All use cases

Prevent / Care

Craving and high-risk-moment intervention

A craving hits at the exact hour it always does, and instead of facing it alone the member gets an urge-surf exercise and a way to reach a person — right when it counts most.

Next best thingGet aheadBehavioral triggerRisk triggerTime-based triggerChatPushIn-appWellness (B2C apps)Health & Telehealth

What it does

In a recovery or sobriety app the dangerous moment is short and specific: a craving spikes, or the member reaches a time of day or a place that has tripped them before. That is when relapse risk peaks and when most apps offer nothing until after the slip. The agent meets the member inside that window. When they say they are struggling, or when a known high-risk pattern lines up, it delivers an immediate, evidence-based coping intervention — an urge-surf exercise, a distraction plan, a prompt to reach out to a sponsor or supporter — and can connect them to a human when the moment needs more than a tool. It is in-the-moment support precisely when it changes the outcome, not a debrief once the chance to help has passed.

How it works

  1. 1
    Trigger. the member signals a craving in chat, or a high-risk pattern lines up — a usual trigger time of day, or a context the member has flagged as risky — raised from the app backend or the member's own message.
  2. 2
    Decision. the agent reads what the member has said and their saved relapse-trigger profile and coping preferences, judges urgency, and chooses the intervention that fits this person and this moment — an urge-surf or breathing exercise, a distraction plan, or a prompt to contact their named supporter.
  3. 3
    Action. it runs the coping intervention in-thread right now, steady and non-judgmental, and offers a one-tap path to a person when the member wants one; every message passes a quality and safety judge before it sends.
  4. 4
    Follow-through. the episode and what helped are written to memory so the next high-risk moment is met better, and if the member's language carries self-harm or crisis signals the agent halts the self-serve flow, surfaces local and national crisis resources, and routes to a human pathway with context, resuming support only after the moment is safe.
Example
AgentIt's around the time you told me cravings usually hit hardest. How are you doing right now — green, amber, or red?
Customeramber, leaning red. really want a drink tonight.
AgentThanks for telling me — that takes strength. Let's ride it out together: a craving usually crests and then passes if you don't act on it. I'll walk you through a short urge-surf now, breath by breath. Want me to also ping your supporter so you're not doing this solo? And if it ever tips past a craving into thoughts of harming yourself, I'll stop and connect you straight to a crisis line and a person.

Configuration

How the agent is wired for this use case.

Triggera craving_reported message or a high_risk_window pattern (a usual trigger time or a member-flagged risky context) raised from the app backend or an inbound chat message.
Tools & actions
  • App backend · detect high-risk time/context patterns and surface them for an in-the-moment touch.
  • Member memory · read the member's relapse-trigger profile, coping preferences, and named supporter; record the episode and what helped.
  • Knowledge base · retrieve the approved coping interventions (urge-surf, breathing, distraction plans) to run in-thread.
  • Messaging channel · deliver the intervention and a one-tap option to reach a supporter or a human.
  • Human and crisis-escalation pathway · connect to a person on request, and on any self-harm or crisis signal halt and surface crisis resources with a warm human hand-off.
Autonomycraving detection and the approved coping interventions run unattended under judge gating, quiet hours, and frequency caps. The agent is a peer-style coping coach, not a clinician — it never diagnoses, never gives medical or medication advice, and never manages a clinical condition; any clinical question is escalated, and a self-harm or crisis signal triggers a mandatory halt to crisis resources and a human under a crisis-safety policy class (detect, then escalate, then close the loop).
Channelschat · push · in-app
Escalationself-harm, crisis, or acute-distress language halts the self-serve flow, surfaces local and national crisis resources, and routes to a human pathway with context; a request to speak to a person connects to the supporter or human-support route; the agent confirms the member is safe before resuming.

What you need

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

Signals

member craving reports, high-risk time/context patterns, app-open and quiet-hour signals, and crisis-language detection on every inbound message.

Data

the member's relapse-trigger profile, coping preferences, named supporter or crisis contacts, channel consent, persistent memory of prior episodes, and the program's approved coping-intervention library and crisis resources.

Guardrails

non-clinical coping support only — no diagnosis, no medication advice; a mandatory crisis halt to helpline resources and a human on any self-harm signal, with loop closure; judge gating on every message; frequency caps and quiet hours so support never tips into pressure; explicit consent and sensitive-data handling for recovery information; visible opt-out.

Metrics it moves

  • adherenceup: support delivered inside the craving window keeps the member engaged with the program through the moments that would otherwise end it.
  • time-to-resolutiondown: a coping intervention runs the instant the risk appears, rather than after a slip has already happened.
  • csatup: being met with steady, judgment-free help at the hardest moment is what members value most in a recovery product.

Related use cases

Prevent / Care

Negative-mood self-help routing

the sub-clinical low-mood sibling that surfaces a coping tool below the crisis line

Prevent / Care

Crisis-language detection and safe handoff

the crisis-threshold safety net this intervention escalates into

Retain / Save

Habit-streak break recovery

the post-break recovery touch that catches a member after a slip, where this card aims to prevent it

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