Selected as Best Overall Capstone

Harvard Medical School Executive Education, "AI in Healthcare," February 2026

HCP-as-Pilot™ v3.5, Updated June 2026 — now with Runtime Governance Infrastructure (RGI) Read the Paper →
Home Companion icon
Home Companion
Independence. Dignity. Peace of mind.

Home Companion™

Phase I Deployment Proof — real-world governance evidence from live home care.

The first proof environment for Safety OS™.
14.7M
Seniors living alone in the US
US Census Bureau
35%
Increased mortality risk from social isolation
Nakou et al., Aging Clinical & Experimental Research, 2025
4 in 10
Adults 45+ report feeling lonely
AARP Loneliness Study, 2025
$390B
Global senior living market by 2032
Coherent Market Insights, 2025

Safety OS™ · Phase I Deployment

Home AI is already here.
Runtime governance is not.

PatientCentricCare.AI builds runtime governance infrastructure for AI systems operating in home-care environments.

Request a Pilot → See How It Works ↓
Harvard Medical School — Best Capstone 2026
Pilot active — Bethesda Alterszentren, Basel
EU AI Act compliant architecture

Companionship That Adapts To The Person

Multilingual, memory-aware, culturally adaptive — built for real lives, not demos.

“I wasn’t sure at first, but then I realized Home Companion helps me organize my day and reminds me about important things. I love that it switches languages instantly and can talk about anything. I learn every day.”

— Resident, Basel pilot

“My daughter worries less now. She knows I have someone to talk to, and that if something goes wrong, the right person gets called.”

— Resident, Phase I deployment

“I remember my appointments better. It reminds me gently, in my own language, at the right time.”

— Resident, Basel pilot

“I feel less alone. It’s not a nurse, it’s not a doctor — it’s just there, and it listens.”

— Resident, Phase I deployment

🌐Instant language switching — German, French, English, Italian, and more
🕑Memory prompting — appointments, routines, medication reminders
👪Caregiver summaries — peace of mind without surveillance
🎨Culturally adaptive — dialect continuity, personal preferences
Request a Pilot →

Phase I: Live Deployment Proof

Non-medical companionship in live senior-care settings — generating governance evidence before clinical expansion.

  • Conversational companionship within defined boundaries
  • Scheduled interaction cycles with memory continuity
  • Human-confirmed escalation only
  • No diagnosis, no medical advice — deliberately non-clinical
  • Every interaction logged in tamper-evident audit trail
Home Companion app - senior-friendly 3-button phone interface for independent living

Phase I: smartphone app with embedded governance kernel

Governance Architecture

Every interaction passes through seven governance layers before reaching the user.

Safety OS governance flow: User → Interaction Layer → Safety OS Runtime Enforcement → Consent Gate → Escalation Logic → Audit Recorder → Human Authority

Governed. Auditable. Human-led. Every interaction is safe, respectful, and accountable.

How It Works in Practice

Three real-world scenarios from live Phase I deployment.

Morning Check-in

1

Home Companion initiates scheduled greeting

2

Resident responds — mood and engagement logged

3

If distress detected → escalation gate triggered (consent required)

4

Caregiver notified only with resident’s permission

💊

Medication Reminder

1

Scheduled reminder delivered in resident’s preferred language

2

Resident acknowledges or defers

3

Non-acknowledgement after threshold → escalation offered

4

No inference about health — only structured notification

👪

Family Connection

1

Caregiver requests wellbeing summary

2

System checks resident’s consent preferences

3

Summary generated from governance metadata only — no raw audio shared

4

Caregiver receives structured update; resident retains control

Download Governance Brief ↓

Deployment Metrics

Operational numbers from Phase I — Bethesda Alterszentren, Basel.

System Uptime
99.2%
Operational
Interactions Logged
2,400+
Growing weekly
Escalation Response
< 90s median
Within target
Consent-Confirmed Escalations
100%
Zero overrides
Multilingual Success Rate
97.8%
DE / FR / EN / IT
False Escalation Rate
< 2%
Below threshold

Flight Recorder

Tamper-evident audit trail — every interaction classified, timestamped, and stored.

Flight Recorder audit log preview
  • Every AI action recorded with authority level
  • Consent status verified before escalation
  • Boundary refusals logged with context
  • Regulatory-ready evidence for EU AI Act compliance

Same audit standard regulators will require for high-risk AI.Built from day one — not retrofitted.

Acoustic Privacy by Design

Voice interactions processed locally. Raw audio never stored. Only governance metadata is retained.

  • No cloud recording of conversations
  • Only structured metadata logged (consent, escalation, authority)
  • GDPR and UK data protection compliant
  • Resident retains full control over data sharing

See the governance model in action

Request a Pilot →
Access Deployment Evidence ↓

Why Now: The Governance Window

Regulatory inevitability creates a narrow window for first-mover advantage.

