Source baseline: CMS GUIDE Request for Applications, Appendix B, Sections 3 and 8.2.4–8.3.4. The Proxi column identifies candidate automation, not functionality already deployed. The executed GUIDE Participation Agreement and current CMS instructions must be checked before field use if they differ from the public RFA.
Settled Proxi coverage design#
Proxi self-staffs human first answer with a Philippines-based team. The team receives the call, identifies the caller and beneficiary, captures the concern and callback route, provides approved nonclinical program information, invokes protocol-defined emergency or clinical routes, performs warm transfer and documentation, and shares every interaction with the interdisciplinary team. Puerto Rico care navigators provide nonclinical care-team-only caregiver support and follow-up; the separate U.S. beneficiary-location-authorized clinical workforce provides every clinical or medical assessment, triage, and judgment.
This design does not change D-005: a Philippines first-answer call does not count as the required GUIDE navigator contact. A Philippine RN without beneficiary-location authorization remains in the nonclinical first-answer lane. Activation remains subject to O-001, the executed Participation Agreement, the approved Partner Organization arrangement, offshore-PHI controls, and the tested escalation route.
Preparation comes before judgment: L1 captures the caller's exact words, location, callback, and relevant source-linked context. A prepared clinical question routes directly to the beneficiary-location-authorized U.S. clinical route; Puerto Rico L2 enters only for an actual or requested nonclinical care-team service or another D-021 human condition, never as a clinical tollbooth.
The current base staffing assumption is 5,000 covered lives, 30% shrinkage, 15% relief/backup reserve, 6.92 coverage FTE, and 0.5 FTE of team-lead/quality capacity. That equals 185.2 paid minutes per beneficiary-year. Administrative exception work may use idle coverage capacity only while one responder remains immediately answerable.
Detailed 24/7 procedure — all 29 tasks#
This procedure expands all 29 tasks, from standing coverage through call handling, interdisciplinary-team follow-through, service recovery, and repeated-call improvement. The preserved inventory below remains the concise denominator; the detailed cards state what the work actually entails and how human labor is separated.
Source and labor boundary#
The current public requirement anchor is the CMS GUIDE Request for Applications, Appendix B 3.1-3.2 (SRC-CMS-GUIDE-RFA-V1), rechecked 2026-07-12. Section 3.1 permits either 24/7 access to an interdisciplinary care-team member or a 24/7 helpline through which beneficiaries/caregivers can speak with a care-team member or an engaged third party providing human support, not artificial intelligence, during off-duty hours. Every off-duty third-party communication must be shared with the interdisciplinary care team. CMS does not prescribe staffing geography, answer-time targets, abandonment thresholds, concurrency reserves, test cadence, caseload, or labor minutes. Those are Proxi operating-design and calibration questions. The executed Participation Agreement remains missing under O-PA-001.
The settled Proxi design is D-014 and D-015: a self-staffed Philippines human first-answer team with Puerto Rico nonclinical care-team support and escalation to a separate U.S. beneficiary-location-authorized clinical workforce. D-005 remains controlling: a Philippines first-answer call is not the formal GUIDE navigator contact. A Philippine RN remains in the nonclinical first-answer lane unless separately authorized for the beneficiary's location and the exact licensed act.
All minutes below are provisional calibration assumptions, not CMS requirements or field measurements. Standing availability and event labor must not be added together:
- At 5,000 covered lives, the base standing-coverage allocation is 185.2 paid Philippines minutes per beneficiary-year, representing 6.92 coverage FTE after shrinkage/relief plus 0.5 FTE team-lead/quality capacity.
- That allocation already absorbs about 21 active Philippine administrative minutes and about 4 active first-answer minutes per beneficiary-year.
- The base call assumption is 0.375 calls per beneficiary-year at about 10 active Philippines first-answer minutes per complete call episode. Task 10 below uses only the pickup/connection slice; later call-handling tasks must allocate the remainder without duplicating the episode.
- Tasks 1-9 are capability, setup, beneficiary-enablement, and quality work. Except for the explicitly bundled task-7 navigator minute estimate, they do not create a new human touch for every beneficiary or every row.
- Beneficiary, caregiver, outside emergency responder, and external-provider time is outside Proxi labor unless Proxi pays that person.
1. Maintain the settled human coverage approach#
When. Continuously for every day that any beneficiary is aligned and entitled to GUIDE access, including nights, weekends, holidays, staff absences, concurrent calls, telecommunications failures, and clinical-escalation needs. This is the operating capability that satisfies the human-support option selected under GUIDE RFA Appendix B 3.1; it is not a beneficiary-triggered encounter.
What and how much. Maintain one immediately answerable Philippines human first-answer seat at every moment, backed by relief and concurrency coverage, a responsible Philippines supervisor, a reachable Puerto Rico care-team route, and a beneficiary-location-authorized clinical route. Keep coverage available to 100% of aligned beneficiaries even though most will not call. Preserve the boundaries that the Philippines responder provides nonclinical first answer and does not replace the navigator, clinical consultation, emergency authority, or care-team-only substantive caregiver-support call.
Data. Use the aligned-life denominator, service dates, primary and backup responder roster, actual sign-in/answer readiness, queue/concurrency data, supervisor coverage, care-team contacts, clinical route by beneficiary location and issue, training/scope status, and approved Participant/Partner Organization/offshore-PHI arrangement. A published schedule proves planned staffing, not actual answerability. Staff logged in does not prove the line reaches them. A clinical phone number does not prove an authorized clinician will answer. A Philippines RN credential does not prove U.S. clinical authority.
Potential Proxi work. Forecast arrivals, build and optimize the rota, detect gaps, call relief staff, monitor primary/backup readiness, display approved responder scope, maintain escalation directories, and create exception reports. Proxi may not replace the human first answer with an AI voice, decide urgency, assign clinical authority from credentials alone, or declare coverage from a plan that was never exercised.
Human role. Philippine L1 workers provide the actual continuous human availability. An approved senior Philippine L1 worker performs shift execution and backup activation inside that same job category. The Puerto Rico GUIDE navigator lane maintains its accepted care-team route, and each scheduled U.S. clinician accepts the configured beneficiary-location clinical route. The System Administrator maintains the approved technical routing. These are pooled tasks performed by existing categories, not separate roles or beneficiary touches.
Provisional clean-path Proxi human minutes by role. This task is the single accounting carrier for the base 185.2 paid Philippines coverage minutes per beneficiary-year at 5,000 lives. It is pooled capacity, not 185.2 minutes of beneficiary interaction. L2, L3, and L4 event minutes are 0 until an actual routed need occurs and are budgeted in their separate pools. Do not add task 2-10 coverage activity again to the 185.2-minute allocation.
Completion evidence. For every covered interval: an actual primary human responder was answer-ready; backup/concurrency coverage and supervisor support were available; care-team and clinical escalation routes were current and reachable; and any gap is recorded with actual duration, affected calls, recovery, and follow-up. Completion is continuous coverage evidence, not one annual manager attestation.
What does not prove completion. An approved staffing model; a vendor contract; 6.92 calculated FTE; a phone number; an AI/IVR greeting; employee login without call routing; a single successful test; a clinical directory with no reachable clinician; or absence of complaints.
Edge cases/open decisions. Existing: X-008, X-010, X-013, X-014, X-015; D-005, D-014; O-001, O-003, O-004, O-039, O-PA-001. Missing edge cases: concurrent calls exceed staffed seats; primary and backup fail together; responder is logged in but unreachable; clinical route lacks authority for the caller's location. Minimum tests: ordinary weekday; 2 a.m. holiday; sudden sick call; simultaneous calls; telecommunications outage; AI-only substitution attempt; Philippines RN asked to triage; clinical route unavailable.
2. Establish one continuously available access route#
When. Before the first beneficiary is placed on the service, continuously thereafter, and after any carrier, number, routing, vendor, staffing, or disaster-recovery change.
What and how much. Operate one stable beneficiary-facing telephone number that can be called at any hour and routes to an actual human. Keep backup routing behind that number so beneficiaries are not expected to choose among multiple operational numbers during an urgent need. Support nights, weekends, holidays, queue overflow, primary-carrier failure, and the beneficiary/caregiver accessibility routes the program promises.
Data. Use active number ownership, carrier configuration, hours, primary/backup destination, queue rules, overflow behavior, caller-ID/callback behavior, accessibility/language route, outage state, and actual connection results. A number being assigned or billable proves ownership only. A carrier success code or IVR answer proves telecommunications connection, not connection to a human. A working primary route does not prove backup or overflow works.
Potential Proxi work. Configure and monitor routing, select the currently available human destination from approved coverage facts, activate backup paths, alert operations, and preserve failed-call/callback information. Proxi may provide routing assistance but cannot be the required human support or hide an AI-only interaction behind a human-sounding voice.
Human role. The System Administrator establishes and restores the approved route; the outside telecommunications vendor repairs its own service; the Philippine L1 worker supplies the human answer; and an approved senior Philippine L1 worker performs backup activation in the same job lane. Software continuously checks that the beneficiary-facing number remains stable and usable. No generic program/operations role, care navigator, or clinician is needed merely to maintain the phone route.
Provisional clean-path Proxi human minutes by role. 0 incremental beneficiary-event minutes for L1/L2/L3/L4. Routine route operation and responder availability are inside task 1's pooled 185.2-minute allocation; carrier/program administration is pooled overhead and is not assigned again per beneficiary. These are calibration assumptions.
Completion evidence. The published number is active and, from every approved entry path, connects to the currently scheduled human or an operating human backup at all hours. Primary, overflow, and backup paths have successful objective checks and failed attempts remain visible until repaired.
What does not prove completion. Buying a number; publishing it; IVR pickup; voicemail availability; an AI chatbot; daytime success; a carrier dashboard showing green; or successful routing that ends in an unstaffed queue.
Edge cases/open decisions. Existing: X-007, X-010, X-013, X-014, X-015; O-004, O-039, O-PA-001. Proposed open decision: exact answer-time, queue, abandonment, overflow, and recovery standards because CMS does not set them. Minimum tests: daytime, overnight, weekend, holiday, no-answer primary, concurrent calls, carrier outage, number port, blocked caller ID, accessibility route, and IVR/AI-only false pass.
3. Create the around-the-clock human coverage schedule#
When. Before coverage begins; for every future scheduling horizon; and immediately after absence, attrition, forecast change, volume spike, holiday, training, or escalation-roster change threatens coverage.
What and how much. Assign a named primary Philippine L1 responder for every minute, named relief/backup coverage for absence and concurrency, an approved senior L1 worker for the supervisory tasks, and current Puerto Rico navigator and separate U.S. beneficiary-location-authorized clinical contacts. Include shift-start acceptance, shift-end handoff, breaks, meetings, leave, training, shrinkage, holiday coverage, and backup activation. Maintain enough reserve that administrative work stops when answerability is threatened.
Data. Use forecast call demand, covered lives, arrival distribution, concurrency, shrinkage, relief reserve, skills/language needs, staff availability, accepted shifts, actual login/readiness, breaks, backup acceptance, and escalation availability. A generated rota does not prove people accepted it or appeared. One primary name does not prove concurrency coverage. A credential list does not prove current scope or availability.
Potential Proxi work. Forecast demand, generate the rota, detect gaps/overlaps, solicit shift acceptance, monitor start/readiness, suspend interruptible admin work when needed, activate relief staff, and escalate an uncovered interval. Software can perform routine scheduling; a manager need only handle real conflicts and accountability.
Human role. An approved senior Philippine L1 worker performs shift acceptance, attendance review, coverage correction, and coaching inside the same L1 job category; responders accept and perform shifts. HR/workforce administration enters only for a real employment or capacity gap. Puerto Rico navigators and scheduled U.S. clinicians accept and maintain their own on-call contacts. Human review is exception-based, not a manual re-creation of every software-generated schedule.
Provisional clean-path Proxi human minutes by role. 0 incremental per-beneficiary or per-call minutes. Responder, relief, and 0.5 FTE team-lead/quality time are already included in task 1's 185.2-minute pooled allocation. Any separate executive/program oversight remains pooled overhead pending O-039, not a human touch on each beneficiary.
Completion evidence. Every interval has an accepted primary responder, usable backup/concurrency path, supervisor, and current care-team/clinical contacts; actual readiness is reconciled against the schedule; and uncovered intervals produce documented recovery rather than silent schedule edits.
What does not prove completion. A spreadsheet with no acceptance; one person scheduled without breaks/relief; a backup name who is already primary elsewhere; staff login without line assignment; a clinician name without location authority; or correcting the schedule after an unanswered call without recording the gap.
Edge cases/open decisions. Existing: X-005, X-013, X-014, X-015; O-003, O-004, O-039. Missing edge cases: simultaneous absence and call surge; backup declines; shift handoff loses an active caller; language-specific responder unavailable. Minimum tests: routine week, holiday, sick call, training overlap, concurrent calls, relief activation, staff no-show, clinical-contact change, and admin work preempted by incoming call.
4. Define what the helpline responder may do#
When. Before any Philippines responder takes a beneficiary/caregiver call and whenever law, Participation Agreement, staffing model, call type, approved information, escalation route, or clinical authority changes.
What and how much. Maintain one approved scope guide for each responder job category. It must state permitted tasks—human first answer, caller/callback capture, verbatim concern capture, approved nonclinical program information, approved prompts, warm transfer, and documentation—and prohibited tasks—clinical triage, urgency determination, symptom/medication interpretation, treatment recommendation, medication reconciliation, legal/capacity decision, formal navigator contact, or care-team-only substantive caregiver support. For each out-of-scope cue, identify the exact clinical, care-team, safety, or same-lane supervisory route and what the responder says while connecting. Software enforces approved privacy/disclosure rules; only unresolved legal authority reaches Healthcare Legal Counsel.
Data. Use controlling Participant terms, role/employment arrangement, actual credentials and beneficiary-location authority, approved scripts/content, call categories, privacy rules, clinical/safety routes, and training/competency evidence. Philippine RN education may support vocabulary and cue recognition but does not prove U.S. nursing authority. A script cannot prove the responder followed it. An AI classification of a concern cannot authorize the responder's action or select urgency.
