Privacy first: 3JAL App reads work-pattern metadata only — meeting hours, app-switch counts, after-hours activity. It never reads message content, screens, or keystrokes. Employees are always nudged privately first; HR is looped in only when indicators stay high.
Team wellbeing — 3JAL App product org
Employee
Stress index
Band
Trend
Early-support reports
How it works
1 · Signals we use
Meeting hours and back-to-back blocks
Task / app switching frequency
Work activity after office hours
Unbroken focus time available
2 · Never collected
Message or document content
Screen capture or webcam
Keystrokes or mouse tracking
Anything from personal accounts
3 · Employee first, always
AI warns the employee privately, with concrete recommendations
HR is informed only if indicators don't improve for days
Reports ask for support, never discipline
What's real in this demo: the backend (Cloudflare Workers API + KV storage), the stress scoring, the escalation flow, and every bot message and HR report are generated live by a real LLM (Llama 3.3 70B on Workers AI) — press “Simulate next day” and watch. What's simulated: the Teams telemetry itself; in production the same signals come from the Microsoft Graph API inside the company's own tenant, and 3JAL App ships as a Teams app + laptop agent.