All use cases

Collect / Recover

Self-serve subscription cancel and refund in chat

A user asks to cancel or get a refund, and instead of a chatbot that can only say "I'll pass you to support," the agent verifies who they are and does it — in the same conversation, with one chance to stay first.

Close the loopBehavioral triggerSystem triggerChatIn-appWellness (B2C apps)Subscription

What it does

For a subscription wellness app, cancellations and refunds are the single biggest support driver, and the in-app assistant usually can't touch them — every one routes to a human queue, so the user waits and the team burns hours on a repetitive, low-judgment task. The agent closes that gap. With verified identity and connected billing, it executes the cancel or the refund directly in chat, and because it is already in the moment of intent it can offer the one reason-matched save before it confirms — a pause for the user who's traveling, a downgrade for the one who finds it pricey. A respectful, immediate resolution replaces the ticket, and the biggest queue in the business becomes a retention surface instead of pure cost.

How it works

  1. 1
    Trigger. a cancel or refund request in chat or in-app — a "cancel my subscription" message, a refund ask, or a cancel event arriving from the app.
  2. 2
    Decision. the agent verifies identity through the app's secure check, reads the subscription and billing state, and decides whether to offer a single reason-matched save first or move straight to the action the user asked for.
  3. 3
    Action. on the user's choice it executes against the billing system — cancels the plan, issues the eligible refund, or applies the agreed save (pause or downgrade) — and confirms the outcome in the same conversation, with no second ticket and no queue.
  4. 4
    Follow-through. the action, the stated reason, and the outcome write back to memory for win-back and product; a refund outside policy or an ambiguous case hands off to a human with full context; the save offer is bounded so cancellation always stays one clear step away.
Example
CustomerI want to cancel my subscription and get a refund for this month.
AgentI can do both right here once I confirm it's you — I'll send a one-time code to your email; nothing to re-enter. While I have you: if it's a quiet stretch, I can pause the plan instead so you keep your history and aren't charged. Or I'll go ahead and cancel — your call.
Customercode's 5-5-2-8-1-9. just cancel and refund please.
AgentDone. Your subscription is cancelled, and this month's charge is on its way back to your original payment method. You'll keep access until the period ends, and your data stays unless you ask me to remove it.

Configuration

How the agent is wired for this use case.

Triggera cancel or refund request in chat or in-app, or a cancellation event from the app backend mid-conversation.
Tools & actions
  • Identity / auth system · run the secure in-chat verification before any account or money action; credentials are never typed to the agent
  • Billing platform · read subscription and billing state; cancel the plan, issue an eligible refund, or apply a pause or downgrade
  • Messaging channel · run the conversation, offer the reason-matched save, and confirm the outcome inline
  • CRM · read plan and contact state; log the request, the stated reason, and the resolution
  • Memory store · read past offers and tenure; write back the reason and outcome for win-back
  • Knowledge base · ground the refund-eligibility and cancellation policy the agent acts within
Autonomyidentity verification and the reason-matched save run unattended under judge gating; the cancel and the refund are money-moving and execute within a bounded refund-and-cancellation policy on verified identity only; a refund outside policy, a dispute, or an ambiguous request escalates rather than auto-resolving, and cancellation always stays one step away — no dark-pattern loop.
Channelschat · in-app
Escalationa refund beyond policy, a chargeback or billing dispute, or a request the agent can't resolve within policy hands off to a human with the full transcript and account context.

What you need

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

Signals

cancel and refund requests in conversation, cancellation events from the app backend, subscription and renewal state

Data

verified identity from the auth system, subscription and billing history, refund-eligibility policy as checkable rules, customer memory of tenure and prior offers, consent state

Guardrails

secure identity verification before any money or account action; refund and cancellation policy as the source of truth, not the model's judgment; judge gating on the conversation and the save offer; cancellation always one step away — no retention dark patterns; human hand-off for out-of-policy refunds and disputes; the agent never takes a new payment, only the user-requested cancel or refund

Metrics it moves

  • ticket-deflectionup, by resolving the top support queue in chat instead of routing every cancel and refund to a human
  • save-rateup, because the agent can offer one reason-matched alternative at the exact moment of cancel intent before it acts
  • time-to-resolutiondown, as the cancel or refund completes in one conversation rather than a ticket and a wait

Related use cases

Retain / Save

Cancellation-intercept save flow

the pre-confirm reason-matched save; this card executes the cancel or refund the user then chooses

Collect / Recover

Data-request self-service for sensitive PII

the sibling action-capable flow behind the same secure identity gate

Prevent / Care

Returns and refunds, judge-gated end-to-end

the judge-gated refund-execution pattern this flow follows

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