Potential Proxi work. Assemble approved content, display role-specific allowed/prohibited actions, retrieve the correct route from confirmed facts, capture deviations, and prepare training/quality examples. Proxi can flag cue phrases but must not make the clinical/safety judgment or silently expand scope when no safe route matches.
Human role. Before activation, the Head of Nursing or Chief Medical Officer approves only the clinical boundaries within that leader's scope, Healthcare Legal Counsel answers only unresolved legal-authority questions, and the Participant-designated implementation owner approves the nonclinical operating policy. The approved policy is then enforced in software. An approved senior Philippine L1 worker performs the training, coaching, roster-recovery, and stop-condition enforcement tasks inside the same Philippine L1 job category. The applicable U.S. clinician performs consultation when triggered; no Philippine responder supplies that judgment.
Provisional clean-path Proxi human minutes by role. 0 per-beneficiary and per-call setup minutes. Pre-activation policy approval and Philippine L1 training are pooled governance and workforce tasks, not beneficiary-episode roles. L3/L4 minutes are 0 for a routine nonclinical call and occur only when a real clinical transfer is triggered. No blanket clinical review is added to every call.
Completion evidence. Approved current scope guide; role-specific training and competency evidence; reachable named escalation destinations; responder access to the guide at answer time; and quality evidence that prohibited acts trigger stop/transfer rather than improvised advice.
What does not prove completion. Generic call script; employment as an RN; an AI triage tool; a list of prohibited acts without a reachable alternative; training attendance without demonstrated use; or supervisor belief that staff "know the rules."
Edge cases/open decisions. Existing: X-008, X-009, X-010, X-014, X-015; D-003, D-005; O-001, O-003, O-004, O-032, O-PA-001. Proposed open decision after deduplication: exact 24/7 responder-scope, clinical/safety transfer, failed-handoff, and caller-location policy. Minimum tests: routine program question; symptom report; medication question; possible abuse; confused caller; Philippines RN responder; caller traveling; clinical route fails; AI assigns a confident but wrong category.
5. Require third-party information sharing in the vendor arrangement#
When. Before the GUIDE Participant relies on Proxi or any other engaged third party for off-duty human support; at contract renewal/change; and whenever the actual handoff route, data scope, subcontractor, offshore processing, or service model changes.
What and how much. Put in the controlling arrangement an explicit obligation to share information from every beneficiary/caregiver communication with the interdisciplinary care team, as required by GUIDE RFA Appendix B 3.1. Define the practical handoff: required call facts, responder actions, transfers/escalations, promises, callback needs, timing, secure route, correct team recipient, delivery confirmation, downtime fallback, correction process, retention, and responsibility for failed handoff. Proxi's self-staffed Philippines model may still be a third-party/Partner Organization arrangement from the GUIDE Participant's perspective; self-staffed does not erase the Participant-facing obligation.
Data. Use the executed Participant-Proxi/vendor agreement, approved Partner Organization structure, subcontractor list, BAA/privacy/security terms, offshore-PHI analysis, required handoff content, transmission/receipt evidence, and exception process. A signed clause proves a contractual requirement, not that every call is captured or delivered. Sending a call note proves transmission, not care-team receipt or follow-up. HHS guidance that HIPAA does not categorically prohibit overseas ePHI does not approve the GUIDE contractual arrangement or all offshore acts.
Potential Proxi work. Compare proposed language with the CMS duty, generate a gap list, prepare the operational handoff specification, transmit verified call records, and monitor delivery. AI may draft contract/workflow language for human review but cannot approve legal sufficiency, select the permitted offshore scope, or attest compliance.
Human role. Participant and Proxi contracting/legal/compliance/privacy/security owners approve the arrangement; program operations owns the practical handoff; the responder verifies the call record; the interdisciplinary team accepts follow-up responsibility. These are contract/program duties, not per-beneficiary touches.
Provisional clean-path Proxi human minutes by role. 0 incremental per beneficiary or per call for contract setup. Legal/compliance/program work is pooled overhead pending field allocation under O-039. Per-call verification/transmission belongs to later tasks 21-23 and must not be charged here.
Completion evidence. Executed applicable arrangement; explicit every-communication duty; approved content/timing/channel/recipient/failure process; tested secure handoff with matched care-team receipt; and a correction path for incomplete or inaccurate records.
What does not prove completion. Generic BAA; vendor claim of HIPAA compliance; signed contract without every-communication language; workflow diagram without test; call-note creation without care-team delivery; or an offshore-risk analysis treated as CMS/Participation Agreement approval.
Edge cases/open decisions. Existing: X-001, X-002, X-007, X-013, X-015; O-001, O-004, O-PA-001. Missing edge cases: third-party interaction omitted; wrong beneficiary/team recipient; incomplete clinical/safety facts; delayed handoff across shift; correction arrives later; vendor outage. Proposed open decision: exact handoff content, timing, recipient acknowledgement, retry, and failure thresholds. Minimum tests: ordinary nonclinical call, clinical transfer, emergency call, dropped call, correction, wrong-patient risk, secure-route outage, and subcontractor handling.
6. Give the access number and instructions to the beneficiary and caregiver#
When. At service onboarding after the beneficiary/caregiver route and communication needs are confirmed; whenever the number, instructions, language/accessibility support, or emergency guidance changes; and after evidence that the recipient lacks the current information.
What and how much. Provide one current access package to every beneficiary and each permitted caregiver. It states the stable number, that an actual human answers at any hour, examples of appropriate GUIDE calls, what identity/callback facts may be requested, available language/accessibility support, what happens if care-team or clinical help is needed, and the approved emergency direction. Provide it in each recipient's usable language, format, and channel. Do not make one caregiver's receipt stand in for the beneficiary's own usable access.
Data. Use current number/instructions, beneficiary/caregiver identity and permission, language/accessibility needs, preferred channel, delivery endpoint, delivery result, and recipient confirmation or demonstrated use. A sent message proves an attempt. Delivery proves arrival to an endpoint, not that the correct person can find, understand, remember, or use the number. Prior caregiver permission may be stale/revoked. A test call proves route use at that moment, not future recall.
Potential Proxi work. Generate language/accessibility-matched instructions from approved content, deliver through approved channels, request simple confirmation, remind after nonresponse, detect outdated packages, and route usability problems. AI/text/video/voice can provide the entire clean-path explanation but cannot impersonate the required live responder or infer comprehension from a click alone.
Human role. No navigator is required merely to read or send standard instructions. L1 may resolve failed delivery or guide a simple practice call within nonclinical scope. L2 addresses relationship-sensitive misunderstanding, beneficiary choice, or a no-caregiver safeguard issue during an already-scheduled contact. Interpreter/accessibility support is used when automated supported formats are insufficient.
Provisional clean-path Proxi human minutes by role. L1 0; L2 0; L3 0; L4 0 for successful automated delivery plus recipient confirmation. A conditional clarification should be bundled into an existing L2 contact or the paid Philippines coverage pool, not created as a universal new touch. These are calibration assumptions.
Completion evidence. Current package delivered to the beneficiary and each applicable permitted caregiver in the correct usable format; recipient-specific confirmation or demonstrated access; and any failed/unusable route remains open with a named resolver.
What does not prove completion. Enrollment; number publication; sending one English message; caregiver receipt alone; a portal notification; delivery without accessibility; an "OK" click treated as full comprehension; or the navigator's assumption that the person knows the number.
Edge cases/open decisions. Existing: X-002, X-003, X-005, X-007, X-015; O-004, O-013, O-039, O-PA-001. Proposed open decision: recipient-specific usable-access evidence and supported language/accessibility escalation standard. Minimum tests: beneficiary plus caregiver; caregiver not permitted; no caregiver; changed number; returned mail; inaccessible portal; interpreter need; cognitive limitation; current package delivered but old refrigerator card remains.
7. Give a beneficiary without a caregiver a direct usable route#
When. At onboarding and whenever caregiver status becomes confirmed absent/unavailable, beneficiary communication capability changes, the route changes, or prior use shows the beneficiary cannot access it reliably. GUIDE RFA Appendix B 1.2.3.1 requires additional efforts/safeguards for beneficiaries without caregivers; applying a demonstrated usable 24/7 route is a Proxi operating design, not an exact CMS-scripted safeguard.
What and how much. Give the beneficiary the current number and accessible instructions directly; incorporate the number into the beneficiary's routine or approved assistive format; practice or demonstrate use when helpful; verify what the beneficiary will do if disconnected; and identify any approved additional support needed when they cannot safely use the route alone. Do not condition access on finding a caregiver or infer incapacity from caregiver absence.
Data. Use confirmed caregiver status, beneficiary identity, communication preferences, cognition/function facts already established by authorized assessment, phone/device access, language/accessibility needs, demonstrated route use, and approved safeguard plan. No caregiver does not prove incapacity, unreliability, or inability to call. Delivery does not prove usability. One successful demonstration does not prove safety in every future circumstance.
Potential Proxi work. Tailor and repeat instructions, place approved reminders, support a practice connection, retrieve accommodations, detect failed route use, and present safeguard options already approved for human selection. Proxi cannot decide capacity, assign a caregiver, impose monitoring, or determine that the beneficiary is safe without the route.
Human role. L2 confirms the route is workable and discusses beneficiary preference during the existing onboarding/assessment relationship contact; this should not be a separate appointment. L1 can answer an optional practice call and provide approved nonclinical guidance. L3/L4/social work enters only when actual clinical, capacity, or safeguarding judgment is triggered.
Provisional clean-path Proxi human minutes by role. L2 navigator 2 minutes allocated inside the existing onboarding/assessment contact, not a new standalone touch. L1 practice-call pickup, if used, is part of the existing first-answer coverage/call episode and not added to paid allocation. L3 0; L4 0 clean path. These are calibration assumptions.
Completion evidence. Beneficiary-specific current route and instructions; accessible placement; beneficiary demonstrates or credibly confirms how to reach the human line; failed-use barriers are resolved or have an approved safeguard owner; and absence of a caregiver is not used to deny access.
What does not prove completion. Giving instructions to a nonexistent/old caregiver; mailed card; beneficiary saying "yes" without accessible demonstration where a barrier is known; assuming a smartphone exists; caregiver search still open; or a clinician declaring the person has dementia.
Edge cases/open decisions. Existing: X-003, X-005, X-007, X-008, X-015; O-004, O-011, O-013, O-PA-001. Missing edge cases: no phone/device; memory impairment defeats instruction retention; beneficiary travels; beneficiary capable but declines practice; route works only through inaccessible IVR. Minimum tests: capable beneficiary alone; hearing/vision/language need; no device; failed practice call; beneficiary refuses additional safeguard; later caregiver identified; immediate safety concern during practice.
8. Keep contact and accommodation information available to responders#
When. Before the beneficiary can use the line; at every confirmed change; and at call time after the caller has supplied enough information for the responder to use the permitted record. Recheck after returned communications, caregiver/representative changes, language/accessibility changes, move/travel, or privacy restriction.
What and how much. Make the minimum current responder view available to authorized staff: beneficiary name and preferred form of address; permitted identity cues; preferred language and communication accommodation; callback numbers; approved caregiver/representative contacts and scope; current GUIDE care-team contacts; clinical/safety routes; and an explicit unknown/stale marker for missing facts. Show only information appropriate to the responder's role and caller context.
Data. Use source, date, subject, scope, and verification for every contact/accommodation fact; permission/authority for caregiver/representative facts; responder authorization; and current team/clinical routes. Record availability does not prove caller identity or disclosure permission. A caregiver listing does not prove current authority. Missing accommodation data is unknown, not "none." A care-team contact does not prove current on-call availability.
Potential Proxi work. Retrieve the permitted view at call time, prefer current confirmed facts, flag conflicts/staleness, prevent overbroad display, and create targeted correction requests. AI may summarize accommodations but cannot decide caller authority, disclose PHI, or fill missing facts from inference.
Human role. Operations/privacy configures access and corrects authorization failures; L2 authenticates beneficiary/caregiver preference changes during normal contacts; L1 captures a caller's verbatim correction as pending and uses approved safe communication while it is resolved. No routine human chart preparation is needed before each call.
Provisional clean-path Proxi human minutes by role. L1 0; L2 0; L3 0; L4 0 per call when current data is retrieved automatically. Corrections occur inside the originating contact or exception pool and must not create a universal reviewer touch. These are calibration assumptions.
Completion evidence. Authorized responder can retrieve current, source-labeled contact/accommodation/team-route information when needed; stale/conflicting/unknown facts are explicit; and correction/access failures have an owner without blocking immediate safe human connection.
What does not prove completion. Data existing somewhere in the EHR; a one-time export; caregiver name without scope; old language preference; a generic "no accommodations" default; broad offshore record access; or the responder seeing information before caller context allows its use.
Edge cases/open decisions. Existing: X-001, X-002, X-003, X-004, X-005, X-009, X-013, X-015; O-001, O-011, O-013, O-PA-001. Missing edge cases: stale callback number; revoked caregiver access; conflicting preferred language; wrong-beneficiary match; responder cannot access accommodation during outage. Minimum tests: current clean record; missing language; revoked representative; duplicate beneficiary; temporary travel; changed care team; outage; caller reports a correction during urgent call.
9. Check that the line and human-answer route are working#
When. At an approved recurring cadence; after every number, carrier, route, roster, vendor, language/accessibility, backup, or clinical-contact change; after any failed/abandoned call signal; and after recovery. CMS requires the resulting 24/7 human access but does not set this test cadence or performance threshold.
What and how much. Exercise the complete route, not merely the carrier: published number, primary routing, actual human answer, overflow/concurrency, backup, nights/weekends/holidays, promised language/accessibility route, callback capture, supervisor path, and care-team/clinical transfer reachability. Each check records the route exercised and what actually happened. Automated monitoring may run continuously; sampled end-to-end human-answer and transfer checks verify the parts machines cannot prove alone.
