Advisor Outcomes PlatformSign in
Guiding principles

Seven principles.
Enforced in the product, not just the pitch.

Advisor Outcomes Platform is built to serve the government's core principles for AI in care. Each principle below is paired with how it shows up in the workspace today and where in the product you can see it operating.

Protect privacy, dignity, and choice

The patient owns their data and the terms it's shared on. Every downstream surface honors those choices, not the other way around.

  • Care Circle with typed, revocable consent grants per relationship (view, comment, receive alerts, respond on behalf of).
  • Companion console shows outcomes to caregivers — never raw transcripts of what the patient said.
  • Audit drawer surfaces every access on the patient's own record with actor, purpose, and time.
Care CircleAudit drawer

Support human-in-the-loop accountability

AI drafts. Devices observe. A licensed human decides. Nothing leaves the workspace without an attributed approval.

  • Every AI output is labeled proposed — ambient notes, triage suggestions, care plans, escalations — and requires care manager approval before it files.
  • Concurrency ledger records who filed each encounter minute against which code family, with model version stamped.
  • Admin audit log captures role changes and sign-ins with actor, IP, and user-agent.
Today queueEncounter workspaceAdmin audit

Support caregivers' well-being and reduce burden

The system does the sorting, drafting, and code-picking so the care manager can spend the visit being human.

  • Prioritized Today queue replaces the inbox — the next best action is already surfaced.
  • Time accrues passively during work; code family is inferred and defensible at file time.
  • Ambient SOAP drafts, cohort-framed predictions, and plan-studio starters cut cold-start writing to near zero.
TodayPlan studio

Supplement, not replace, human connection

Automation earns the care team more time with the patient — it never stands in for them at moments that matter.

  • Virtual clinic kanban keeps the care manager in the room; the video panel is the primary surface, notes are secondary.
  • Companion escalates to a human on any risk-tier trigger; kill-switch is one click.
  • Care plans require named human touchpoints (RN, pharmacist, dietitian, social worker) across the 90-day horizon.
Virtual clinicCompanion

Allow personalized and flexible care

One plan per person, grounded in that person's twin — not a template applied to a diagnosis code.

  • Plan studio drafts a 90-day plan from the patient's own observations, predictions, language, and risk tier.
  • Any section can be regenerated with a plain-English instruction; thresholds, cadences, and modalities are per-patient.
  • Scenario selector lets the care team rehearse the same plan against high-risk, stable, and post-discharge cohorts.
Plan studioPatient predictions

Promote safety, reliability, and transparency

Every prediction shows its cohort, model version, and drivers. Every device stream shows its FDA class and provenance.

  • Predictions carry cohort framing, model version, calibration bin, and driver attribution in plain language.
  • Device stream cards show FDA class, transmission counts, and last-seen. Missing data is labeled, not imputed silently.
  • Model evaluations and subgroup fairness metrics are versioned alongside MPFS/OPPS rule versions — invariants documented in docs/invariants.md.
DevicesPatient predictions

Ensure affordability and access

Modality follows the patient. Every billable minute is captured defensibly so care reaches people who couldn't otherwise afford it.

  • Virtual clinic and async touchpoints let one care manager reach a rural panel without a windshield tour.
  • CCM, RPM, RTM, BHI, and TCM code families are all first-class — the concurrency ledger prevents double-billing while capturing every eligible minute.
  • Care Circle brings the family into the plan without requiring them to travel or take time off work.
Virtual clinicToday
Who we serve

Caregiving, as the National Strategy defines it

Advisor Outcomes Platform is built for care delivered where people live — homes and communities, not institutions. That includes family caregivers of every age and relationship, direct support professionals working across community settings, and the licensed care managers who coordinate around them.

Family caregivers

Spouses, adult children, sandwich-generation parents, grandparents, siblings, friends, and neighbors — living together or apart. Rarely formally trained; carrying medication management, symptom monitoring, transportation, and health decisions alongside their own lives.

Care CircleCompanion
Direct care workers

Personal care attendants, home health aides, CNAs, direct support professionals, job coaches, and employment support staff. Community-based, individualized, and — per ANCOR 2025 — critically understaffed. Tooling must reduce burden, not add documentation.

TodayEncounter workspace
Care managers (RN / SW)

The licensed accountability layer. Coordinate across health and social providers, own the plan, sign the note, file the code. Everything AI drafts is theirs to approve, edit, or discard.

TodayPlan studioVirtual clinic
Eligible care settings in scope
  • Care in the home and the community
  • In-home hospice
  • HCBS providers delivering community-based day services
  • Employment support and vocational rehabilitation
  • Adults and children with disabilities, including neurodivergent
  • Serious/progressive illness — dementia, cancer, I/DD across the lifespan
Out of scope for this challenge: institutional settings such as assisted living and memory care facilities.
Technology readiness

TRL 6 — Application Development

Advisor Outcomes Platform exceeds the Phase 1 minimum of TRL 3. The workspace demonstrated here is a running, end-to-end system with real schema, live AI drafting, versioned rule sets, and audited role-based access — ready to be robustified against one or more care management use-cases.

Eligibility: Phase 1 ≥ TRL 3 ✓
  1. TRL 0-2
    First principles → Proof of principle
    Care management invariants and bitemporal data model designed.
  2. TRL 3
    Systems development
    Schema, RLS, concurrency ledger, and agent registry deployed.
  3. TRL 4
    Proof of concept
    8-patient demo cohort with observations, predictions, and filed encounters.
  4. TRL 5
    ML capability (R&D → product)
    Ambient notes, triage, and plan-studio drafting via managed AI gateway, cohort-framed predictions.
  5. TRL 6
    Application development
    Care Manager workspace live: Today, Patients, Virtual clinic, Plan studio, Companion, Admin.
  6. TRL 7
    Integrations
    EHR (FHIR), device fleets, MPFS/OPPS rule feeds, identity federation.
  7. TRL 8
    Mission-ready
    Regional pilot with audited billing outputs and subgroup fairness reports.
  8. TRL 9
    Deployment
    Multi-tenant production with continuous model evaluation.

A note on scope. These principles are architectural commitments, not marketing copy. The invariants that make them true — attributed time capture, versioned rule sets, cohort-framed predictions, typed consent grants, kill-switched agents — are documented in docs/invariants.md and enforced at the schema layer so that no future feature can regress them silently.