All use cases

Retain / Save

Weight-stall plateau intervention coach

The scale has not moved in weeks and the patient has quietly concluded the medication has stopped working — the agent catches the flat line in the data and reframes it before they decide to quit.

Get aheadBehavioral triggerTime-based triggerChatPushWhatsAppIn-appGLP-1 & PeptidesHealth & TelehealthSubscription

What it does

A real weight plateau is one of the top reasons GLP-1 patients quit somewhere in the middle of their course, and the danger is that it reads to the patient as failure when it is normal biology. By the time someone says "it stopped working," they have usually watched a flat line on the scale for weeks and already half-decided to stop. Rather than wait for that message, the agent watches the logged weight trend itself: when the data goes flat for a sustained stretch it reframes the plateau as expected, coaches the levers that actually help here — protein, resistance training, an honest look at intake — and, if the stall persists, surfaces it to the care team to review dose and response. It coaches behavior and escalates; it never adjusts the dose itself.

How it works

  1. 1
    Trigger. the patient's logged weight stays flat for a sustained stretch of weeks — a data signal, not a calendar date — optionally alongside engagement that has started to slip.
  2. 2
    Decision. the agent confirms the stall against the patient's own trend and memory (titration stage, recent logging, what has already been tried), distinguishes a genuine multi-week plateau from normal day-to-day noise, and decides between a behavioral coaching touch and, for a persistent stall, a care-team surface. A judge reviews every unprompted message.
  3. 3
    Action. a supportive touch lands on a program-owned channel — the plateau-is-normal reframe plus one concrete lever (protein target, a resistance-training nudge, or an intake review) — written to re-open the coach conversation, not to lecture.
  4. 4
    Follow-through. replies continue in the coach chat; a stall that persists past the coaching window is surfaced to the care team to review dose and response, with the trend attached; outcomes write back to memory; queued touches cancel on self-re-engagement, and the lift is measured against a holdout.
Example
AgentYour weight's held flat for about three weeks now. Before you read anything into that — a stall here is normal biology, not the medication quitting on you. The two levers that move it most at this stage are getting enough protein and adding a little resistance training. Want to start with protein this week?
PatientI figured it had just stopped working and I was about to give up on it.
AgentCompletely understandable, and you're far from the first to feel that — but a flat stretch like this is exactly when the basics matter most. Let's set a daily protein target together, and if the scale's still stuck in a couple of weeks I'll loop in your care team to take a look at how you're responding.

Configuration

How the agent is wired for this use case.

Triggera sustained multi-week flat in the patient's logged weight (weight_stall_detected), optionally with engagement decay, sourced from the weight-tracking and program systems.
Tools & actions
  • Weight-tracking / app backend · detect a sustained flat stretch against the patient's own trend and distinguish it from normal fluctuation
  • Patient record / CRM · read titration stage, logging history, and what has already been tried
  • Memory store · read coach-conversation history; write the plateau touch and outcome back for the next touch
  • Messaging channel · send the judged reframe-plus-lever touch on a program-owned channel and re-open the coach chat
  • Clinical escalation path · surface a persistent stall to the care team to review dose and response, with the trend attached
Autonomystall detection and the behavioral coaching touch run unattended, judge-gated on every unprompted message, using approved coaching content only under a never-new-medical-advice clinical policy; any dose or response question is the clinician's, and sends are suppressed on self-re-engagement.
Channelschat · push · whatsapp · in-app
Escalationa stall that persists past the coaching window, or any dose/response concern, is surfaced to the care team with the weight trend attached; clinical questions route to a human.

What you need

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

Signals

logged weight entries and trend, sustained-flat detection, engagement-decay detection, patient replies.

Data

weight history and trend, titration stage, coaching history, persistent memory, channel consent state.

Guardrails

the agent coaches behavior and escalates only — never adjusts or advises a dose; judge review on every unprompted message; clinical questions always route to a human; approved coaching content only; frequency caps; suppression on self-re-engagement; holdout measurement of save-rate.

Metrics it moves

  • save-rateup, by catching the plateau in the data and reframing it before the patient decides to stop.
  • adherenceup, by replacing the "it stopped working" conclusion with a concrete behavioral lever that keeps doses on schedule.
  • churndown, by defending one of the top mid-course quit triggers with a signal-driven touch rather than a calendar guess.

Related use cases

Retain / Save

Drop-off-window cadence coach (GLP-1)

the calendar-window coach this complements with a data-driven trigger

Retain / Save

Four-month drop-off re-engagement

the window-based save fired by the calendar rather than a weight signal

Engage / Lifecycle

Muscle-preservation protein and resistance coaching loop

the standing program the protein and training levers plug into

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