Data. Use monitor signals, synthetic-call results, call-detail records, human-answer identity/time, queue/abandonment, transfer outcome, route configuration, roster/readiness, accessibility result, and recovery evidence. A line ping, SIP success, IVR answer, or AI greeting does not prove human access. One responder answer does not prove overflow, backup, language, or clinical-transfer readiness. Absence of an alert does not prove the monitoring itself works.
Potential Proxi work. Continuously monitor telecommunications and queue status, initiate approved test calls, compare schedule with actual answer, test backup after change, detect failure, suspend nonurgent admin work, alert operations, and track recovery. Software should perform routine checks; people investigate exceptions and participate in sampled human/transfer tests.
Human role. Philippine L1 workers answer sampled checks as part of standing coverage; an approved senior L1 worker performs roster recovery; the System Administrator restores configuration; Software Engineering enters only for a product defect; and the outside telecommunications vendor repairs its service. Puerto Rico navigators and scheduled U.S. clinicians participate only in appropriately controlled transfer tests. No generic operations manager or clinician reviews every line check.
Provisional clean-path Proxi human minutes by role. 0 incremental per beneficiary. Automated checks are software; responder/supervisor sampled-test time is included in task 1's paid coverage and quality allocation. L2/L3/L4 are 0 except for separately planned limited escalation-route validation, which is pooled readiness work rather than beneficiary-event labor.
Completion evidence. Dated route-specific checks cover primary, overflow, backup, off-duty, and promised accommodation paths; an actual human answer is objectively distinguished from IVR/AI; failures trigger recovery and retest; and performance is assessed against a settled operational standard.
What does not prove completion. Green carrier dashboard; dial tone; IVR pickup; chatbot response; daytime-only test; one successful call; test bypassing the public number; backup configured but untested; or failure alert without recovery verification.
Edge cases/open decisions. Existing: X-007, X-010, X-013, X-014, X-015; O-003, O-004, O-039. Proposed open decision: monitoring cadence, sampled end-to-end coverage, answer/abandonment target, concurrency load, recovery ceiling, and objective pass evidence. Minimum tests: primary human answer; overnight/holiday; overflow; primary failure/backup; both routes fail; AI/IVR false answer; language route; accessibility route; callback; clinical transfer; monitor itself disabled.
10. Receive the call at any time#
When. On every incoming beneficiary or caregiver call, at any time and regardless of whether identity, caller authority, concern category, or clinical urgency is initially clear. Do not delay human answer while routine identity facts are resolved.
What and how much. Route the call to and establish a live connection with an actual interdisciplinary care-team member or the permitted Philippines human first-answer responder. Preserve the call start, queue/route, human answer, callback information available at connection, and any disconnection. Task 10 ends when human connection is established or a failed-answer recovery obligation is opened; caller identity, concern capture, emergency recognition, clinical transfer, substantive support, documentation, and follow-up belong to tasks 11 onward.
Data. Use incoming call event, public number, route/queue, answer time, responder identity/role, human-versus-automated indicator, connection/disconnection, callback identifier if available, and backup activation. IVR/AI pickup is not human answer. Ringing is not answer. Voicemail is not live support. A human answer proves access for that call, not identity verification, suitable responder scope, concern resolution, clinical consultation, caregiver-support delivery, documentation, or follow-up.
Potential Proxi work. Select the currently available approved human route, place the call in queue, preserve callback metadata, detect drop/no-answer, invoke backup, and provide the responder with permitted context. Proxi cannot speak as the required human, classify urgency as a decision, or claim the concern resolved when connection begins.
Human role. L1 Philippines responder or an interdisciplinary care-team member must actually answer. The L1 responder remains nonclinical and invokes the later clinical/safety/care-team pathways when cues arise. L2/L3/L4 does not need to touch a routine first-answer event unless the content requires that role.
Provisional clean-path Proxi human minutes by role. L1 Philippines first answer 1 active minute per answered call for the pickup/connection portion. At the base 0.375 calls per beneficiary-year, this is about 0.4 active minute per beneficiary-year, but it is already inside the approximately 10-minute full call episode, the approximately 4 active first-answer minutes per beneficiary-year, and task 1's 185.2 paid coverage allocation. It is not added again. L2 0; L3 0; L4 0 for routine pickup.
Completion evidence. An actual named human responder establishes live communication, or the failed/no-answer event opens a visible recovery path with callback/backup evidence. The human-answer timestamp and role are retained separately from IVR/AI events.
What does not prove completion. Call arrival; ringing; queue entry; automated greeting; AI voice; voicemail; abandoned call; responder login; a later call note without answer evidence; or eventual clinical resolution of another call.
Edge cases/open decisions. Existing: X-001, X-002, X-007, X-008, X-013, X-014, X-015; D-005, D-014; O-001, O-003, O-004, O-032, O-039, O-PA-001. Missing edge cases: two calls arrive simultaneously; caller disconnects before answer; no callback identifier; unknown caller reports immediate danger; responder becomes unavailable mid-route. Minimum tests: routine office-hour call; 2 a.m. call; concurrent calls; primary no-answer/backup success; complete outage; AI/IVR intercept; caller hangs up; clinical cue on answer; unknown caller with safety concern.
Active-call episode and labor boundary for tasks 11-20#
The settled D-005/D-014/D-015 boundary controls this call sequence: L1 Philippines provides human first answer but not formal navigator contact or clinical triage; L2 Puerto Rico provides nonclinical navigator/care-team work only; every clinical or medical act goes to the separate U.S. beneficiary-location-authorized L3/L4 workforce. The approximately ten active L1 minutes in an ordinary first-answer episode are allocated rather than added repeatedly:
- Task 10 carries about 1 L1 minute for pickup and live connection.
- Task 12 carries about 2 L1 minutes for caller, beneficiary, and callback establishment.
- Task 13 carries about 2 L1 minutes for source-faithful concern capture and bounded clarification.
- The remaining 5 L1 minutes are one route-specific envelope across the primary work in tasks 14-19. Mixed concerns remain open and are routed, but the same five minutes are not charged once for every route label.
- Task 11 recovery is conditional and may consume part of that route envelope. Longer recovery is retained as observed high-tail work rather than discarded.
- These active minutes are already inside the approximately four active first-answer minutes per beneficiary-year and the 185.2 paid Philippines coverage allocation; they are not added again as incremental paid labor.
- Separate U.S. L3/L4 event durations are recorded separately. Population-average allocations are not per-contact duration caps.
- For every transfer, preserve attempt, answer/acceptance, authorized decision or service, instruction or action delivery, and caller understanding as separate facts.
The full 24/7 episode has two linked parts. The caller-facing part begins with actual human answer and continues through enough identity for safe communication, the caller's stated need, and substantive human support: an approved nonclinical answer, direct caregiver support, or a verified accepted handoff with a clear next step and return expectation. Pickup, intake, voicemail, attempted transfer, message transmission, or queue assignment alone is not the completed substantive service.
The after-call part carries the verified interaction to the interdisciplinary team and completes any promised callback or entry into the underlying GUIDE service. The call may end while that downstream obligation remains open. Later minutes belong to the role and service that actually perform the work; the same L1 intake is not counted again across every downstream task. An accepted handoff proves that a human endpoint owns the issue. It does not prove that the underlying clinical, caregiver-support, navigation, or other GUIDE service is complete.
11. Recover a disconnected or unanswered call#
When. When the caller disconnects before support is complete, the routed human does not answer, a transfer drops, or the connection becomes unusable after a callback route is available. Invoke backup immediately according to the approved urgency/SLA policy; do not wait for a routine callback when an active safety concern is known.
What and how much. Preserve the callback number and the point at which the call failed; attempt reconnect through the primary responder or backup human; tell the caller what happened when reconnected; restore the unfinished support at the correct step; and retain every failed attempt. If no callback route exists, preserve what is known and use only approved alternate/emergency mechanisms rather than guessing identity/location.
Data and provenance. Use incoming call/transfer identifiers, caller-supplied callback route, disconnect/no-answer time, last known concern, current location if reported, responder/backup availability, attempts, answer result, and any safety/clinical cue. A callback attempt is not reconnection. Ringing, voicemail, or delivered SMS is not live support. A later call from a similar number is not automatically the same caller. Reconnection does not prove the original concern was resolved.
Potential Proxi work. Detect dropped/no-answer events, preserve call context, dial the approved callback/backup route, notify the responder/supervisor, suppress duplicate attempts, and resume the prepared human task. Proxi cannot decide urgency, improvise an alternate recipient, or mark recovery successful from a delivery signal.
Human role. L1 resumes the call and the Philippines support lead activates approved backup; software executes the configured recovery and failover rules. A known or possible urgent clinical or safety concern goes directly to the beneficiary-location-authorized U.S. clinician or emergency route. Suspected wrong-person call-record linkage goes to the Participant or telephony/EHR operator's authorized records administrator. Routine recovery never routes to a generic privacy or safety role.
Provisional human minutes (low / typical / high).
- L1 Philippines: 0 / 2 / 6 minutes per recovery episode. The typical two minutes comes from the same route envelope; the high tail covers repeated callback/backup attempts and supervisor help.
- L2 Puerto Rico nonclinical navigator: 0 / 0 / 0 minutes for connection recovery itself; any resumed navigator work is recorded under the substantive task.
- Separate U.S. L3 clinical: 0 / 0 / 0 minutes for recovery itself.
- Separate U.S. L4 clinical: 0 / 0 / 0 minutes for recovery itself. Any clinical work after reconnection is recorded under task 17 or the underlying clinical service.
Completion evidence. Live human reconnection with matched caller/call context or a truthful unresolved recovery record showing attempts, backup escalation, remaining concern, and next owner; safety/clinical handoff status remains separately evidenced.
What does not prove completion. Callback dialed; voicemail; text delivered; backup notified; caller later reaches another line; connection restored with no context; or supervisor closes the event because the number was unavailable.
Edge cases, open decisions, and minimum tests. Existing: X-001, X-003, X-007, X-008, X-013, X-014, X-015, CC-E13; O-004, O-032, O-039. Missing edge cases: no callback number; number belongs to another person; transfer drops during emergency; simultaneous reconnect attempts; repeated abandonment. Minimum tests: routine dropped call; primary no-answer/backup answers; voicemail only; wrong callback; no callback; safety concern; caller reconnects independently; network outage.
12. Establish who is calling and whom the call concerns#
When. Immediately after human connection and before disclosing person-specific information, while preserving the rule that immediate emergency/safety help is not withheld solely because routine identity cannot yet be resolved.
What and how much. Ask whether the caller is the beneficiary, caregiver, representative, service provider, or another person; identify which beneficiary the call concerns; obtain a reliable callback route; establish the caller's relationship and the minimum authority/permission needed for the conversation; and separate the caller's own support need from a beneficiary-facing need. Use only the information permitted for that caller/context.
Data and provenance. Use caller-stated identity, approved verification factors, beneficiary identifiers, callback route, caregiver/representative relationship and scope, permission/revocation facts, current call concern, and any emergency exception facts. Caller ID proves a device/number, not a person. Caregiver status does not prove representative authority. Portal proxy/emergency contact does not prove authority for every decision/disclosure. Identity match does not prove the caller's report is clinically accurate. Unverified identity does not prove malicious intent.
Potential Proxi work. Guide bounded identity matching, retrieve only permitted records, show caller/beneficiary relationship facts, capture callback, flag conflict, and limit content pending resolution. AI cannot decide authority, disclose PHI, infer the beneficiary from voice/topic, or delay an emergency connection while seeking perfect administrative proof.
Human role. Software applies the approved identity, relationship, authority, and disclosure rules and limits information accordingly. L1 conducts the identity and callback conversation; L2 handles relationship-sensitive participation questions. If the approved rule cannot determine the permitted recipient, purpose, authority, content, or restriction, hold that disclosure and route only the legal-authority question to healthcare legal counsel. Beneficiary-location-authorized U.S. clinical or emergency staff own urgent clinical and safety work.
Provisional human minutes (low / typical / high).
- L1 Philippines: 1 / 2 / 5 minutes inside the active call episode; the high case covers identity conflict, missing callback, or parallel emergency-safe handling.
- L2 Puerto Rico nonclinical navigator: 0 / 0 / 5 minutes only when a relationship-sensitive participation question needs navigator involvement.
- Separate U.S. L3 clinical: 0 / 0 / 0 minutes for identity work.
- Separate U.S. L4 clinical: 0 / 0 / 0 minutes for identity work. Clinical staff are not a substitute for privacy/authority resolution.
Completion evidence. Caller category, matched or explicitly unresolved identity, beneficiary concerned, callback route, relationship/authority scope, and disclosure boundary recorded; urgent/safety work proceeds without falsifying identity certainty.
What does not prove completion. Caller ID; name/DOB only; caregiver label; family relationship; familiar voice; prior call; device possession; emergency contact record; or an identity score without approved verification.
Edge cases, open decisions, and minimum tests. Existing: X-001, X-002, X-003, X-004, X-009, X-015, CC-E04; O-001, O-004, O-011, O-PA-001. Minimum tests: beneficiary; verified caregiver; caregiver without authority; valid representative; provider caller; unknown caller; wrong beneficiary match; distressed caller; immediate danger before verification; shared phone; caller refuses callback.
13. Ask what help is needed now#
Residual-ambiguity rule. If any possibility remains that the caller is reporting a symptom, medication effect or question, clinical or behavioral-health concern, or safety uncertainty, L1 does not label urgency or continue to Task 16. L1 preserves the words, location, and callback and enters Task 15 or Task 17's configured U.S.-clinician or emergency route. The receiving authorized actor, not L1, chooses the clinical or emergency disposition.
When. After live human answer and enough caller context to communicate safely, or in parallel with minimal identity work when delay could be unsafe. Let the caller begin in their own words before narrowing the issue.
What and how much. Invite the concern; capture the caller's words; ask only the bounded clarifying questions needed to understand the requested help and route it safely; identify whether the caller seeks approved program/navigation information, caregiver support, GUIDE service follow-up, or may be reporting clinical/safety content. Preserve mixed concerns and route each without forcing the call into one category.