2025
Conversational AI enters homes at scale — without governance
2026
EU AI Act enforcement begins — Home Companion generates compliance evidence
WE ARE HERE
2027
First enforcement precedents — ungoverned systems face sanctions
2028
Payer requirements emerge — governance evidence becomes table stakes
2029+
Clinical expansion — only systems with proven governance history qualify

Designed to Scale — Only When Earned

Governance maturity unlocks capability expansion. Never the reverse.

Phase I

Non-medical companionship. Governance proof.

Phase II

Structured wellbeing monitoring. Expanded boundaries.

Phase III

Clinical-grade integration. Regulatory inheritance.

  • Each phase requires demonstrated governance maturity from the previous
  • Regulatory inheritance: evidence from Phase I directly supports Phase II authorization
  • No capability expansion without proven safety at the current tier

Why Non-Medical First Is Strategic

Deliberate scope constraint that accelerates the long-term regulatory pathway.

Faster Deployment

Non-medical scope avoids SaMD classification, enabling immediate real-world deployment and evidence generation.

Governance Evidence

Every interaction generates the audit data regulators will require for higher-risk AI — built before it’s mandated.

Regulatory Inheritance

Phase I evidence directly supports Phase II/III authorization — creating a compounding compliance advantage.

PatientCentricCare.AI — Deployment Proof

Home Companion:
Phase I Proof Environment

Runtime-governed home care. Real-world evidence. Scalable only when earned.
Home Companion is not a product launch.
It is the first deployment proof for runtime-governed healthcare AI.
Every interaction governed. Every escalation logged. Evidence generated before clinical expansion.
What the Deployment Proof Demonstrates
🏠
Home Companion
Safety OS™ Control Plane
  • Governs every AI interaction
  • Enforces escalation and boundaries
  • Logs full audit trail
  • Generates real-world evidence (RWD)
Ecosystem Logic — Each Stakeholder Unlocks the Next
🏥
NHS / Care Systems
Access + Pilot Environments
They Give
Access • Patient cohorts • Pilot sites • Protocols
They Get
Safe AI deployment • No integration burden • Audit trail • EU AI Act pathway
Urgency
Runtime enforcement will be required. Policy documents alone will not be enough.
↓ enables evidence generation
💊
Pharma Sponsors
The RWE Funding Layer
They Give
Sponsorship capital • RWE programme budget • Strategic input
They Get
Behavioural RWD in early cognitive decline + adherence — outside the clinic, at scale
Insight
No clinical trial can generate this data.
↓ funds and validates
🤝
Patient Advocacy Groups
Trust + Distribution Layer
They Give
Credibility • Endorsement • Patient trust • Distribution reach
They Get
Governed AI that preserves patient autonomy in the home • Ethical deployment • Policy influence
Urgency
AI is entering home care without governance. Patient protection must lead.
↓ de-risks and amplifies
📈
Investors
Capital + Category Acceleration
They Give
Growth capital • Strategic network • Category conviction
They Get
Defensible infrastructure moat • First-mover RWD dataset • Regulatory defensibility
Urgency
Category window closes as EU AI Act enforcement begins (Aug 2026).
The Flywheel
Each step compounds the next
1
🏥
NHS enables safe pilot environments
2
📊
Real-world evidence is generated
3
💊
Pharma funds RWE programmes
4
🤝
PAGs validate and distribute
5
🔄
Capital scales deployment
Why the Model Compounds
Real-world evidence (RWD)
Repeatable pilot model
Regulatory pathway data
Multi-stakeholder alignment
Moat
The first governed AI interaction dataset in home-based senior care — impossible to replicate without running the pilots
PatientCentricCare.AI — Runtime Governance Infrastructure
Confidential · 2026

Important Notice: Home Companion is a non-medical system. It does not provide diagnosis, treatment, or clinical decision-making. Escalation is permitted only with user consent and subject to governance controls. Future phases are subject to regulatory review and formal authorization.

Governance-first AI infrastructure for the home.

Built deliberately. Scaled only when earned.

Empowering independence.
Preserving dignity.

Frequently Asked Questions

What makes Home Companion "non-medical"?

Home Companion supports daily living — reminders, companionship, wellbeing check-ins — without providing clinical diagnosis, prescriptions, or treatment. This non-medical scope means faster deployment and lower regulatory burden, while still generating governance evidence that builds toward clinical-grade AI.

How does the audit trail work?

Every interaction is logged in a tamper-evident "Flight Recorder" — recording what the AI did, what authority it operated under, whether consent was given, and whether escalation was triggered. This produces the same kind of governance evidence regulators require for high-risk AI systems under the EU AI Act.

Is this ready for deployment in my organisation?

Phase 1 is designed for care homes and home care environments. Deployment requires minimal IT integration — the system operates independently with its own governance layer. Contact us to discuss a pilot in your setting.

What data is collected and how is privacy protected?

Home Companion uses acoustic privacy by design — it processes voice interactions locally and only records governance metadata (consent given, escalation triggered, authority level). Raw audio is not stored. All data handling is designed to meet GDPR and UK data protection requirements.

Continue Reading

Request a Pilot →