All use cases

Prevent / Care

Pre-symptom illness alert

The morning before the sore throat shows up, the watch already sees it — the agent quietly tells you to slow down and rest, days before you would have known to.

Get aheadRisk triggerSystem triggerPushIn-appChatWellness (B2C apps)Health & Telehealth

What it does

A wearable streams skin temperature, resting heart rate, and heart-rate variability every night, and a few days before someone feels a cold or flu coming on those readings often drift together in a recognizable precursor pattern. Most apps bury that drift inside a dashboard the member never opens, so the one window where rest actually changes the outcome passes unnoticed. The agent watches for that illness-precursor deviation against the member's own baseline and reaches out before any symptom appears, with a steady, non-clinical message to prioritize rest, hydration, and an easier day. It is one of the most striking get-ahead moments a consumer wellness product can offer, and the kind of save that makes wearing the device feel worth it.

How it works

  1. 1
    Trigger. an overnight biometric sync produces an illness-precursor signal — elevated skin temperature plus a resting-heart-rate rise and a heart-rate-variability drop, deviating together from the member's personal baseline.
  2. 2
    Decision. the agent confirms the deviation is meaningful against this member's own trend rather than a one-night blip or a known confounder (a hard workout, alcohol, travel), checks it has not already flagged this episode, and decides whether a caring heads-up genuinely helps now.
  3. 3
    Action. it sends one concise message on the member's channel naming what shifted in plain language and suggesting lifestyle steps only — rest, fluids, an easier training day — with every unprompted message passing a quality and tone judge before it leaves.
  4. 4
    Follow-through. the reading and the alert are written to memory so the agent can tell the member when the pattern settles, and if the member replies with anything that reads as an actual medical symptom or asks for a diagnosis, the agent stops guiding and routes to the program's clinical pathway.
Example
AgentHeads up — overnight your skin temperature was up, your resting heart rate was higher than usual, and your recovery dipped, all at once. That pattern often shows up a couple of days before you feel run down. Nothing to worry about, but today's a good day to go easy, keep water close, and get an early night. I'll keep watching with you.
Customerhuh, I do feel a bit off actually. should I train today?
AgentI'd keep it light or rest it out while your body's working on something — your numbers should tell us within a day or two. If a real symptom turns up, like a fever or a sore throat that worsens, check in with a clinician rather than me, and I'll have your recent readings ready to share.

Configuration

How the agent is wired for this use case.

Triggeran illness-precursor biometric event (illness_precursor_pattern — combined skin-temperature, resting-heart-rate, and HRV deviation from baseline) raised on the overnight sync from the wearable/device feed.
Tools & actions
  • Wearable/device feed · ingest nightly skin temperature, resting heart rate, and HRV, and detect the multi-signal precursor pattern against baseline.
  • App backend · read the member's personal baseline and recent trend, and screen out known confounders (hard session, travel, alcohol) before flagging.
  • Member memory · confirm no prior alert is open for this episode and record the alert so the agent can follow up when readings settle.
  • Messaging channel · deliver one judge-gated, lifestyle-only heads-up on the member's preferred channel.
  • Clinical-escalation pathway · hand off to the program's clinical route if a reply surfaces an actual symptom or a request for medical interpretation.
Autonomyprecursor detection, baseline comparison, and the single rest-and-hydration nudge run unattended under judge gating and frequency caps. The agent gives general lifestyle guidance only; it never diagnoses an illness, predicts a specific condition, or gives medical instruction from a biometric reading — anything that reads as a symptom or a clinical question is a mandatory hand-off to the clinical pathway under a clinical-safety policy class (detect, then escalate).
Channelspush · in-app · chat
Escalationa reply reporting a real symptom (fever, worsening sore throat, breathing difficulty), a reading pattern outside safe lifestyle-guidance bounds, or any request for a diagnosis routes to the clinical pathway with the recent readings attached, rather than being answered by the agent.

What you need

The inputs this use case runs on. Your channels stay yours; the agent supplies the judgment.

Signals

nightly wearable events (skin temperature, resting heart rate, HRV) with baseline-deviation thresholds, plus member replies for follow-up.

Data

the member's personal biometric baselines and recent trend, confounder context (workouts, travel), channel and quiet-hour preferences, explicit consent for health-data use, and persistent memory of any open episode.

Guardrails

the agent gives lifestyle and rest guidance only — no diagnosis, no condition prediction, no medication advice from a reading — with a stated boundary and mandatory clinical hand-off on any reported symptom; judge gating on every alert with "stay silent" as a valid outcome; frequency caps and quiet hours; explicit consent and compliant handling of sensitive biometric data; one alert per episode; visible opt-out.

Metrics it moves

  • csatup: a genuinely prescient, caring heads-up before someone feels unwell is a memorable moment that members credit to the product.
  • dau-mauup: a timely, data-tied alert gives the member a reason to open and engage rather than ignore a passive dashboard.
  • adherenceup: members who feel the device looks out for them keep wearing it and stay inside the daily habit loop the program depends on.

Related use cases

Prevent / Care

Overtraining and strain-creep early warning

the sibling biometric alert for sustained strain rather than an illness precursor

Engage / Lifecycle

Wearable-driven rest and recovery nudge

the single-reading rest nudge this builds on with a multi-signal precursor pattern

Engage / Lifecycle

Recovery-score readiness coach

the daily readiness loop the same biometric stream also powers

See it on your own customer journey

Bring one drop-off, one churn cliff, or one silent segment. We will show you what a proactive agent with memory and judgment does with it.

Book a demo