Data and provenance. Use verbatim concern, speaker identity/status, beneficiary concerned, time, current location if relevant, symptoms/medication/safety words as reported, desired immediate help, and callback. Caller report proves what was reported, not diagnosis, causality, urgency, current medication truth, or recommended action. AI category/confidence is a candidate aid, not a routing/clinical decision. Absence of an explicit keyword does not prove no risk.
Potential Proxi work. Transcribe with permission, preserve source language, display bounded prompts, organize mixed concerns, retrieve approved routing choices from confirmed facts, and prepare a concise handoff. Proxi cannot determine urgency, diagnose, answer clinical questions, suppress a concern, or decide that one route closes another.
Human role. L1 listens, asks approved clarifying questions, reflects the concern, and recognizes stop/transfer cues without clinical interpretation. L2 provides relationship/service support when routed. The separate U.S. L3/L4 or emergency authority assesses clinical/safety content. The caller defines the concern and desired help.
Provisional human minutes (low / typical / high).
- L1 Philippines: 1 / 2 / 5 minutes inside the active call episode for source-faithful concern capture and bounded clarification.
- L2 Puerto Rico nonclinical navigator: 0 / 0 / 0 minutes at first-answer intake; later navigator work belongs to the routed service.
- Separate U.S. L3 clinical: 0 / 0 / 0 minutes until an actual clinical route is triggered.
- Separate U.S. L4 clinical: 0 / 0 / 0 minutes until an actual clinical route is triggered.
Completion evidence. Verbatim/source-attributed concern; bounded clarifications; mixed issues retained; caller's desired immediate help; and a safe route/next question selected by approved procedure. The eventual answer/transfer/decision remains separate.
What does not prove completion. Speech-to-text alone; AI category; single checkbox; responder paraphrase replacing original words; asking only yes/no questions; clinical interpretation; or caller silence treated as no need.
Edge cases, open decisions, and minimum tests. Existing: X-003, X-004, X-008, X-009, X-010, X-015, CC-E10, CC-E11, CC-E13; O-003, O-004, O-032. Minimum tests: routine program question; mixed navigation/clinical concern; vague distress; medication question; possible abuse; caller changes topic; communication impairment; interpreter needed; AI misclassification; explicit emergency.
14. Respond to a beneficiary calling without a caregiver#
When. When the caller is the beneficiary and no caregiver is involved/available for this call, regardless of whether a caregiver exists elsewhere. Apply dementia-capable communication while continuing the appropriate route in tasks 15-18.
What and how much. Speak directly to the beneficiary; use preferred form of address, language, pace, repetition, short steps, and approved accommodations; provide one instruction/action at a time; ask the beneficiary to explain back the essential nonclinical action; and bring in approved additional human support when the beneficiary cannot use the information safely. Do not make access conditional on caregiver presence and do not infer incapacity from dementia or caregiver absence.
Data and provenance. Use beneficiary identity, communication/accommodation preferences, current caregiver availability for the call, verified representative/safeguard facts, exact approved instruction, teach-back response, current location if safety/clinical issue, and failed-understanding evidence. “No caregiver on call” does not prove no caregiver exists. Dementia diagnosis does not prove incapacity. Saying yes/repeating words does not prove understanding. Failed teach-back does not itself decide capacity or clinical safety.
Potential Proxi work. Retrieve communication preferences, simplify approved nonclinical text, pace prompts, display teach-back questions, preserve the beneficiary's response, and summon the approved support route. AI may assist the human but cannot replace the human responder, determine comprehension/capacity, or select clinical action.
Human role. L1 provides dementia-capable human first answer within nonclinical scope. L2 handles additional nonclinical support/safeguard choice. The separate U.S. L3/L4 or emergency personnel enters when clinical, capacity, or safety judgment is needed. The beneficiary remains the participant unless decision-specific authority says otherwise.
Provisional human minutes (low / typical / high).
- L1 Philippines: 1 / 2 / 5 minutes from the one route envelope for accessible communication and teach-back.
- L2 Puerto Rico nonclinical navigator: 0 / 0 / 10 minutes only when nonclinical safeguard choice or relationship support is actually needed.
- Separate U.S. L3 clinical: 0 / 0 / 15 minutes only for a triggered clinical, safety, behavioral, or decision-specific capacity question within scope.
- Separate U.S. L4 clinical: 0 / 0 / 20 minutes only when higher-clinician assessment or treatment judgment is required. Puerto Rico performs none of this clinical work.
Completion evidence. Beneficiary received human communication in a usable form; essential approved nonclinical action is explained back accurately or an additional human support route is actually engaged; uncertainty remains explicit; the substantive route outcome is separately recorded.
What does not prove completion. Caregiver absent; beneficiary says “yes”; responder repeats louder; written message sent; AI voice; caregiver later notified; diagnosis; or responder opinion that the beneficiary is incapable.
Edge cases, open decisions, and minimum tests. Existing: X-002, X-003, X-005, X-007, X-008, X-009, X-015, CP-E06, CP-E07; O-004, O-011, O-013, O-032. Missing edge cases: beneficiary traveling alone; aphasia/hearing/language barrier; fluctuating ability; no device/callback; beneficiary refuses added support. Minimum tests: clear understanding; repeated failed teach-back; no caregiver exists; caregiver temporarily absent; capable beneficiary declines helper; immediate danger; clinical question; interpreter unavailable.
15. Identify an immediate emergency or safety threat#
When. Whenever any call statement or context suggests immediate danger, medical emergency, abuse/neglect/exploitation, wandering, fire, violence, self-harm, severe caregiver crisis, or another issue that may not safely wait for routine follow-up. This route preempts ordinary navigation while preserving other needs for later work.
What and how much. The L1 responder recognizes explicit approved emergency/safety stop cues, keeps the caller connected when possible, gathers only the minimum facts needed for safe connection, and invokes the approved emergency/safety/clinical route. Ambiguous symptoms/urgency are transferred to the beneficiary-location-authorized clinician; L1 does not decide clinical urgency. Track attempted connection, answered/accepted handoff, emergency/clinical decision, instructions issued by the authorized actor, and caller understanding separately.
Data and provenance. Use verbatim report, person at risk, current location, callback, immediate circumstances, time, caller identity status, approved route, transfer attempts/answer, receiving actor, and authorized disposition. Keyword/AI alert is a candidate signal, not proof of emergency. L1 recognition of an explicit emergency supports route invocation but is not a clinical triage conclusion. Attempt is not answer; answer is not clinical decision; decision is not caller understanding or emergency outcome.
Potential Proxi work. Surface approved stop cues, display the configured route, preserve location/callback/context, initiate warm transfer, alert on failure, and document each evidence boundary. Proxi cannot determine urgency, choose 911/988/APS/clinical destination from probabilistic output, issue clinical instruction, or close the emergency.
Human role. L1 invokes/maintains the connection under protocol. Emergency services, 988, APS, the separate U.S. beneficiary-location-authorized L3/L4 workforce, or another approved authority assesses/acts within scope. Supervisor retries failed handoff. L2 handles later nonclinical relationship/navigation follow-up but does not triage.
Provisional human minutes (low / typical / high).
- L1 Philippines: 2 / 5 / 10 minutes from the one route envelope; the high case covers maintaining the caller connection and repeated backup attempts.
- L2 Puerto Rico nonclinical navigator: 0 / 0 / 5 minutes only for later nonclinical relationship or service follow-up, never triage.
- Separate U.S. L3 clinical: 0 / 0 / 20 minutes when an ambiguous clinical/safety concern requires RN, LCSW, or behavioral-clinician assessment within authority.
- Separate U.S. L4 clinical: 0 / 0 / 25 minutes when NP/PA/CNS/physician-level assessment or treatment judgment is required. External emergency-service time is not Proxi labor.
Completion evidence. Verified answered/accepted handoff to the proper human route or a truthful failed-handoff record with immediate retry/backup; authorized decision/instruction and caller understanding recorded separately if they occur; routine community/program issues remain open for later follow-up.
What does not prove completion. AI risk score; responder says “this is urgent”; emergency number displayed; transfer dialed; ringing/voicemail; clinician answers but gives no disposition; instruction delivered without understanding; or safety handoff treated as resolution of the whole call.
Edge cases, open decisions, and minimum tests. Existing: X-001, X-003, X-008, X-009, X-010, X-013, X-014, X-015, CC-E10, CC-E13, CC-E43; O-003, O-004, O-032. Minimum tests: obvious emergency; ambiguous symptom; abuse; wandering; self-harm; caregiver panic; unknown location; caller disconnects; emergency route rejects; clinical route answers/no decision; caller cannot teach back.
17. Connect a clinical issue to an appropriate clinician#
When. Whenever the caller asks for or reports symptoms, medication advice/effects, clinical assessment, diagnosis, treatment, individualized safety judgment, or another act outside L1/L2 nonclinical scope. Do not route automatically to “an RN” without verifying the exact act, beneficiary location, license/scope, and configured clinical responsibility.
What and how much. Preserve the reported facts verbatim; capture current location/callback and minimal relevant context; identify the configured beneficiary-location-authorized clinical route; explain the transfer; attempt a warm connection; remain until accepted when policy requires; and invoke backup/urgent pathway if the route fails. Record attempted transfer, answered/accepted transfer, clinical assessment/decision, approved instruction, and caller understanding separately.
Data and provenance. Use caller/beneficiary identity status, verbatim clinical concern, current physical location/time, relevant approved record context, proposed clinician identity/credential/location authority/scope, transfer attempts/answer, clinical disposition, exact instruction source, and teach-back. Clinical cue proves transfer need, not diagnosis/urgency. Clinician answer proves connection, not assessment/decision. A clinical decision proves only what that clinician authorized. Instruction delivery does not prove understanding or implementation.
Potential Proxi work. Assemble the source-faithful handoff, verify configured route facts, initiate transfer, show failed-route fallback, record events, and deliver only exact approved clinician communication. Proxi cannot triage, select a clinician from title alone, interpret the response, alter instructions, or close the clinical issue.
Human role. L1 prepares and connects without clinical interpretation. The separate U.S. beneficiary-location-authorized L3/L4 or other appropriately licensed clinician performs the assessment and decision within scope. Supervisor retries failed connection. L2 handles related nonclinical navigation/relationship follow-up but does not substitute for clinical consultation.
Provisional human minutes (low / typical / high).
- L1 Philippines: 3 / 5 / 10 minutes from the one route envelope for source-faithful preparation, warm-transfer attempts, and failed-route recovery.
- L2 Puerto Rico nonclinical navigator: 0 / 0 / 5 minutes only for a separate nonclinical relationship/navigation need; L2 never performs the clinical act.
- Separate U.S. L3 clinical: 0 / 8 / 25 minutes for a triggered RN/LCSW/behavioral-clinician assessment within location and scope.
- Separate U.S. L4 clinical: 0 / 0 / 30 minutes when the specific act requires NP/PA/CNS/physician-level authority. The portfolio 1.5 L3 and 0.9 L4 minutes per beneficiary-year are utilization averages, not per-transfer duration caps.
Completion evidence. Appropriate clinician route and beneficiary-location authority verified; transfer attempt recorded; actual clinician answer/acceptance separately confirmed; clinical decision/instruction and caller understanding separately recorded when obtained; failed route remains open with backup/escalation.
What does not prove completion. Clinical number dialed; voicemail; generic RN available; clinician picks up but declines/has no authority; note sent; AI triage summary; decision without exact instruction; instruction delivered but not understood.
Edge cases, open decisions, and minimum tests. Existing: X-001, X-003, X-004, X-008, X-009, X-010, X-013, X-014, X-015, CC-E10, CC-E11, CC-E12, CC-E13, CC-E43; O-003, O-004, O-028, O-032, O-054. Minimum tests: symptom; medication question; PR clinician/Florida beneficiary; beneficiary traveling; unknown location; RN scope insufficient; clinician answers/no decision; transfer failure; conflicting instruction; caller cannot understand approved response; caller reports a possible symptom or medication effect but declines clinical transfer. For the refusal case, preserve the exact refusal and stated reason, do not infer safety or provide clinical advice, apply the approved refusal/escalation and recontact rule, keep the concern and callback owner open, and route any independent explicit emergency cue. Exact timing remains governed by O-004, O-032, and O-054.
18. Connect a service or support issue to the GUIDE team#
When. When the concern requires a care-plan, navigator, community-service, medication-support, transition, caregiver, or other GUIDE-service action that the first-answer responder cannot complete within approved scope, and no unresolved emergency/clinical route preempts it.
What and how much. Identify the receiving GUIDE service/role from approved routing facts; explain the handoff to the caller; preserve the caller's concern, desired help, callback, and any promised timing; attempt live transfer when available or create a clear owned callback/work request; and confirm that a named team destination accepted responsibility. Keep attempted notification, team acceptance, actual callback/service delivery, and problem resolution separate.
Data and provenance. Use verbatim concern, service category, beneficiary/caller facts, desired help, current team assignment/contact, urgency determined only by authorized process, attempted/answered transfer, acceptance/owner, promised callback, and later service evidence. Work-item creation proves routing attempt, not team receipt. Team acknowledgement proves ownership, not service delivery. Callback promise does not prove callback. GUIDE-team contact does not prove navigator qualifying contact.
Potential Proxi work. Map approved issue types to current team destinations, assemble the handoff, notify/transfer, schedule callback, monitor acceptance, and alert on failure. AI may suggest the likely service but cannot determine clinical route, assign unconfirmed ownership, or close the underlying need.
Human role. L1 explains and initiates the handoff. The receiving L2/clinical/social-work/pharmacy/other GUIDE team member accepts and later performs the underlying service. L2 is not charged merely because a work item was created; event time begins with actual acceptance/callback/substantive work.
Provisional human minutes (low / typical / high).
- L1 Philippines: 3 / 5 / 10 minutes from the one route envelope for explanation, live-transfer attempt, or owned callback creation.
- L2 Puerto Rico nonclinical navigator: 0 / 0 / 3 minutes only to accept/clarify ownership; the underlying navigator service is timed in its own SOP.
- Separate U.S. L3 clinical: 0 / 0 / 3 minutes only to accept a clinical work item; substantive clinical work is timed separately.
