Loading the interactive rules engine… This summary is replaced when it starts.
People often give up a healthy routine after a lapse.
ZenAI Behaviour Twin helps adults in Abu Dhabi restart a movement routine they chose, with smaller, feasible actions instead of resetting their progress. Rules decide; the person chooses. The initial pilot customer is a workplace or public wellness programme that needs adaptive participant support and aggregate-only outcome reporting.
What the demonstration shows (10 steps, synthetic persona)
- The person chooses an action and time
- A seven-day baseline sets the first plan
- One small action, plan version 1
- Daily check-ins
- A lapse is detected
- The plan shrinks without blame; progress is never reset
- The person keeps or undoes the change (24 hours)
- Recovery time is measured
- Programme outcomes for groups of 20 or more only
- A trace of every decision to a versioned rule
How it works
- Six versioned DMN 1.4 decision tables (DEC-01 day status, DEC-02 safety, DEC-03 visibility, DEC-04 message mode, DEC-05 plan change, DEC-06 reminders) run in the page and in the Azure API, with the same answer.
- Live on Azure Container Apps (UAE North): decision API, Application Insights (OpenTelemetry), and Azure OpenAI embeddings that rank approved restart actions after the rules. A generative model is not configured; every message is a reviewed template.
- Safety first: red-flag symptoms stop coaching and show 998; pain pauses suggestions. Wellbeing coaching, not medical advice.
- Standards: DMN 1.4, HL7 FHIR R4 with LOINC, Open mHealth, CloudEvents, W3C trace context.
- Synthetic data only: no sign-in, no cookies, nothing stored. 25 auditor tests and an Architecture & evidence view are built into the app.
If this text stays, JavaScript is turned off or still loading: service health · submission deck (PDF) · auditor guide (PDF).