partial | L2 owns the nonclinical relationship exchange; software presents approved standard content and enforces approved disclosure and routing rules; the Medicaid Program Specialist receives only a bounded unresolved Medicaid-program question, Healthcare Legal Counsel receives only a legal-authority question that the approved rule cannot resolve, and separate U.S. clinical staff receives only a clinical-content question. | |
| Medicaid HCBS — obtain the Medicaid service picture | Learn which community-based services and HCBS the beneficiary currently receives through Medicaid, including responsible organizations and known frequency or scope. | RFA App. B §6.2 | Create a comparison-ready summary from case-manager information and available documents. | Partial | The outside Medicaid waiver/HCBS case manager supplies or confirms Medicaid information; L2 clarifies the relationship exchange, and L1 retrieves permitted administrative records. | |
| Medicaid HCBS — review GUIDE and Medicaid services together | Compare the services received through GUIDE and Medicaid to understand how they interact. | RFA App. B §6.2 | Place both service lists side by side by need, provider, responsible program, and known scope. | Partial | L2 and the outside Medicaid waiver/HCBS case manager perform the nonclinical coordination review; separate U.S. clinical staff enters only for a clinical-service equivalence or safety question. | |
| Medicaid HCBS — identify gaps and duplication | Determine whether a need is unserved, both programs appear to be arranging the same service, or responsibility is unclear. | RFA App. B §6.2 | Highlight unmatched needs, overlapping service descriptions, and conflicting responsibility information. | Yes | Software applies approved duplication rules; L2 and the outside Medicaid waiver/HCBS case manager discuss nonclinical operational responsibility; the existing Billing role receives only a bounded GUIDE payment or claim-integrity question, the Medicaid Program Specialist receives only a bounded unresolved Medicaid-program question, and separate U.S. clinicians own clinical-content judgment. | |
| Medicaid HCBS — attempt to coordinate the response | Seek a clear follow-up from each program or organization to address gaps or avoid duplication without promising a service another program has not authorized. If the case manager is unreachable, declines coordination, or disputes responsibility, record the attempts and unresolved result accurately and continue appropriate GUIDE follow-up. | RFA App. B §6.2 | Track outreach, draft the proposed responsibility list, and preserve the confirmed, declined, disputed, or unanswered result without converting an attempt into agreement. | Yes | L2 attempts nonclinical relationship and referral coordination with the outside Medicaid case manager and relevant outside provider staff; L1 pursues administrative responses; software enforces approved disclosure and routing rules; each outside organization authorizes and delivers its own services. | |
| Medicaid HCBS — explain responsibilities to the beneficiary | Tell the beneficiary which program or organization is expected to handle each need, what remains uncertain, and whom to contact. | RFA App. B §6.2 | Prepare a plain-language side-by-side summary of confirmed responsibilities. | Yes | Puerto Rico GUIDE navigator (L2) owns the nonclinical explanation and relationship-sensitive questions; separate U.S. clinical staff explains only actual clinical implications. | |
| Medicaid HCBS — reflect the coordinated services in the care plan | Add confirmed GUIDE and Medicaid services, responsible contacts, gaps, and agreed follow-up to the care plan. | RFA App. B §6.2; RFA care-plan requirements | Draft the relevant updates from the reviewed comparison. | Partial | L2 validates nonclinical coordination content; software enforces approved recipient-specific disclosure and content rules; the beneficiary participates, separate U.S. clinical staff reviews only triggered clinical implications, and Healthcare Legal Counsel receives only a legal-authority question that the approved rule cannot resolve. | |
| Medicaid HCBS — re-coordinate when circumstances change | Repeat coordination when Medicaid eligibility, HCBS services, case manager, GUIDE services, or beneficiary needs materially change. | RFA App. B §6.2 | Detect recorded changes, prepare an updated comparison, and prompt renewed contact. | Partial | L1 verifies administrative changes; L2 re-engages the beneficiary and outside Medicaid waiver/HCBS case manager for substantive nonclinical coordination; separate U.S. clinical staff enters only for triggered clinical or safety work. |
Requirement, value, and clinical classificationReference table
| Task | GUIDE standing | Customer-value position | Clinical lane | Why |
|---|---|---|---|---|
| Community connection — receive an identified need | Necessary delivery work | Core customer value | No clinical judgment | Opens beneficiary- or caregiver-facing community support work. |
| Community connection — define the desired result | Necessary delivery work | Core customer value | No clinical judgment | The service seeker defines the outcome they want. |
| Community connection — confirm who should participate | Necessary delivery work | Core customer value | No clinical judgment | Preserves the service seeker's participation and representation choices. |
| Community connection — gather service constraints and preferences | Necessary delivery work | Core customer value | No clinical judgment | Makes the referral usable for the service seeker's circumstances. |