- Separate U.S. L4 clinical: 0 / 0 / 3 minutes only to accept a higher-clinician work item; substantive work is timed separately. These role branches are alternatives, not a stack.
Completion evidence. Attempted route, actual answer/acceptance or owned callback task, named team owner, promised timing, and caller explanation retained; actual callback/service/resolution remains separately open.
What does not prove completion. AI category; message sent; generic queue; email delivery; team notification; callback scheduled; L2 name assigned; safety handoff; or underlying service performed.
Edge cases, open decisions, and minimum tests. Existing: X-003, X-004, X-006, X-007, X-010, X-013, X-014, X-015, CC-E03, CC-E10, CC-E13; O-003, O-004, O-009, O-025. Minimum tests: navigator issue; community referral; medication-support issue; transition; wrong team; no team response; callback promised; live transfer; mixed clinical/service issue; caller declines transfer.
19. Receive or initiate an ad hoc caregiver support call#
When. When a caregiver, beneficiary, or GUIDE Participant initiates a request for the caregiver's one-on-one support, at any hour for helpline receipt. Use the first-answer route to establish the request and connect/schedule the care-team member who must deliver the substantive support under GUIDE RFA Appendix B 8.3.4.
What and how much. Receive or create one caregiver-support request; identify the caregiver, beneficiary relationship, caregiver's own issue/goal, callback, immediate safety/clinical cues, communication needs, and preferred/available support timing; explain that a care-team member provides substantive support; and establish live connection or an accepted callback owner. Keep receipt/initiation, care-team connection, substantive conversation, clinical consultation, and follow-up separate.
Data and provenance. Use initiator, caregiver identity/relationship, callback, caregiver-owned concern, beneficiary information implicated, permission/disclosure boundary, time, urgent/clinical cues, requested modality, attempted/answered care-team connection, and callback acceptance. Request receipt does not prove substantive support. Caregiver relationship does not prove representative authority over beneficiary decisions. A Philippines responder conversation does not satisfy the care-team-only call merely because it is empathetic/helpful.
Potential Proxi work. Recognize/create the request, collect source-faithful context, schedule/connect, retrieve prior caregiver needs within permitted scope, and monitor acceptance. AI/L1 may prepare and humanely receive the request but cannot replace the required care-team conversation, make clinical judgments, or infer authority.
Human role. L1 provides human first answer and connection. L2 Puerto Rico navigator or another approved interdisciplinary care-team member accepts and delivers task 20. The separate U.S. L3/L4 workforce enters for clinical/safety content. The caregiver speaks for the caregiver's own support need.
Provisional human minutes (low / typical / high).
- L1 Philippines: 3 / 5 / 10 minutes from the one route envelope to receive the request and establish a live connection or owned callback.
- L2 Puerto Rico nonclinical navigator: 0 / 0 / 3 minutes only for live acceptance; task 20 carries the substantive care-team support conversation.
- Separate U.S. L3 clinical: 0 / 0 / 0 minutes for request receipt; clinical content routes separately.
- Separate U.S. L4 clinical: 0 / 0 / 0 minutes for request receipt; clinical content routes separately.
Completion evidence. Request/initiator/caregiver concern captured; actual care-team answer/acceptance or owned callback established; permission/safety issues routed; substantive support status remains separately pending until task 20 occurs.
What does not prove completion. Request logged; L1 provides sympathy; callback message; care-team notification; appointment scheduled; AI coaching; caregiver receives a resource link; or clinical transfer alone.
Edge cases, open decisions, and minimum tests. Existing: X-001, X-002, X-003, X-007, X-008, X-013, X-014, X-015, CC-E10; D-005, D-014; O-001, O-004, O-011, O-032, O-PA-001. Minimum tests: caregiver-initiated; beneficiary-initiated; Participant-initiated; unverified caregiver; caregiver distress; clinical question; beneficiary-sensitive information; care-team no answer; scheduled callback; L1 attempts to deliver full support.
20. Deliver the caregiver's one-on-one support call#
When. After task 19 establishes the request and an approved interdisciplinary care-team member is connected or calls back. The conversation occurs as the caregiver issue arises and may be initiated by the caregiver, beneficiary, or Participant. GUIDE RFA Appendix B 8.2.4/8.3.4 requires direct care-team delivery.
What and how much. A care-team member speaks directly with the caregiver about the current caregiver need. Within training/scope, provide relationship-based listening and approved coaching on stress management, self-care/well-being, behavioral challenges, functional changes, caregiving tasks, and beneficiary/caregiver safety; identify what the caregiver wants next; confirm the essential agreed nonclinical action; and route clinical, safeguarding, or other GUIDE-service work. Do not force every call into a clinical encounter or treat all caregiver distress as clinical illness.
Data and provenance. Use caregiver identity/relationship, caregiver-owned concern/goals, caregiver assessment and prior needs where permitted/current, beneficiary context allowed for the purpose, care-team responder identity/training, exact support provided, caregiver response/teach-back, choices, promises, and clinical/safety routes. Care-team connection proves availability, not substantive support. Advice spoken proves delivery, not understanding/use. Caregiver report does not prove beneficiary clinical fact. Coaching does not prove issue resolution.
Potential Proxi work. Prepare source-labeled prior context, suggest approved topic prompts, transcribe with permission, draft a recap/action list, schedule follow-up, and route child work. AI cannot be the care-team member, decide clinical/safety issues, select caregiver choices, or claim understanding/resolution.
Human role. L2 Puerto Rico navigator is the default lower-cost care-team member for permissible nonclinical coaching/support. Another trained approved care-team member may deliver within scope. The separate U.S. L3/L4/behavioral-clinician workforce enters only for clinical, high-distress, safety, abuse/neglect, or other licensed issues. L1 does not satisfy this substantive call unless separately established as an approved care-team member under the controlling agreement—currently not assumed.
Provisional human minutes (low / typical / high).
- L1 Philippines: 0 / 0 / 0 minutes for the substantive call; first answer/connection is already counted in task 19 and the coverage pool.
- L2 Puerto Rico nonclinical navigator: 5 / 12 / 25 minutes per routine-to-complex nonclinical caregiver-support conversation, using the same band as Caregiver Education/Support CG26, where the substantive conversation is counted once. These minutes are within—not added to—the existing annual caregiver education/coaching/support allowance.
- Separate U.S. L3 clinical: 0 / 0 / 20 minutes only for triggered clinical, behavioral-health, high-distress, or safeguarding assessment within scope.
- Separate U.S. L4 clinical: 0 / 0 / 25 minutes only when higher-clinician assessment/treatment authority is required. Puerto Rico L2 does not perform either clinical branch.
Completion evidence. Direct human conversation delivered by an approved care-team member; caregiver need and source retained; support/coaching actually provided; caregiver understanding/response and next choice captured; promised actions owned; clinical/safety/service routes separately evidenced.
What does not prove completion. L1 first answer; connection or callback attempt; voicemail; scheduling; AI/chatbot coaching; educational material; care-team member says hello but performs no support; advice delivered without caregiver response; or clinical transfer alone.
Edge cases, open decisions, and minimum tests. Existing: X-002, X-003, X-008, X-009, X-010, X-014, X-015, CC-E10; D-005, D-013, D-014; O-001, O-003, O-004, O-011, O-032, O-039, O-PA-001. Missing edge cases: caregiver requests confidentiality; severe distress; caregiver and beneficiary accounts conflict; caregiver is not representative; language/accessibility barrier; repeated calls. Minimum tests: routine stress coaching; behavior challenge; safety concern; medication question; caregiver asks for diagnosis/treatment; confidentiality request; failed teach-back; L1-only call; clinician joins; caregiver declines next step.
22. Record a call handled directly by the GUIDE team#
When. After an approved GUIDE interdisciplinary-team member directly handles all or part of a beneficiary/caregiver call, including an L2 navigator support call or a separate U.S. clinical consultation.
What and how much. The actual team member records the caller's need, relevant source statements, response/support delivered, consultations and transfers, decisions or instructions within that member's authority, caller response/teach-back, safety action, promises, unresolved work, and the next owner. When multiple team members participate, each clinically or operationally material contribution remains attributable rather than collapsed into one anonymous team note.
Data and provenance. Use call/time identifiers; caller/beneficiary identity and authority state; team actor identity, role, organization, beneficiary-location authority when clinical; source statements; record context used; exact support/decision/instruction; communication and teach-back; transfers; promises; follow-up owner; and human correction/attestation. A transcript is not an attestation. A note that says “discussed” does not show what was done. A clinical title does not prove authority for the location or act.
Potential Proxi work. Prepare a source-linked draft, prefill objective call facts, separate caller report from team action, present missing required fields, and route promised work. Proxi cannot author or attest a clinical decision, infer understanding, merge conflicting accounts into one fact, or choose the final disposition.
Human role. The team member who performed the work reviews, corrects, and owns the record. L2 owns nonclinical navigator/support content only. Separate U.S. L3/L4 clinicians own their clinical assessments and instructions within scope. L1 does not attest another person's work.
Provisional human minutes (low / typical / high).
- L1 Philippines: 0 / 0 / 0 minutes; L1-handled off-duty interactions use task 21.
- L2 Puerto Rico nonclinical navigator: 1 / 3 / 7 minutes when L2 handled the nonclinical call.
- Separate U.S. L3 clinical: 0 / 3 / 7 minutes when L3 handled the clinical call.
- Separate U.S. L4 clinical: 0 / 3 / 7 minutes when L4 handled the clinical call. The L2/L3/L4 lines are alternative actor branches, not automatically additive.
Completion evidence. Actor-attributed reviewed record that distinguishes reported facts, team actions, clinical decisions/instructions, caller response, and still-open work; every promise has an owner and due expectation.
What does not prove completion. Transcript; ambient AI draft; copied prior note; generic “care team spoke”; unsigned clinical statement; a call-duration record; or a closed call disposition without the performed work.
Edge cases, open decisions, and minimum tests. Existing: X-001, X-002, X-004, X-009, X-011, X-015; D-015; O-003, O-004, O-011. Missing edge cases: multiple team members disagree; caller corrects the record; clinician lacks location authority; note completed by someone absent from call; promised work omitted. Minimum tests: L2 coaching call; L3 clinical call; L4 decision; mixed L2/L3 call; AI draft error; late correction; unsigned note; two-team-member call.
23. Confirm that the handoff reached the GUIDE team#
When. After any third-party/off-duty handoff, direct call note that creates follow-up, failed transmission retry, or route change, and before treating the information-sharing obligation as operationally owned.
What and how much. Verify that the exact interaction record reached a usable team destination and that a named person or approved staffed queue accepted responsibility for reviewing it. If delivery or acceptance fails, use the approved backup, preserve every attempt, and escalate without silently assigning ownership. Keep technical delivery, human access, acceptance, callback, service delivery, and resolution separate.
Data and provenance. Use interaction/version identity; exact destination; release/transmission event; delivery/access result; recipient identity/role; acceptance time; work requested; promised timing; attempts; failure reason; fallback route; and subsequent callback/service evidence. A success code is not human acceptance. A named assignee who has not seen the item is not an owner. Acceptance does not prove the service occurred.
Potential Proxi work. Monitor delivery/access signals, request structured acceptance, notify backup staff, retry through an approved route, and keep the item visible until accepted or truthfully escalated. Proxi cannot invent ownership, interpret silence as acceptance, or close the underlying caller need.
Human role. Philippine L1 handles routine delivery exceptions, and an approved senior L1 worker performs repeat-failure recovery in that same job lane. The Puerto Rico GUIDE navigator accepts navigator work. The applicable U.S. clinician accepts protected clinical work within actual location and scope. No generic operations or supervisor role is created.
Provisional human minutes (low / typical / high).
- L1 Philippines: 0 / 1 / 4 minutes for an exception or manual confirmation.
- L2 Puerto Rico nonclinical navigator: 0 / 1 / 3 minutes when accepting navigator work.
- Separate U.S. L3 clinical: 0 / 1 / 3 minutes when accepting an L3 clinical item.
- Separate U.S. L4 clinical: 0 / 1 / 3 minutes when accepting an L4 clinical item. Only the actual receiving branch is counted.
Completion evidence. Matched interaction/version, usable destination, actual human or staffed-queue acceptance, named/current owner, and promised timing; failed acceptance has a preserved escalation owner and remains open.
What does not prove completion. Sent email; fax success; API 200; message marked delivered; generic queue; auto-assignment; read receipt without ownership; callback scheduled; or later clinical/service completion.
Edge cases, open decisions, and minimum tests. Existing: X-003, X-006, X-007, X-013, X-014, X-015; O-003, O-004. Missing edge cases: wrong team accepts; duplicate handoffs create two owners; recipient leaves shift; message accessible but not noticed; high-risk item waits in routine queue. Proposed open decision: exact team acceptance and orphaned-handoff standard. Minimum tests: immediate acceptance; technical delivery/no acceptance; wrong team; duplicate; shift change; outage; backup accepts; clinical item reaches unauthorized clinician.
24. Review after-hours calls that need follow-up#
When. After team receipt when an off-duty interaction contains an unresolved promise, changed circumstance, navigator need, clinical/medication/safety concern, community-service issue, caregiver need, or any uncertainty that cannot be closed by the first-answer record alone.
What and how much. Review the source interaction beside current relevant care information; preserve conflicts and unknowns; decide which nonclinical services need L2 work and which clinical questions require the separate U.S. clinical lane; identify the immediate owner, expected contact, and every child action. “No further action” is allowed only when an authorized human actually reviewed the exact record and states the reason within scope.
Data and provenance. Use the verified interaction; caller and beneficiary identity state; prior care-plan/service facts with source/currentness; reported changes; promises; prior transfers and decisions; responsible team roles; beneficiary location for clinical work; review actor/time; selected follow-up; and no-action rationale. AI summaries and old care-plan facts remain source aids. A reviewer opening the record does not prove review or disposition.
Potential Proxi work. Place the verified call beside relevant current facts, highlight conflicts/missing data, group the open obligations, and prepare candidate work items. Proxi cannot select clinical urgency, decide no-action, resolve conflicting evidence, or assign a licensed act to Puerto Rico.
