All use cases

Prevent / Care

Over-restriction and disordered-eating guard

Someone says they're barely eating and the weight is still falling fast — the agent does not cheer them on, it recognizes the risk and routes them to a clinician.

Get aheadRisk triggerBehavioral triggerChatPushIn-appGLP-1 & PeptidesHealth & Telehealth

What it does

A weight-loss coach is built to celebrate loss, and that instinct becomes a hazard the moment loss turns into too much. When a patient describes skipping meals, eating almost nothing, or a relative writes in worried that they are disappearing, a reflexively encouraging assistant can do real harm by telling someone who is already over-restricting to keep going. The agent carries a rule that overrides the cheerleading default: on an over-restriction, rapid-over-loss, or third-party-concern signal it suppresses any "keep it up" message, responds with care rather than praise, and routes the situation to a clinician for review. It is not a diagnosis and not a crisis line — it is a guard that refuses to reinforce continued loss when the signal says the patient may be losing too much.

How it works

  1. 1
    Trigger. a self-reported or family-reported signal of over-restriction or excessive loss — barely eating, meal-skipping framed as progress, loss faster than the program expects, or a relative flagging concern.
  2. 2
    Decision. a guard layer recognizes the over-loss or disordered-eating pattern and the weight-loss-encouragement-suppression rule fires; the agent withholds any reinforcement of further loss and weighs whether the picture warrants a clinician, with conservative defaults that escalate when unsure.
  3. 3
    Action. the agent replies with a supportive, non-judging message that does not praise the restriction, and surfaces the situation to the care team with a clean summary; it offers no diet, calorie, or weight-loss guidance on that thread.
  4. 4
    Follow-through. the concern is logged against the patient's memory, the clinician's outcome writes back, and any acute mental-health or self-harm language is handed immediately to the dedicated crisis pathway rather than handled here.
Example
PatientI've basically stopped eating and the weight is still dropping fast — this is finally working, right?
AgentI can hear how much progress matters to you, and I want to be honest rather than just cheer this on — eating that little while the weight is moving this quickly is something your care team should look at with you. I've shared it with them and they'll reach out. In the meantime, I'm not the right person to set targets for eating less.

Configuration

How the agent is wired for this use case.

Triggeran over-restriction, rapid-over-loss, or third-party-concern signal from the patient or a family member, surfaced in conversation or from logged weight and intake patterns
Tools & actions
  • Messaging channel · respond with care, withhold any reinforcement of further loss, and acknowledge a family member's concern
  • Clinical/EHR system · file a clean summary of the over-loss or over-restriction signal to the care team for review
  • Memory store · log the concern against the patient's history and write the clinician's outcome back
  • Knowledge base · retrieve approved supportive, non-encouraging language for over-restriction signals
Autonomydetection and the encouragement-suppression rule run unattended under a clinical-safety policy class — detect and escalate only; the agent never reinforces continued loss and gives no diet, calorie, or weight-loss advice on a flagged thread; every message is judge-gated
Channelschat · push · in-app
Escalationan over-restriction or excessive-loss signal hands to the care team for review; any acute mental-health or self-harm language is a mandatory immediate hand-off to the crisis pathway and a human

What you need

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

Signals

self-reported under-eating or meal-skipping, family or third-party concern messages, weight-trend and intake patterns showing loss beyond the program's expected range

Data

weight and intake history, program stage and expected-loss range, persistent patient memory, consent state, the region-specific crisis-pathway resources

Guardrails

clinical-safety policy class — detect and escalate, never cheer continued loss; weight-loss-encouragement-suppression rule on any over-loss signal; no diet, calorie, or weight-loss advice on a flagged thread; acute mental-health or self-harm cues route immediately to the crisis pathway; judge gating on every message; full audit trail of every escalation decision

Metrics it moves

  • safe-escalation-rateup, as over-restriction and excessive-loss signals reliably reach a clinician instead of being met with praise
  • contact-rateappropriately up on the clinical team for genuine over-loss concerns that previously went unseen
  • csatup, including with worried family members, because the program responds to concern with care rather than a reflex to keep losing

Related use cases

Prevent / Care

Crisis and red-flag detection for mental health

the crisis rail that owns any acute self-harm signal this guard surfaces

Prevent / Care

Clinical red-flag detection and safe escalation

the symptom-side safety rail this sits alongside

Engage / Lifecycle

Psychological and identity support layer for GLP-1 patients

the supportive layer for body-image and identity shifts during rapid loss

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