All use cases

Prevent / Care

Pancreatitis and gallbladder red-flag detector

A patient on a GLP-1 medication describes severe stomach pain that won't let up and is creeping toward their back — the agent treats it as an emergency, not a chat to soothe.

Get aheadRisk triggerChatPushVoiceGLP-1 & PeptidesHealth & Telehealth

What it does

On a GLP-1 medication, persistent severe abdominal pain — especially pain that radiates to the back, with or without vomiting — is a recognized signal of pancreatitis or a gallbladder problem, not a side-effect to wait out. The danger is a reassuring reply: a patient describes exactly that picture, gets told it is probably the usual nausea, and loses hours that matter. The agent does the opposite. It carries this specific symptom pattern as a hard rule on the safety rail, recognizes it the moment it is described in any channel, and escalates straight to urgent care or a clinician with the full picture. It never reassures, never offers a self-care tip, never tries to resolve it in the conversation.

How it works

  1. 1
    Trigger. a free-text symptom description, or a logged symptom on the medication, that matches the pancreatitis/gallbladder pattern — severe persistent abdominal pain, radiation to the back, accompanying vomiting.
  2. 2
    Decision. the agent matches the description against the encoded red-flag pattern for this drug class and checks it against the patient's medication memory; this pattern is a designated hard-escalate, so there is no self-care branch and no severity negotiation — a match routes immediately.
  3. 3
    Action. the agent gives the patient a plain, bounded instruction to seek urgent care now, and routes a clinician handover carrying the symptom description, medication, and stage; on voice, it stays on a steady, scripted urgent-care prompt.
  4. 4
    Follow-through. automated coaching and any pending nudges on that thread stand down, the escalation and its full context write to memory, and the clinician's outcome closes the loop so nothing is dropped.
Example
PatientI've had really bad stomach pain since last night, it's not easing off and now it's going through to my back. I've been a bit sick too.
AgentThat combination on your medication needs to be seen urgently — please contact urgent care or go in now rather than waiting for this to pass. I've flagged your care team with exactly what you've told me and which medication you're on, so they have the full picture the moment you're in touch. I'm not able to advise on this one myself — getting you seen is what matters.

Configuration

How the agent is wired for this use case.

Triggera symptom description or logged symptom matching the pancreatitis/gallbladder red-flag pattern (severe persistent abdominal pain, radiating to the back, with or without vomiting) on an active GLP-1 medication
Tools & actions
  • Messaging channel · receive the symptom description across chat, push, or voice and hold a bounded urgent-care script
  • Knowledge base · match the description against the encoded red-flag pattern for the drug class
  • Memory store · confirm the active medication and stage, write the escalation and clinician outcome back
  • Clinical/EHR system · route an immediate clinician handover with the symptom description, medication, and context
  • Scheduling · surface the urgent-care or clinician path so the patient knows where to go now
Autonomydetection and the immediate hard-escalate run unattended under a clinical triage policy — detect and escalate only, no dose advice, no diagnosis, no self-resolution; this pattern is a designated hard-escalate with no self-care branch, and every patient-facing line is judge-gated
Channelschat · push · voice
Escalationany match hands off to a clinician or urgent care immediately and stands automated replies down on that thread; ambiguous-but-concerning descriptions escalate under the conservative default rather than being resolved in chat

What you need

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

Signals

in-conversation symptom descriptions and logged symptoms on the medication; medication and program-stage events

Data

active medication and stage, the encoded pancreatitis/gallbladder red-flag pattern for the drug class, persistent patient memory, consent state

Guardrails

clinical triage policy with conservative defaults — detect and escalate only; no dose advice, no diagnosis, no self-resolution; designated hard-escalate pattern with no self-care branch; judge gating on every message; automated replies stand down once a clinician owns the thread; full audit trail of every escalation

Metrics it moves

  • safe-escalation-rateup, because the highest-acuity GLP-1 symptom pattern reliably reaches a clinician instead of getting a soothing reply
  • time-to-resolutiondown, since a recognized emergency routes the patient to urgent care in the moment rather than after hours of waiting
  • contact-ratemeasured as the share of red-flag descriptions caught and escalated, evidence the safety rail is doing its job

Related use cases

Prevent / Care

Clinical red-flag detection and safe escalation

the general safety rail this hard-codes a specific emergency pattern onto

Prevent / Care

GLP-1 side-effect education with red-flag escalation

the everyday side-effect layer that routes here when a description turns acute

Prevent / Care

Symptom-triage urgency-ranking layer

orders many symptoms by acuity; this card is the top of that order

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