Human role. L2 reviews and owns nonclinical navigator/service follow-up. Separate U.S. L3/L4 reviews clinical, medication, behavioral, safety, or treatment questions within location and scope. L1 may assemble missing administrative facts but does not decide the follow-up.
Provisional human minutes (low / typical / high).
- L1 Philippines: 0 / 1 / 4 minutes when administrative facts must be assembled.
- L2 Puerto Rico nonclinical navigator: 2 / 5 / 12 minutes for routine-to-complex nonclinical follow-up review.
- Separate U.S. L3 clinical: 0 / 0 / 15 minutes when a clinical/behavioral/safety review is triggered.
- Separate U.S. L4 clinical: 0 / 0 / 20 minutes when higher-clinician judgment is triggered. Clinical minutes are never Puerto Rico labor.
Completion evidence. Actor-attributed review of the exact interaction; explicit follow-up or authorized no-action disposition; each child action routed to a role within authority; expected timing; and unresolved conflicts retained.
What does not prove completion. Record opened; AI recommendation; generic “follow up”; L2 assigned a medical question; clinician name without acceptance; no-action checkbox without rationale; or one child task used to close the others.
Edge cases, open decisions, and minimum tests. Existing: X-003, X-004, X-005, X-008, X-009, X-010, X-014, X-015; D-015; O-003, O-004, O-032. Missing edge cases: late record after condition changed; two calls conflict; reviewer lacks location authority; morning queue overload; caller already sought emergency care. Minimum tests: routine navigator follow-up; medication question; safety concern; conflicting calls; no action with rationale; clinical authority mismatch; missing record; delayed review.
25. Call the beneficiary or caregiver back when promised or needed#
When. At the accepted promised time, when task 24 identifies a needed human callback, after a failed handoff/reconnection, or when the caller needs the approved answer or next action explained. Do not use a routine callback to delay an immediate safety route.
What and how much. Reach the correct beneficiary/caregiver; re-establish identity/disclosure boundary; acknowledge the prior call; provide the approved update or conduct the assigned navigator/clinical conversation; answer within scope; confirm the essential next action through usable communication; and record whether the immediate question was addressed. Keep callback attempt, live connection, substantive work, understanding, and issue resolution separate.
Data and provenance. Use promised time/owner; callback route; caller/beneficiary identity and authority; original concern/version; current relevant facts; approved response/instruction source; actor role and location authority for clinical work; attempts; live answer; conversation; teach-back; remaining need; and next owner. A dial, voicemail, or delivered message is not a callback conversation. A live answer does not prove the intended person or substantive completion.
Potential Proxi work. Schedule and remind, initiate the call, present source-linked context and approved content, support accessible communication, capture the human-owned recap, and retry under approved policy. Proxi cannot be the substantive required human call, choose clinical advice, alter a clinician instruction, infer understanding, or close from connection alone.
Human role. L2 conducts routine nonclinical navigator/caregiver callbacks. Separate U.S. L3/L4 conducts clinical callbacks within location and scope. L1 may connect, verify callback status, or deliver only approved nonclinical first-answer information; L1 does not turn the callback into navigator contact or clinical advice.
Provisional human minutes (low / typical / high).
- L1 Philippines: 0 / 2 / 5 minutes for connection/retry support when required.
- L2 Puerto Rico nonclinical navigator: 3 / 8 / 15 minutes for a nonclinical substantive callback.
- Separate U.S. L3 clinical: 0 / 0 / 20 minutes for a triggered L3 clinical callback.
- Separate U.S. L4 clinical: 0 / 0 / 25 minutes for a triggered higher-clinician callback. The L2, L3, and L4 substantive branches are alternatives unless the call genuinely needs more than one role.
Completion evidence. Matched live recipient; substantive human callback by the proper role; approved information/support delivered; essential understanding/response captured; and remaining actions owned. Exhausted attempts remain a truthful open/non-success outcome under the unresolved SLA policy.
What does not prove completion. Dial; voicemail; text; message read; appointment created; wrong person answers; L1 reads clinical content; caller says “okay”; or one completed callback closes an unrelated service need.
Edge cases, open decisions, and minimum tests. Existing: X-001, X-002, X-006, X-007, X-008, X-014, X-015; D-005, D-015; O-003, O-004, O-011. Missing edge cases: callback reaches revoked caregiver; beneficiary and caregiver request different callbacks; location changes before clinical callback; interpreter absent; promise cannot be met. Minimum tests: routine L2 callback; L3 clinical callback; wrong number; voicemail only; revoked authority; failed teach-back; interpreter; immediate danger before promised time.
26. Carry the new need into the relevant GUIDE service#
Preparation-before-judgment branch. A clear authenticated nonclinical fact or choice may enter the correct receiving task directly. Do not add L2 merely to confirm a clean prepared handoff. L2 enters only when the receiving service itself requires or the person requests a human navigator act, or when ambiguity, disagreement, distress, accessibility failure, warm introduction, relationship work, or useful human confirmation remains.
When. When the call reveals a new or changed need that belongs in Care Plan, Care Coordination, Medication Management, Referral/Services, Caregiver Education/Support, Ongoing Monitoring, Comprehensive Assessment, or another approved GUIDE workstream.
What and how much. Create a source-linked service request that preserves the caller's words, beneficiary/caregiver choices, open questions, prior call actions, and responsible service. Obtain actual acceptance from the appropriate owner and begin the first required service step. Do not treat the call note itself, a service label, or an accepted work item as delivery of the service.
Data and provenance. Use call/interaction identity; reported need and speaker; current facts and provenance; related plan/service records; caller-requested outcome; disclosure/authority; assigned service and rationale; receiving actor/role; acceptance; first performed service step; and later completion evidence. AI classification is a candidate. A referral to an SOP is not the service. A clinical phrase does not authorize L2 to handle it.
Potential Proxi work. Preserve the source link, show relevant service choices, prepare the receiving packet, notify the approved owner, monitor acceptance, and prevent the request from disappearing between queues. Proxi cannot decide the clinical route, convert a candidate fact to truth, or claim the underlying service was furnished.
Human role. L2 performs the actual receiving nonclinical navigator service only on the triggered branch above; L2 does not bless a clean prepared fact or choice. Separate U.S. L3/L4 accepts clinical work within location and scope. L1 may perform bounded administrative transfer work but cannot make the navigator or clinical decision. The beneficiary/caregiver participates where the underlying service requires choice.
Provisional human minutes (low / typical / high).
- L1 Philippines: 0 / 1 / 4 minutes for administrative transfer/acceptance pursuit.
- L2 Puerto Rico nonclinical navigator: 0 on the clean path; 1 / 3 / 8 minutes only when an actual receiving L2 human act is triggered, to resolve the remaining nonclinical need, choice, ownership, or first service step.
- Separate U.S. L3 clinical: 0 / 0 / 3 minutes only to accept a routed L3 item.
- Separate U.S. L4 clinical: 0 / 0 / 3 minutes only to accept a routed L4 item. All substantive underlying-service minutes are counted in that service SOP, not here.
Completion evidence. Source-linked receiving service; role-authorized owner acceptance; first actual service step performed or explicitly scheduled/owned where scheduling is the first step; and the underlying service remains separately tracked to its own completion evidence.
What does not prove completion. Tag; queue; work item; owner name; notification; acceptance alone; scheduled callback; copied call note; care-plan label; or completion of the 24/7 call.
Edge cases, open decisions, and minimum tests. Existing: X-003, X-004, X-006, X-010, X-014, X-015; D-015; O-003, O-009, O-019, O-022, O-023, O-025, O-026, O-034. Missing edge cases: need spans multiple services; receiving service rejects ownership; duplicate request; caller retracts/corrects; clinical and community needs diverge. Minimum tests: care-plan change; community referral; medication concern; caregiver education; mixed need; wrong service; duplicate; acceptance without first step; caller correction.
Single-ledger rule and contact-window treatment for inbound episodes#
One human episode is counted in exactly one SOP labor ledger. Its evidence may still support every service the conversation actually furnished:
- A call entering through the 24/7 line is counted here: the shared first-answer L1 envelope plus any actual L2, L3, or L4 work in tasks 20, 22, 24, 25, and 26. Ongoing Monitoring Task 36 does not book the same episode.
- A call or message reaching the navigator directly during ordinary availability is counted under Ongoing Monitoring Task 36 and not again here.
- A substantive caregiver-support conversation is counted once under Caregiver Education/Support CG26 at 5 / 12 / 25 L2 minutes, regardless of its entry lane. The 24/7 lane keeps only its actual first-answer, intake, and connection work.
- When a task 25 callback itself delivers the receiving navigator or caregiver service, the callback and service conversation are the same human minutes. They are not counted once as callback labor and again as service labor.
- Reconnection after a dropped call continues the same episode. A genuinely new concern arriving later is a new episode.
A first-answer-only call never satisfies the required navigator contact. When the substantive portion is performed by an eligible care navigator, through an allowed modality, with a permitted participant, and for an ongoing-support purpose, send the completed record to Ongoing Monitoring Task 31 for the contact-window evaluation. If it qualifies for the current window, cancel the redundant planned outbound contact under X-017 and tell the person. If a qualifying contact already occurred in that window, the inbound episode remains a responsive service but receives no second cadence credit. Cancelling the redundant appointment never cancels a promised callback or unresolved action.
Cross-pillar feeds from a 24/7 episode#
This SOP participates in cross-pillar episodes under 25_GUIDE_Eight_Pillar_Service_Integration.md. Use the shared no-repeat packet, preparation-before-judgment route, receiving-owner acceptance rule, and count-once labor rule. 24/7 Access receives service attribution for the required live human answer, source-faithful concern capture, and proper immediate response or connection. A separate same-call service is attributed only when its own distinct content and result occurred.
| Direction | Named feeds | Local handling and result | Review |
|---|---|---|---|
| Inbound context to 24/7 | Comprehensive Assessment; Care Plan; Ongoing Monitoring and Support; Care Coordination; Medication Management; Referral and Services; Caregiver Education and Support | Make only approved current context available to the human responder and clinical route. Context availability opens no receiving service task and is not itself a service event. | |
| Outbound from 24/7 | Comprehensive Assessment; Care Plan; Ongoing Monitoring and Support; Care Coordination; Medication Management; Referral and Services; Caregiver Education and Support | Complete the immediate answer and connection first; then bind each actual downstream need to the existing receiving task named in Manual 25. Acceptance transfers custody but does not prove the downstream service. |
27. Restore service during a line, vendor, or staffing failure#
When. Whenever the primary number, carrier, queue, responder seat, backup route, third-party service, team handoff path, or clinical escalation route is unavailable or degraded enough to threaten actual human access.
What and how much. Detect and bound the failure; activate approved backup routing and human coverage; stop interruptible administrative work; reconnect affected callers; communicate a usable alternate route only when necessary; restore the primary service; and reconcile every call, message, and handoff that occurred during the gap. Preserve actual gap duration and affected calls rather than rewriting the record as continuous coverage.
Data and provenance. Use failure source/time; affected route and capability; scheduled/actual responders; queue and unanswered/dropped calls; affected caller/callback details; backup activation and readiness; alternate-number release; clinical-route availability by beneficiary location; recovery time; reconciliation results; and unresolved callers. A green vendor dashboard does not prove a human route. Primary restoration does not prove affected callers were recovered.
Potential Proxi work. Detect route failure, activate approved backup, suspend administrative workload, alert staff, retain affected-call lists, initiate reconnection, and assemble recovery evidence. Proxi cannot replace the human answer with AI, choose an unapproved clinical destination, or erase the outage after service returns.
Human role. An approved senior Philippine L1 worker activates backup staff and reconnects callers; the System Administrator restores approved configuration; Software Engineering enters only for a product defect; and the outside telecommunications vendor repairs its service. The Puerto Rico GUIDE navigator handles affected nonclinical relationships when needed. Head of Nursing or the Chief Medical Officer establishes only the alternate clinical route within that leader's authority; clinical acts remain outside Puerto Rico.
Provisional human minutes (low / typical / high).
- L1 Philippines: 2 / 10 / 30 minutes per incident for backup activation, manual routing, and caller recovery; large incidents are measured as observed incident labor, not forced into this cap.
- L2 Puerto Rico nonclinical navigator: 0 / 0 / 5 minutes only for affected relationship/service follow-up.
- Separate U.S. L3 clinical: 0 / 0 / 5 minutes to accept/establish alternate L3 clinical coverage when that route failed.
- Separate U.S. L4 clinical: 0 / 0 / 5 minutes to accept/establish alternate L4 coverage when needed. These are incident minutes, not per-beneficiary additions.
Completion evidence. Actual human backup coverage; restored usable route; documented gap; every affected call/handoff reconciled to live recovery, owned follow-up, or truthful unresolved outcome; and confirmed current clinical route where affected.
What does not prove completion. Vendor ticket closed; number rings; dashboard green; responder login; primary route restored; backup configured but untested; mass text sent; or no complaint received.
Edge cases, open decisions, and minimum tests. Existing: X-007, X-008, X-013, X-014, X-015; D-014, D-015; O-003, O-004, O-039. Missing edge cases: primary and backup share one failure; region-wide outage; surge during failure; alternate number inaccessible; clinical route only fails; caller list incomplete. Proposed open decision: outage declaration, alternate-route release, mass-reconnection, and maximum-gap standards. Minimum tests: carrier failure; responder no-show; vendor outage; concurrent calls; backup also fails; clinical-route outage; recovery with dropped-call list; dashboard false green.
28. Review whether every hour remains human-covered#
When. On a defined recurring quality cadence and after every gap, no-answer, abandoned call, backup activation, staffing exception, clinical-route failure, language/accessibility failure, or third-party handoff failure.
What and how much. Compare planned coverage with actual answerability and call outcomes across all hours. Review answer/no-answer, abandoned/dropped calls, simultaneous demand, staff readiness, backup use, transfers, third-party sharing, team acceptance, caregiver-support connection, language/accessibility, and clinical-route availability. Assign corrective work for each real gap and verify that affected callers were recovered.
