AI never makes clinical decisions
The recommendation engine produces candidate care plans. The Care Navigator approves the plan. The clinician adjusts in session. The algorithm's job is to surface options; the human's job is to choose.
Refill Health is a clinical mental healthcare platform first — built to use AI in service of care, not as a substitute for it. The Refill Health AI Suite is seven named products — built around a predictive intelligence layer — designed to make care faster, more precise, and more measurable. Every clinical decision rests with a human. Every algorithm is supervised by clinicians. Every output is auditable.
THE PRINCIPLE
Mental healthcare is too important for unsupervised automation. Refill Health's AI is operationally bounded, clinically supervised, and continuously audited — every clinical decision rests with a human.
The recommendation engine produces candidate care plans. The Care Navigator approves the plan. The clinician adjusts in session. The algorithm's job is to surface options; the human's job is to choose.
Every AI output that touches a member's care is monitored by Refill Health's clinical leadership. Models are reviewed for clinical appropriateness, drift, and bias. Models that produce unexpected outputs are paused and reviewed before continued use.
Every algorithmic recommendation that influences a care plan is logged with its inputs, its reasoning, and the human decision that followed. A clinical reviewer or procurement auditor can trace any recommendation back to its source signal.
Clinical note: The operational supervision of the AI suite is owned by Refill Health's clinical leadership team — the same clinicians who supervise the network and own the clinical model.
THE PREDICTIVE INTELLIGENCE LAYER
Most workforce mental health platforms are reactive. Members reach out when they're already struggling; the platform responds. Refill Foresight™ is built so the platform sees what is coming and acts before the member has to ask. Foresight is what makes Refill Health a proactive platform.
| QUESTION | WHAT FORESIGHT SURFACES | WHERE IT REACHES THE HUMAN | |
|---|---|---|---|
| 01 | Who may need support now? | Members whose recent screening signals, engagement patterns, or behavioural indicators suggest unaddressed clinical need — even if they haven't yet reached out. | Care Navigator dashboard (Refill Navigator™) |
| 02 | Who is likely to worsen if unsupported? | Members whose trajectory data indicates a likely decline in clinical state without intervention — flagged for proactive outreach. | Care Navigator dashboard; session-prep view (Refill Notes™) |
| 03 | What level of care is appropriate? | Risk-stratified care plan recommendations — therapy, coaching, or self-care — based on validated clinical signals and member preferences. | Member matching flow (Refill Match™) |
| 04 | Who may disengage from care? | Members whose engagement signals suggest disengagement risk before the next scheduled session — surfaced early enough for the Navigator to act. | Care Navigator dashboard (Refill Navigator™) |
| 05 | When should a Care Navigator intervene? | Specific moments when human outreach is more likely to succeed — based on member context, engagement state, and clinical signals. | Care Navigator dashboard (Refill Navigator™) |
The platform surfaces signals; humans act on them. Foresight does not contact members, does not change care plans, does not escalate to crisis. Predictions are bounded by the principle in Section 2. Every output is logged, supervised, and auditable. Foresight is what makes the rest of the suite work. Match, Navigator, Notes, and Insights all draw on Foresight's outputs.
Methodology note:
Detailed model documentation — including signal definitions, prediction methodology, supervision protocols, and accuracy review cadence — is available in the clinical procurement demo.
THE REFILL HEALTH AI SUITE
Refill Health's AI suite is designed to integrate seamlessly across the clinical lifecycle, from initial matching to ongoing supervision and outcome measurement.
Proprietary matching algorithm that uses clinical state, lived experience, and member preference to ensure the right fit from day one.
Assists Care Navigators in creating risk-stratified care plans and managing proactive outreach for members at risk of disengagement.
Ambient clinical documentation that transcribes sessions, extracts clinical themes, and drafts SOAP notes for clinician review.
Enterprise-level analytics that surface population health trends, clinical outcomes, and ROI data for HR and benefits leaders.
A member-facing tool providing AI-guided self-care exercises, mood tracking, and crisis detection between clinical sessions.
Provides clinicians with real-time training and performance feedback based on session patterns and outcome trends.
Methodology note: Detailed product roadmaps and technical specifications for each AI suite component are available for review during the clinical due diligence phase.

FEATURED PRODUCT
Transforming the administrative burden of clinical work into a source of insight.
Foresight™ surfaces key clinical themes from member check-ins and Companion™ data, giving the clinician a rich starting point for the conversation.
Ambient recording captures the session with medical-grade precision. Clinicians focus on the member, not the keyboard.
Notes™ generates a high-fidelity SOAP note and identifies potential risks for follow-up. The clinician reviews, edits, and signs off in minutes.
Refill Notes™ doesn't just save time; it improves the quality of care. By providing a structured, high-fidelity record of every interaction, it ensures that clinicians and care managers have a shared, accurate understanding of member progress and risk.
Clinical note: All ambient recording is subject to explicit member consent and data is encrypted and handled in strict accordance with HIPAA and regional privacy regulations.
WHAT THIS MEANS FOR EACH AUDIENCE
The Refill Health AI Suite serves four distinct stakeholders. Each audience sees only what they need — and only the parts they're authorised to see.
Faster matches. Better care. Smoother journey.
AI helps members reach the right care in less time and with less guesswork... Between sessions, contextually-appropriate self-care content keeps the work going.
PRODUCTS: REFILL MATCH™ · REFILL COMPANION™
Earlier signals. Better priorities. Fewer cases that fall through.
AI helps Care Navigators see who needs attention before the next scheduled touchpoint. Drop-off risk, outcome trajectory shifts, and crisis flags surface in a single dashboard view...
PRODUCTS: REFILL NAVIGATOR™
More context. Less paperwork. More time with members.
AI gives clinicians better preparation before each session, structured documentation during, and finished notes immediately after... Hours that used to disappear into documentation come back to members.
PRODUCTS: REFILL NOTES™ · REFILL MATCH™
Anonymised insights. Engagement campaigns. Defensible ROI.
AI transforms anonymised workforce data into engagement patterns, outcome trajectories, and ROI signals. Recommended engagement campaigns target cohorts whose signals warrant attention...
PRODUCTS: REFILL INSIGHTS™
ETHICAL AI
Most AI ethics statements are aspirational. The five commitments below are operational — each with a defined practice and a defined check.
Data minimisation across every product. Encryption at rest and in transit. India-resident data architecture. The principle is enforced in code, not in policy documents.
Outcomes are tracked across cohorts (gender, role, function, geography) for parity. Where outcome gaps emerge, models are paused and reviewed before continued use.
Members can ask why a particular care recommendation was surfaced. The Care Navigator can explain the inputs that shaped the recommendation in plain language.
Each product has a defined scope of action. Refill Companion never accesses clinical session data. Refill Insights never accesses individual member records. Boundaries are enforced architecturally.
Model performance, drift, output quality, and outcome equity are reviewed by clinical leadership and engineering on the same cadence as clinician supervision.
Methodology note: Detailed governance documentation — including audit cadences, supervision protocols, bias-monitoring methodology, and incident response — is reviewed in a clinical procurement demo.
WHAT WE PUBLISH VS WHAT WE DEMO
AI is the most IP-sensitive topic in our platform. This page covers the principles... The operational detail is reserved for the clinical procurement demo.
The demo is built around your procurement requirements, not around our marketing slides. It is the right next step for any procurement review that requires operational depth.
DEMO CTA
Write to clinical@refillhealth.com to start.
Refill Health uses AI to make care faster, more precise, and more measurable — bounded by clinical supervision, accountable to human judgement, and audited continuously. The architecture is on this page. The detail comes in the demo.