All use cases

Engage / Lifecycle

Treatment-stage hero card with guidance-only feedback

The first thing a GLP-1 patient sees when they open the app speaks to exactly where they are in treatment — and never crosses into telling them what to do with their dose.

Get aheadBehavioral triggerSystem triggerIn-appPushGLP-1 & PeptidesHealth & TelehealthSubscription

What it does

Most home screens show the same static surface to a patient on week one and a patient on month nine. But the moment that matters changes constantly: a logged symptom, a dose taken off schedule, protein progress for the day. The agent powers a home-screen hero card whose feedback is generated from where the patient sits in treatment and the events they just logged, so the first thing they see is relevant rather than generic — which is what keeps them opening the app and staying on protocol. By design it is guidance-only: it reflects, encourages, and points to the right next step, and it stays firmly out of clinical and dosage territory, with every output quality-gated before it shows.

How it works

  1. 1
    Trigger. the patient opens the app, or an event fires — a logged symptom, a missed or off-schedule dose, a protein-goal update — paired with the patient's current treatment stage.
  2. 2
    Decision. the agent reads treatment stage and the triggering event from memory, selects the guidance-appropriate message, and a quality judge checks every output for tone and the hard clinical / dosage boundary before it renders.
  3. 3
    Action. the hero card on the home screen shows the generated feedback — a stage-aware nudge, an encouragement pegged to today's progress, or a pointer to log or to ask the care team — with an optional push if the patient is away.
  4. 4
    Follow-through. what the card showed and how the patient responded is written back to memory so the next render is continuous; anything that reads as a clinical concern is surfaced for the care team rather than answered, and dose questions are routed, never resolved on the card.
Example
Agent (hero card)You logged some nausea after yesterday's dose. Around this point in treatment that's common and usually settles. Lighter meals early in the day tend to help — want the quick guide? If it's getting worse, I can pass this to your care team instead.
PatientIt's manageable, just annoying. Show me the guide.
Agent (hero card)Here it is. You also hit your protein goal three days running this week — that's the part that protects your results. Nice work; I'll keep an eye on how the nausea tracks.

Configuration

How the agent is wired for this use case.

Triggeran app open or a logged event (symptom, missed / off-schedule dose, protein progress) joined with the patient's treatment stage, sourced from the app backend and dose-logging system.
Tools & actions
  • App backend · reads treatment stage, logged symptoms, dose-log status, and protein progress; renders the hero card.
  • Memory store · reads prior card history and responses; writes each render and reply back for continuity.
  • Knowledge base · retrieves stage-appropriate, guidance-only content (symptom comfort, expectation-setting, next step).
  • Messaging channel · sends an optional push when the patient is away from the app.
  • Care-team escalation path · surfaces clinical concerns and routes dose questions to a human.
Autonomythe hero card renders unattended only after a quality judge clears each output against a guidance-only policy class; the card never recommends, confirms, or changes a dose and never diagnoses — a clinical concern is surfaced to the care team and a dose decision is routed to a clinician, never resolved on the card.
Channelsin-app · push
Escalationa clinical concern or a dose decision hands off to the care team.

What you need

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

Signals

app-open events, logged-symptom events, missed / off-schedule dose events, protein-progress updates, treatment-stage milestones.

Data

current treatment stage, dose-log status, symptom and protein logs, hero-card history and responses in memory.

Guardrails

judge review on every generated output; a hard guidance-only boundary — no dose recommendation, confirmation, change, or diagnosis; clinical concerns and dose questions route to a clinician; frequency caps on the optional push.

Metrics it moves

  • dau-mauup, a home screen that speaks to the patient's current moment is a reason to keep opening the app.
  • adherenceup, stage-aware prompts on logged symptoms and off-schedule doses keep the patient on protocol.
  • churndown, a relevant daily surface sustains the engagement that is the program's retention.
  • ltvup, longer adherent membership on high-CAC patient economics.

Related use cases

Engage / Lifecycle

Proactive health-coach companion

the coaching engine the hero card surfaces inside the app

Engage / Lifecycle

Between-doses lull check-in

the weekly-trough cadence that complements the home-screen surface

Retain / Save

Month-4 cliff companion coach (GLP-1)

the highest-risk window the same engagement loop defends

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