All use cases

Engage / Lifecycle

Proactive health-coach companion

Between human sessions, every member has a coach who remembers their dose, their last weigh-in, and their goal — and reaches out at the moments that decide whether they stay.

Get aheadNext best thingClose the loopTime-based triggerBehavioral triggerRisk triggerIn-appPushChatSMSWhatsAppEmailHealth & TelehealthSubscriptionWellness (B2C apps)

What it does

A weekly session with a human coach covers an hour. The other 160 hours, the member is alone with a new dose, a side effect, a missed log, and a growing doubt about whether the program is working. In subscription health, continued engagement is the revenue: when the program goes quiet between sessions, the member cancels. The agent is an always-on coach assigned to every member, with one memory across every channel — in-app chat, push, SMS, WhatsApp, email — that knows the current dose, the last weigh-in, the last side effect reported, and the weeks until the next titration step, and uses those moments to keep the member moving toward their own goal.

How it works

  1. 1
    Trigger. program milestones on the clock (dose reminder due, titration step approaching), behavioral signals (missed injection log, a weigh-in gap, engagement decay), and risk signals (a poor-sleep stretch from a connected wearable, weight trajectory drifting from the member's stated goal).
  2. 2
    Decision. the agent reads the member's persistent memory — dose history, side effects, goal, what was discussed last — and weighs whether reaching out now actually helps. Every unprompted message passes a quality judge before it sends; quiet hours and frequency caps apply.
  3. 3
    Action. a short, personal nudge on the channel this member actually answers: a dose reminder that anticipates the side effect they reported last time, a check-in after a missed injection, encouragement pegged to their own trajectory, a rest suggestion after a rough week of sleep data.
  4. 4
    Follow-through. replies write back into the same memory, so the next human session opens on progress instead of from scratch. If the member shows up in the app first, the pending nudge is cancelled. Red-flag replies — concerning symptoms, distress — escalate straight to the care team.
Example
AgentYour dose increase is scheduled this week. Last time you stepped up, you mentioned nausea in the first days — here's the routine that helped then. Want me to check in on Friday to see how it's going?
CustomerYes please. I was actually a bit nervous about this one.
AgentThat's normal, and you've handled it before. Keep meals light early in the week, and if anything feels off, message me here — I'll bring in your care team if we need them.

Configuration

How the agent is wired for this use case.

Triggera program milestone, behavioral signal, or risk signal — a due dose reminder or approaching titration step from the program backend, a missed injection log or weigh-in gap, or a poor-sleep stretch from a connected wearable.
Tools & actions
  • Program backend / EHR · reads dose history, titration schedule, side effects, and stated goal; writes replies back to the member record.
  • Memory layer · reads the cross-channel conversation memory; writes each exchange back so the next session opens on progress.
  • Wearable / logging feed · ingests injection logs, weigh-ins, meals, and sleep data as trigger and context.
  • Messaging channel · sends the personal nudge on the channel the member answers (in-app chat, push, SMS, WhatsApp, email).
  • Care-team escalation path · routes red-flag replies to a clinician.
Autonomyunprompted nudges send unattended only after a quality judge clears them under quiet hours and frequency caps; the agent coaches within the program's protocol and never improvises medical advice — any clinical red flag is a hard human handoff, never an automated reply.
Channelsin-app · push · chat · sms · whatsapp · email
Escalationconcerning symptoms or distress in a reply escalate straight to the care team.

What you need

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

Signals

program schedule events (dose due, titration step), logging events (injection, weigh-in, meals), wearable data where the member has connected one, engagement-decay signals.

Data

member profile and stated goal, dose and program history, the cross-channel conversation memory, consent and quiet-hours preferences per channel.

Guardrails

judge review of every unprompted message; clinical red-flag detection with human escalation; the agent coaches within the program's protocol and never improvises medical advice; per-channel consent and frequency caps.

Metrics it moves

  • adherenceup, because the moments that break a program (a feared titration step, a missed injection) get a timely, personal intervention instead of silence.
  • churndown, the program stays present in the member's week instead of going quiet between sessions.
  • ltvup, longer adherent membership in a model where engagement is the revenue.
  • dau-mauup, nudges open into real coaching conversations rather than dead-end notifications.

Related use cases

Retain / Save

Month-4 cliff companion coach (GLP-1)

the same companion focused on the highest-risk window

Retain / Save

Adherence guardian for medication programs

the risk-signal side of keeping members on protocol

Engage / Lifecycle

Commitment check-in

closing the loop on what the member promised themselves

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