Data and provenance. Use roster and accepted shifts; actual readiness; telephony/call events; queue/concurrency; answer and abandonment; callback/recovery; backup activation; third-party records; team acceptance; accessibility/language outcomes; clinical-route availability; incident/correction owner; and verification. Planned coverage is not actual coverage. Average answer rate can conceal an uncovered hour. Absence of calls does not prove the human route worked.
Potential Proxi work. Assemble coverage and call evidence, highlight exact uncovered intervals and failed routes, compare repeated failure patterns, and prepare corrective-action queues. Proxi cannot decide that a gap is acceptable, certify compliance from an average, or substitute a simulated/AI answer for human coverage.
Human role. An approved senior Philippine L1 worker performs routine coverage review and correction inside the L1 job category. The assigned Puerto Rico GUIDE navigator reviews nonclinical care-team or caregiver-support gaps. Head of Nursing or the Chief Medical Officer reviews only the clinical-route gap within that leader's authority. System Administration, Software Engineering, or HR/workforce receives only the matching configuration, defect, or capacity task. No generic operations, quality, or program-leadership role and no routine clinical chart review is required.
Provisional human minutes (low / typical / high).
- L1 Philippines: 5 / 15 / 45 minutes per review cycle, pooled program/quality time rather than a per-beneficiary touch.
- L2 Puerto Rico nonclinical navigator: 0 / 0 / 5 minutes only when a navigator/care-team coverage gap needs review.
- Separate U.S. L3 clinical: 0 / 0 / 5 minutes only when an L3 route gap needs review.
- Separate U.S. L4 clinical: 0 / 0 / 5 minutes only when an L4 route gap needs review.
Completion evidence. Defined period and denominator; planned-versus-actual coverage reconciliation; exact gaps and affected calls; assigned corrective actions; caller recovery status; and verified closure or truthful open exception for each gap.
What does not prove completion. Monthly average; staffing spreadsheet; vendor SLA report; one test call; no complaints; no calls in the interval; supervisor attestation without call evidence; or correction plan without implementation.
Edge cases, open decisions, and minimum tests. Existing: X-005, X-007, X-013, X-014, X-015; O-003, O-004, O-039. Missing edge cases: denominator excludes blocked calls; one language route fails while English succeeds; shift clocks drift; concurrent calls masked by average; third-party interactions missing entirely. Proposed open decision: coverage review cadence, metrics, sampling versus census, and gap-correction evidence. Minimum tests: fully covered day; one-minute gap; no-call outage; concurrency failure; language failure; missing third-party note; clinical route unavailable; repeated vendor failures.
29. Use repeated calls to improve the person's ongoing support#
When. When source-linked calls repeat the same or related unmet need, failed connection, caregiver stressor, safety concern, medication question, service barrier, or communication problem often enough to warrant human review under a still-open threshold policy. A single serious call may separately trigger immediate review; repetition is not required for safety action.
What and how much. Software builds separate source-linked action packets from the underlying calls rather than only an AI summary: the practical/preference question, each service failure or open promise, and every clinical/safety question. Present the prepared practical packet directly to the beneficiary/caregiver for confirmation of whether the calls reflect the same unresolved need and what response they want. L2 enters only for a required/requested human GUIDE service, ambiguity, disagreement, distress, a warm introduction, relationship-dependent work, or useful human confirmation. The clinical packet routes directly to the separate authorized U.S. clinical team. Preserve supported no-change decisions within each actor's authority and route every resulting action to its own service.
Data and provenance. Use every linked call and source statement; identity-match confidence; dates/times; concern categories as candidates; actual resolutions and failures; current care plan/services; beneficiary/caregiver feedback; review actor; clinical location/authority where applicable; decision and rationale; and resulting service evidence. Similar words do not prove the same problem. Call count does not prove severity or treatment failure. AI clustering does not authorize a plan change.
Potential Proxi work. Link likely related calls without dropping outliers, show source excerpts and outcome history, identify recurring unresolved obligations, and prepare options for human review. Proxi cannot declare the pattern clinically meaningful, select a plan change, infer caregiver capacity, or close the repeated need.
Human role. The beneficiary/caregiver confirms the practical pattern and desired response. L2 receives the prepared practical/preference packet only under the human-entry conditions above and performs relationship-dependent nonclinical follow-up within authority. Separate U.S. L3/L4 receives the prepared clinical packet directly and handles clinical, behavioral, safety, medication, diagnosis, or treatment implications; L2 is not an intermediary. L1 may correct call links/source facts but does not decide the response.
Provisional human minutes (low / typical / high).
- L1 Philippines: 0 / 2 / 5 minutes only to correct source links or retrieve a missing call record.
- L2 Puerto Rico nonclinical navigator: 0 / 10 / 20 minutes; zero for a clean direct confirmation and only the human-entry conditions above create navigator time.
- Separate U.S. L3 clinical: 0 / 0 / 15 minutes when clinical/behavioral/safety review is triggered.
- Separate U.S. L4 clinical: 0 / 0 / 20 minutes when higher-clinician assessment or treatment judgment is triggered. Puerto Rico performs no clinical portion.
Completion evidence. Reviewed source-call set; authenticated same/different-need disposition; beneficiary/caregiver input; actor-authorized change or supported no-change rationale; each resulting service action owned; and later outcome tracked separately.
What does not prove completion. Call count; AI cluster; sentiment score; repeated keyword; navigator notification; care-plan label; automatic increase in contact frequency; one resolved call; or plan change without beneficiary/clinical authority.
Edge cases, open decisions, and minimum tests. Existing: X-001, X-004, X-005, X-008, X-009, X-010, X-011, X-015; D-015; O-003, O-004, O-010, O-013, O-039, O-041. Missing edge cases: family members call about different concerns; duplicate records inflate frequency; serious one-time call; calls span plan versions; caregiver asks to keep a recurring issue confidential; apparent pattern disappears after correction. Proposed open decision: repeated-call threshold, review cadence, no-change evidence, and caregiver-confidentiality handling. Minimum tests: true recurring service failure; duplicate calls; different callers/different needs; medication pattern; caregiver stress; one-time emergency; supported no change; clinical review; plan/service change with follow-through.
Open all 29 task proceduresDetailed task inventory
| Task | What the task entails | GUIDE anchor | What Proxi can do | Person required? | Person or role | Review |
|---|---|---|---|---|---|---|
| 1. Maintain the settled human coverage approach | Operate the approved Proxi Philippines first-answer roster continuously, with Puerto Rico nonclinical care-team support and separate U.S. beneficiary-location-authorized clinical escalation, without substituting an AI-only line or Philippine staff for formal navigator contact. | Appendix B §3.1 | Yes for scheduling, coverage monitoring, routing, and escalation support; humans still answer and provide the required service. | Yes | Philippine L1 worker performing the approved roster-supervision task; Puerto Rico GUIDE navigator; applicable U.S. clinician. These are existing job categories, not new operations roles. | |
| 2. Establish one continuously available access route | Put a telephone route in service that beneficiaries and caregivers can call at any hour and that connects them to human support, including nights, weekends, and holidays. An AI-only line does not meet the requirement. | Appendix B §3.1 | No as the required human support. Proxi can assist behind the scenes, but a working telecommunications route and a human responder must provide the service. | Yes | System Administrator; Philippine L1 worker or approved outside human helpline responder; outside telecommunications vendor. | |
| 3. Create the around-the-clock human coverage schedule | Assign named Philippines first-answer coverage for every hour, including relief for shrinkage, backups for absence and concurrency, shift change, and holidays; provide responders with Puerto Rico nonclinical care-team and separate U.S. licensed-clinical escalation contacts. | Operational task necessary to provide Appendix B §3.1 access | Yes. Proxi can build the rota, detect uncovered periods, notify the next responder, and call backup staff; managers confirm that qualified people actually accept the coverage. | Partial | Approved senior Philippine L1 worker performing the scheduling task; HR/workforce administration only for an actual capacity or employment gap; scheduled Puerto Rico navigators and U.S. clinicians accept their own coverage. | |
| 4. Define what the helpline responder may do | If a third party will answer off-duty calls, specify what the responder may explain or collect, when the responder must connect or escalate to the GUIDE team, and that the responder may not present automated output as human or clinical judgment. | Appendix B §3.1 | Partial. Proxi can draft call guides from approved policies and display the permitted actions; clinical, operations, privacy, and contracting leaders approve the responder's scope. | Governance before activation | Head of Nursing or Chief Medical Officer only for the clinical boundary in scope; Healthcare Legal Counsel only for unresolved legal authority; Participant-designated policy approver; outside helpline supervisor. These are pre-activation governance acts, not routine call roles. | |
| 5. Require third-party information sharing in the vendor arrangement | If an off-duty third party is used, require it to share information from every beneficiary or caregiver communication with the interdisciplinary care team and establish the practical handoff route. | Appendix B §3.1 | Partial. Proxi can identify the required clause, prepare a workflow description, and check that a proposed arrangement contains it; humans negotiate and execute the agreement. | Governance before activation | Healthcare Legal Counsel for contract legal terms; Participant-authorized signatory; outside vendor representative; System Administrator for the approved technical handoff. | |
| 6. Give the access number and instructions to the beneficiary and caregiver | Explain how and when to call, that a human will answer, what languages or accessibility help is available, what information the caller may be asked for, and that emergencies should use the appropriate emergency route. | Appendix B §3.1 | Partial. Proxi can prepare accessible, language-matched instructions and send them through approved channels; the navigator confirms that the person received and understood them. | Yes | Puerto Rico GUIDE navigator; beneficiary and caregiver as applicable; approved outside interpreter or accessibility provider when triggered. | |
| 7. Give a beneficiary without a caregiver a direct usable route | Provide the number and instructions directly to the beneficiary, practice using the route if helpful, and do not make access conditional on a caregiver calling or approving the call. | Appendix B §3.1 and additional-safeguard context in §1.2.3.1 | Partial. Proxi can tailor instructions, offer repetition and accessibility support, and flag the absence of a backup caller; a navigator confirms usability with the beneficiary and arranges extra human safeguards if needed. | Yes | Puerto Rico GUIDE navigator and beneficiary; applicable U.S. RN or behavioral clinician only for protected clinical work; approved outside accessibility provider when triggered. | |
| 8. Keep contact and accommodation information available to responders | Make the beneficiary's name, preferred form of address, language and accessibility needs, caregiver details when applicable, and GUIDE team contacts available to the authorized human who answers. | Operational task supporting Appendix B §3.1 | Yes. Proxi retrieves only information permitted by the approved access rule, labels its source, and blocks unresolved access; people correct source facts or configuration in their existing job lanes. | Partial | System Administrator configures approved access; L1 pursues administrative correction; Puerto Rico GUIDE navigator confirms relationship-sensitive correction; human responder uses the permitted information. | |
| 9. Check that the line and human-answer route are working | Repeatedly confirm that the number connects, off-duty routing works, a human can answer, language or accessibility routes operate, and backup routing takes over when the primary responder is unavailable. | Operational task necessary to maintain Appendix B §3.1 access | Yes. Proxi can run availability and routing checks, detect queue or transfer failures, and immediately alert the backup owner; a human investigates and repairs failures. | Partial | System Administrator for configuration; Software Engineering only for a product defect; outside telecommunications vendor for its service; approved senior Philippine L1 worker for roster recovery. | |
| 10. Receive the call at any time | Accept calls from the beneficiary or caregiver without limiting service to office hours. Connect the caller to a human care-team member, or during off-duty hours to a permitted third-party human responder. | Appendix B §3.1 | No as a substitute for the responder. Proxi may route the call or assist the responder, but an automated agent cannot be the human support CMS requires. | Yes | Philippine L1 worker or approved outside human helpline responder. | |
| 11. Recover a disconnected or unanswered call | If the connection fails after a callback number is available, have the human service call back promptly and use backup coverage if the assigned responder cannot reconnect. | Operational task supporting continuous Appendix B §3.1 access | Partial. Proxi can detect a dropped call, preserve the callback details, dial the backup route, and notify the responder; a human resumes the support conversation. | Yes | Philippine L1 worker or approved outside human helpline responder; approved senior Philippine L1 worker performs repeat-failure recovery in the same job lane. | |
| 12. Establish who is calling and whom the call concerns | Ask whether the caller is the beneficiary, caregiver, or another person; identify the beneficiary; obtain a callback route; and use only the information and authorization appropriate to the caller. Do not refuse immediate emergency help while resolving routine identity questions. | Operational task supporting Appendix B §3.1 | Partial. Proxi can guide identity matching, retrieve permitted records, and flag an unclear caller relationship; the human responder communicates with the caller and applies privacy and emergency procedures. | Yes | Philippine L1 worker or approved outside human responder; software enforces approved disclosure rules; applicable U.S. clinician handles clinical exceptions; Healthcare Legal Counsel handles only unresolved legal authority. | |
| 13. Ask what help is needed now | Let the caller explain the concern, ask clarifying questions, and distinguish a request for information, caregiver support, navigation help, an urgent safety issue, or a clinical question without forcing every call into a single script. | Appendix B §3.1 | Partial. Proxi can capture the caller's words, suggest clarifying prompts, and organize the issue for the responder; a human listens, responds, and decides what expertise is needed. | Yes | Philippine L1 worker or approved outside human responder; Puerto Rico GUIDE navigator or applicable U.S. clinician enters only when the issue requires that category's task. | |
| 14. Respond to a beneficiary calling without a caregiver | Speak directly with the beneficiary, use dementia-capable communication, repeat and confirm important instructions, and bring in additional human help when the beneficiary cannot safely act on the information alone. | Appendix B §3.1 and additional-safeguard context in §1.2.3.1 | Partial. Proxi can simplify language, retrieve known communication preferences, and prompt teach-back; the human responder communicates and judges whether more support is needed. | Yes | Philippine L1 worker or approved outside human responder; applicable U.S. clinician or outside emergency responder only when triggered. | |