| Community connection — redirect urgent or clinical needs | Necessary delivery work | Core customer value | Clinical review on trigger | Safe referral delivery requires removing urgent or clinical matters from the routine route; the clinician or safety authority decides the issue. |
| Community connection — search the available inventory | Necessary delivery work | Value through better execution | No clinical judgment | Uses the required inventory to find relevant options. |
| Community connection — build a relevant option list | Necessary delivery work | Core customer value | No clinical judgment | Converts a directory into realistic choices without choosing for the person. |
| Community connection — confirm community-support fit | Necessary delivery work | Core customer value | No clinical judgment | Confirms nonclinical service fit against the stated need and setting. |
| Community connection — verify current service availability | Necessary delivery work | Value through better execution | No clinical judgment | Prevents referrals to closed or unavailable services. |
| Community connection — verify eligibility, cost, waitlist, and intake needs | Necessary delivery work | Core customer value | No clinical judgment | Provides practical facts while the service organization decides eligibility. |
| Community connection — explain the options | Necessary delivery work | Core customer value | No clinical judgment | Supports an informed beneficiary or caregiver choice. |
| Community connection — obtain the service seeker's choice | Necessary delivery work | Core customer value | No clinical judgment | The beneficiary or caregiver chooses their own service. |
| Community connection — confirm permission to share information | Necessary delivery work | Core customer value | No clinical judgment | Protects the service seeker's control of their information. |
| Community connection — prepare the referral or intake request | Necessary delivery work | Value through better execution | No clinical judgment | Assembles the factual packet needed for connection. |
| Community connection — send the referral or make the introduction | Public RFA care-delivery requirement | Core customer value | No clinical judgment | RFA §§6.1 and 6.3 require navigator referral and connection. |
| Community connection — help complete intake | Necessary delivery work | Core customer value | No clinical judgment | Removes administrative barriers after referral. |
| Community connection — address access barriers | Necessary delivery work | Core customer value | No clinical judgment | Resolves practical obstacles without making clinical decisions. |
| Community connection — confirm referral receipt | Necessary delivery work | Value through better execution | No clinical judgment | Verifies the receiving organization got the referral. |
| Community connection — confirm intake disposition | Necessary delivery work | Value through better execution | No clinical judgment | Distinguishes progress, denial, waitlist, and stalled intake. |
| Community connection — respond to denial, waitlist, or failed intake | Necessary delivery work | Core customer value | No clinical judgment | Pursues the service seeker's chosen nonclinical next step. |
| Community connection — confirm service start | Public RFA care-delivery requirement | Core customer value | No clinical judgment | Actual service start proves connection rather than referral alone. |
| Community connection — check whether the service meets the need | Beyond the public GUIDE minimum | Additional evidence-backed value | Clinical review on trigger | Outcome follow-up adds value; clinical or safety concerns are escalated. |
| Community connection — pursue another option when needed | Necessary delivery work | Value through better execution | No clinical judgment | A failed first option should not end the required connection effort. |
| Community connection — explain the outcome | Necessary delivery work | Core customer value | No clinical judgment | Gives the service seeker a clear result and next action. |
| Community connection — update the care plan | Necessary delivery work | Value through better execution | Clinical review on trigger | Service facts are delegable; clinical implications require review. |
| Community connection — continue unresolved follow-up | Necessary delivery work | Value through better execution | Clinical review on trigger | Routine tracking is delegable; emerging clinical concerns are escalated. |
| Community inventory — identify needed service categories | Necessary delivery work | Compliance infrastructure | No clinical judgment | Aligns inventory coverage with assessed HRSNs and common needs. |
| Community inventory — join or maintain a referral inventory system | Public RFA care-delivery requirement | Compliance infrastructure | No clinical judgment | RFA §6.3 directly requires joining or maintaining the system. |
| Community inventory — discover candidate organizations | Necessary delivery work | Value through better execution | No clinical judgment | Populates the required inventory with usable local options. |
| Community inventory — capture useful service details | Necessary delivery work | Compliance infrastructure | No clinical judgment | Records the facts needed for accurate referrals. |
| Community inventory — verify an organization directly | Necessary delivery work | Value through better execution | No clinical judgment | Confirms material facts not reliably available elsewhere. |