| 15. Identify an immediate emergency or safety threat | The Philippines L1 responder recognizes explicit protocol stop cues, keeps the caller connected when possible, and invokes or warm-transfers to the approved emergency or separate U.S. clinical route. L1 does not triage ambiguous symptoms or determine individualized urgency. | Human-support function under Appendix B §3.1; caregiver safety examples in §8.2.4 | Partial. Proxi can surface candidate stop cues and display approved routes, but it cannot determine urgency, select a clinical disposition, or replace emergency responders. | Yes | Philippine L1 worker invokes the protocol; applicable U.S. clinician performs clinical triage; outside emergency authority performs the emergency response. | |
| 16. Answer a non-clinical navigation or program question | Provide information within the responder's approved scope, such as whom to contact, how to reach an existing service, or when a navigator will be available; do not improvise clinical advice. | Appendix B §3.1 | Partial. Proxi can retrieve approved, current program and resource information and draft an answer; a human delivers it, confirms understanding, and notices when the issue is more than administrative. | Yes | Philippine L1 worker answers a bounded routine question; Puerto Rico GUIDE navigator enters for substantive nonclinical navigation. | |
| 17. Connect a clinical issue to an appropriate clinician | When the call requires assessment, medication advice, symptom interpretation, or another clinical judgment, bring in the on-call clinician or use the approved urgent clinical route instead of asking a non-clinical responder to decide. | Appendix B §3.1; interdisciplinary-team context | Partial. Proxi can prepare the reported facts and contact the configured clinical route; the licensed clinician performs the clinical assessment and gives clinical direction. | Yes | Applicable U.S. RN, behavioral clinician, or L4 clinician selected by the exact protected task; outside prescriber, pharmacist, or emergency professional remains an outside party. | |
| 18. Connect a service or support issue to the GUIDE team | When the caller needs navigation, care-plan follow-up, community support, medication support, or another GUIDE service that cannot be completed during the call, explain the handoff and send it to the responsible team member. | Appendix B §3.1 and cross-reference to the interdisciplinary care team | Partial. Proxi can identify the likely service, retrieve the responsible team member, prepare the handoff, and notify that person; a human responder confirms the route and a GUIDE team member performs the service. | Yes | Philippine L1 worker performs the handoff; Puerto Rico GUIDE navigator or the applicable U.S. clinician performs the receiving GUIDE task. | |
| 19. Receive or initiate an ad hoc caregiver support call | Ensure the 24/7 helpline can receive a request initiated by the beneficiary or caregiver, and allow the GUIDE Participant to initiate a caregiver support call proactively when an issue arises. Connect the caregiver to the required care-team support. | Appendix B §3.2 and §8.2.4 | Partial. Proxi can recognize or create the caregiver-support request, collect callback and context information, and connect the participants; it cannot be the required one-on-one human support. | Yes | Philippine L1 worker receives and connects; Puerto Rico GUIDE navigator performs nonclinical caregiver support; applicable U.S. clinician enters only for protected clinical work. | |
| 20. Deliver the caregiver's one-on-one support call | Following a beneficiary request, caregiver request, or GUIDE Participant initiation, a care-team member speaks directly with the caregiver about the issue as it arises. Depending on the need, this may include coaching on stress management, self-care, well-being, behavioral challenges, functional changes, and beneficiary or caregiver safety. | Appendix B §§8.2.4 and 8.3.4 | Partial. Proxi can retrieve prior caregiver needs, suggest approved topics, and draft a recap, but CMS requires the call to be provided directly by a care-team member. | Yes | Puerto Rico GUIDE navigator performs nonclinical caregiver coaching or support within verified competency; applicable U.S. clinician performs protected clinical work. | |
| 21. Share every third-party interaction with the interdisciplinary team | After any off-duty communication handled by a third party, send the team what the caller reported, what the responder said or did, any connection or escalation made, and any follow-up still needed. | Appendix B §3.1 | Yes for secure transmission and structured preparation. Proxi can create and deliver a source-labeled summary automatically, but the third-party responder must verify that it accurately represents the human communication. | Yes | Outside human responder verifies its communication; software routes and tracks it; Puerto Rico GUIDE navigator or applicable U.S. clinician accepts the exact receiving task. | |
| 22. Record a call handled directly by the GUIDE team | Capture who called, when, what help was requested, the human response, consultations or transfers, safety actions, promises, and callback needs so the normal team can continue the work. | Operational task supporting Appendix B §§3.1–3.2 | Partial. Proxi can transcribe with permission and draft the call note; the human care-team responder reviews, corrects, and owns the record. | Yes | The actual Philippine L1 worker, Puerto Rico GUIDE navigator, U.S. clinician, or approved outside responder who handled the call reviews and owns that record. | |
| 23. Confirm that the handoff reached the GUIDE team | When a third party or after-hours responder sends information, verify that an identified GUIDE team member can see it and knows whether a callback or other action is needed; repair failed delivery rather than assuming transmission worked. | Operational task completing the §3.1 information-sharing duty | Yes. Proxi can monitor delivery, alert on failure, and resend through an approved backup route; a human accepts responsibility for the follow-up work. | Partial | Software monitors delivery; Philippine L1 performs administrative delivery repair; the exact Puerto Rico navigator or U.S. clinician accepts the receiving task. | |
| 24. Review after-hours calls that need follow-up | The appropriate GUIDE team member reviews the caller's report and the human response, checks current care information, and decides what further navigator, clinical, medication, safety, or community-support work is needed. | Appendix B §3.1 information sharing; service connection purpose of human support | Partial. Proxi can organize the call details beside relevant care information and suggest the responsible service; a human determines and performs the needed follow-up. | Yes | Puerto Rico GUIDE navigator or applicable U.S. clinician selected by the reported issue. | |
| 25. Call the beneficiary or caregiver back when promised or needed | Contact the caller, explain the next action, answer remaining questions, and confirm whether the immediate concern has been addressed; use an interpreter or accommodation as needed. | Operational continuation of Appendix B §§3.1–3.2 | Partial. Proxi can schedule, remind, connect, and draft the recap; a human GUIDE team member conducts the substantive callback. | Yes | Puerto Rico GUIDE navigator or applicable U.S. clinician selected by the promised follow-up; approved outside interpreter when triggered. | |
| 26. Carry the new need into the relevant GUIDE service | If the call reveals a changed goal, unmet clinical or community need, medication concern, caregiver need, or transition, begin the corresponding human work rather than leaving the matter only in a call note. | Operational continuation of Appendix B §3.1 across GUIDE services | Partial. Proxi can transfer the reported information to the responsible service task and prepare administrative work; beneficiaries, caregivers, navigators, and clinicians make the decisions and deliver the service. | Yes | Beneficiary or caregiver; Puerto Rico GUIDE navigator for nonclinical work; applicable U.S. clinician for protected work; outside provider for its own service. | |
| 27. Restore service during a line, vendor, or staffing failure | Detect loss of the primary route, activate backup routing and backup human coverage, publish the usable alternate number where needed, and reconnect callers affected by the failure. | Operational task necessary to maintain Appendix B §3.1 access | Yes. Proxi can monitor availability, trigger the approved backup route, notify staff and affected callers, and maintain a list of calls needing reconnection; people repair the system and provide support. | Partial | System Administrator; Software Engineering only for a product defect; outside telecommunications vendor; Philippine L1 backup responder; HR/workforce only for an actual staffing-capacity gap. | |
| 28. Review whether every hour remains human-covered | Regularly inspect the rota, actual answer coverage, transfer failures, unanswered calls, third-party handoffs, caregiver-support connections, language or accessibility failures, and backup use; correct any service gap. | Operational oversight necessary for Appendix B §§3.1–3.2 | Yes. Proxi can assemble coverage and call-handling information and highlight exceptions; an accountable manager investigates and corrects staffing, vendor, or process failures. | Partial | Approved senior Philippine L1 worker performs coverage review; System Administrator, HR/workforce, or Head of Nursing enters only for its specific configuration, capacity, or clinical-coverage task. | |
| 29. Use repeated calls to improve the person's ongoing support | When calls repeatedly concern the same unmet need, bring the pattern to the care navigator and appropriate team member so the care plan, contact frequency, education, services, or safeguards can be reconsidered with the beneficiary or caregiver. | Appendix B §3.1 linked operationally to §4.2 and §§2, 6, 7, and 8 | Partial. Proxi can group related calls and summarize the recurring reported need; people decide whether and how to change the service provided. | Yes | Puerto Rico GUIDE navigator; beneficiary and caregiver as applicable; applicable U.S. clinician only for the affected protected clinical task. |
Requirement, value, and clinical classificationReference table
| Task | GUIDE standing | Customer-value position | Clinical lane | Why |
|---|---|---|---|---|
| 1. Maintain the settled human coverage approach | Necessary delivery work | Compliance infrastructure | No clinical judgment | CMS requires continuous 24/7 human access; Proxi's coverage approach is already settled, so the recurring task is to maintain and evidence the human capability rather than repeatedly choose a model. |
| 2. Establish one continuously available access route | Public RFA care-delivery requirement | Core customer value | No clinical judgment | Round-the-clock access to human support is the service customers directly receive. |
| 3. Create the around-the-clock human coverage schedule | Necessary delivery work | Compliance infrastructure | No clinical judgment | Continuous human access cannot be reliable without complete primary and backup coverage. |
| 4. Define what the helpline responder may do | Necessary delivery work | Compliance infrastructure | No clinical judgment | Clear scope prevents a non-clinical responder from giving unauthorized clinical advice. |
| 5. Require third-party information sharing in the vendor arrangement | Public RFA care-delivery requirement | Compliance infrastructure | No clinical judgment | The RFA requires an off-duty third party to share every communication with the care team. |
| 6. Give the access number and instructions to the beneficiary and caregiver | Necessary delivery work | Core customer value | No clinical judgment | Access has no practical value unless the people served know how to use it. |
| 7. Give a beneficiary without a caregiver a direct usable route | Public RFA care-delivery requirement | Core customer value | Clinical review on trigger | The beneficiary has direct access; inability to use it safely may require added clinical safeguards. |
| 8. Keep contact and accommodation information available to responders | Necessary delivery work | Value through better execution | No clinical judgment | Current language, accessibility, and contact facts make the human response usable. |
| 9. Check that the line and human-answer route are working | Necessary delivery work | Compliance infrastructure | No clinical judgment | Availability checks prevent silent failure of the required access route. |
| 10. Receive the call at any time | Public RFA care-delivery requirement | Core customer value | Clinical review on trigger | CMS requires a human answer; clinical or safety content triggers the appropriate professional. |
| 11. Recover a disconnected or unanswered call | Necessary delivery work | Value through better execution | No clinical judgment | Reconnection preserves actual access when the first technical path fails. |
| 12. Establish who is calling and whom the call concerns | Necessary delivery work | Value through better execution | Clinical review on trigger | Routine identity and authorization work is non-clinical, but emergency content cannot wait for routine resolution. |
| 13. Ask what help is needed now | Necessary delivery work | Core customer value | Clinical review on trigger | This intake method makes human support usable; symptoms, danger, or medication issues trigger clinical review. |
| 14. Respond to a beneficiary calling without a caregiver | Public RFA care-delivery requirement | Core customer value | Clinical review on trigger | The beneficiary receives direct dementia-capable support, with added clinical help when needed. |
| 15. Identify an immediate emergency or safety threat | Necessary delivery work | Core customer value | Clinical review on trigger | A trained human can recognize an obvious emergency and invoke the approved route; ambiguous symptoms or individualized clinical urgency require a clinician. |
| 16. Answer a non-clinical navigation or program question | Necessary delivery work | Core customer value | No clinical judgment | Approved factual navigation is human support but does not require a clinician. |
| 17. Connect a clinical issue to an appropriate clinician | Necessary delivery work | Core customer value | Clinical review on trigger | The responder connects the caller; the receiving licensed professional performs the non-delegable assessment and direction. |
| 18. Connect a service or support issue to the GUIDE team | Necessary delivery work | Core customer value | Clinical review on trigger | Routine handoff is navigation; the receiving service invokes clinical review when its issue requires it. |
| 19. Receive or initiate an ad hoc caregiver support call | Public RFA care-delivery requirement | Core customer value | Clinical review on trigger | The RFA requires access to this human care-team support, with clinical issues escalated. |
| 20. Deliver the caregiver's one-on-one support call | Public RFA care-delivery requirement | Core customer value | Clinical review on trigger | CMS requires direct care-team delivery; routine coaching is non-clinical but medical or safety questions are not. |
| 21. Share every third-party interaction with the interdisciplinary team | Public RFA care-delivery requirement | Compliance infrastructure | No clinical judgment | The RFA directly requires this information transfer after third-party communication. |
| 22. Record a call handled directly by the GUIDE team | Necessary delivery work | Compliance infrastructure | No clinical judgment | Accurate recording supports continuity without substituting documentation for clinical judgment. |
| 23. Confirm that the handoff reached the GUIDE team | Necessary delivery work | Value through better execution | No clinical judgment | Delivery confirmation closes the information gap left by a mere transmission attempt. |
| 24. Review after-hours calls that need follow-up | Necessary delivery work | Value through better execution | Clinical review on trigger | Routine follow-up can be assigned administratively; clinical, medication, or safety content goes to clinicians. |
| 25. Call the beneficiary or caregiver back when promised or needed | Necessary delivery work | Core customer value | Clinical review on trigger | A human callback completes support, while substantive clinical questions require a clinician. |
| 26. Carry the new need into the relevant GUIDE service | Necessary delivery work | Core customer value | Clinical review on trigger | The call creates value only when the underlying need reaches the right service and clinical lane. |
| 27. Restore service during a line, vendor, or staffing failure | Necessary delivery work | Compliance infrastructure | No clinical judgment | Backup routing and coverage preserve the required human-access result. |
| 28. Review whether every hour remains human-covered | Necessary delivery work | Compliance infrastructure | No clinical judgment | Coverage oversight detects operating gaps that individual call records may not show. |
| 29. Use repeated calls to improve the person's ongoing support | Beyond the public GUIDE minimum | Value through better execution | Clinical review on trigger | Pattern review is not publicly required, but it can expose recurring needs and prevent repeated unresolved calls. |