| Community inventory — confirm service-area relevance | Necessary delivery work | Compliance infrastructure | No clinical judgment | Prevents listing services that do not serve the covered area. |
| Community inventory — correct changed or inaccurate information | Necessary delivery work | Value through better execution | No clinical judgment | Keeps the required resource inventory usable. |
| Community inventory — recheck entries regularly | Necessary delivery work | Value through better execution | No clinical judgment | Reduces repeated failed referrals caused by stale facts. |
| Community inventory — identify coverage gaps | Beyond the public GUIDE minimum | Additional evidence-backed value | No clinical judgment | Proactively finds communities and needs with no realistic option. |
| Community inventory — share relevant resources | Public RFA care-delivery requirement | Core customer value | No clinical judgment | RFA §6.4 requires appropriate sharing with beneficiary and caregiver. |
| AAA or Tribal Aging Program route — choose how the Participant will meet §6.1 | Necessary delivery work | Compliance infrastructure | No clinical judgment | Selects direct referral, the written-agreement route, or both. |
| AAA or Tribal Aging Program route — identify the correct organization | Necessary delivery work | Compliance infrastructure | No clinical judgment | Ensures the agreement uses the correct local AAA or Title VI program. |
| AAA or Tribal Aging Program route — establish the relationship | Necessary delivery work | Compliance infrastructure | No clinical judgment | Creates the counterpart relationship needed for the chosen route. |
| AAA or Tribal Aging Program route — define the assistance to be provided | Necessary delivery work | Compliance infrastructure | No clinical judgment | Makes the required partner assistance operationally specific. |
| AAA or Tribal Aging Program route — execute a written agreement when using that option | Public RFA care-delivery requirement | Compliance infrastructure | No clinical judgment | RFA §6.1 directly requires a written agreement for this option. |
| AAA or Tribal Aging Program route — establish the working referral process | Necessary delivery work | Compliance infrastructure | No clinical judgment | Turns agreement language into a usable referral path. |
| AAA or Tribal Aging Program route — send and follow an individual referral | Necessary delivery work | Core customer value | No clinical judgment | The written agreement must require assistance, while this exact participant referral-and-follow-up step is the chosen operating method. |
| AAA or Tribal Aging Program route — resolve operating failures | Necessary delivery work | Value through better execution | No clinical judgment | Repairs rejected, unclear, or unanswered referrals. |
| AAA or Tribal Aging Program route — maintain the arrangement | Necessary delivery work | Compliance infrastructure | No clinical judgment | Keeps contacts and the written route usable over time. |
| Medicaid HCBS — identify when §6.2 applies | Necessary delivery work | Compliance infrastructure | No clinical judgment | Applies the duty only when the beneficiary receives state Medicaid HCBS. |
| Medicaid HCBS — identify the waiver or HCBS program and case manager | Necessary delivery work | Value through better execution | No clinical judgment | Finds the specific person CMS requires GUIDE to contact. |
| Medicaid HCBS — confirm participation and information-sharing preferences | Necessary delivery work | Core customer value | No clinical judgment | Preserves beneficiary choice in the coordination exchange. |
| Medicaid HCBS — contact and attempt coordination | Public RFA care-delivery requirement | Core customer value | No clinical judgment | RFA §6.2 directly requires contact and an attempt to coordinate. |
| Medicaid HCBS — explain GUIDE | Public RFA care-delivery requirement | Compliance infrastructure | No clinical judgment | RFA §6.2 requires sharing information about GUIDE. |
| Medicaid HCBS — obtain the Medicaid service picture | Public RFA care-delivery requirement | Value through better execution | No clinical judgment | RFA §6.2 requires review of services received through Medicaid. |
| Medicaid HCBS — review GUIDE and Medicaid services together | Public RFA care-delivery requirement | Value through better execution | No clinical judgment | The required comparison makes gaps and duplication visible. |
| Medicaid HCBS — identify gaps and duplication | Public RFA care-delivery requirement | Core customer value | Clinical review on trigger | Operational comparison is routine; clinical overlap questions are escalated. |
| Medicaid HCBS — attempt to coordinate the response | Public RFA care-delivery requirement | Core customer value | Clinical review on trigger | Coordination is required as an attempt; clinical responsibility questions go to clinicians. |
| Medicaid HCBS — explain responsibilities to the beneficiary | Necessary delivery work | Core customer value | No clinical judgment | Makes confirmed and unresolved program responsibilities understandable. |
| Medicaid HCBS — reflect the coordinated services in the care plan | Necessary delivery work | Value through better execution | Clinical review on trigger | Coordination facts are delegable; clinical implications require review. |
| Medicaid HCBS — re-coordinate when circumstances change | Necessary delivery work | Value through better execution | Clinical review on trigger | Repeats routine coordination and escalates new clinical conflicts. |