Onboard / Activate
In-app activation companion
A newcomer's first weeks, guided by one agent that knows the product, notices the stall, and suggests the right next step at the right moment.
What it does
Most new users don't churn because the product failed them; they churn because they never reached the moment it clicked. The companion lives inside the app through the first weeks: it answers plan and feature questions in chat, deep-links the user to the setup steps that matter (connect the health tracker, book the first coach session), and when behavioral signals show a stall — a key feature never opened, onboarding gone quiet by day 14 — it nudges with one specific, supportive next action. Product guidance and support run as a single agent, not two bolted-together bots.
How it works
- 1Trigger. behavioral and time signals from the product —
feature_unused, an onboarding step left idle, activity gone quiet past a stall threshold. - 2Decision. the agent weighs the user's progress against the activation path, recalls what this user has already done and asked, and picks one next action instead of a generic tour. Every unprompted nudge passes a quality judge before it is shown or sent.
- 3Action. an in-app message or push that deep-links to the exact step — "you logged breakfast yesterday, let's pick today's action" — and any question it raises is answered in the same conversation.
- 4Follow-through. if the user completes the step on their own, the queued nudge is cancelled before it sends; activation events write back to the client's analytics; lift is measured against a holdout group, not assumed.
Configuration
How the agent is wired for this use case.
feature_unused, an onboarding step left idle, activity gone quiet past a stall threshold.- Messaging channel · deep-link an in-app message or push to the exact next step and answer any question it raises in the same conversation.
- Product analytics · read product events and stall timers in, and write activation events back out.
- Knowledge base · weigh the user's progress against the activation path and the plan/feature catalog to pick one next action.
- Scheduling system · queue a nudge and cancel it if the user completes the step on their own before it sends.
What you need
The inputs this use case runs on. Your channels stay yours; the agent supplies the judgment.
Signals
product events (onboarding_step_completed, feature_unused, session activity) plus stall timers on the activation path.
Data
the activation path itself (which steps predict retained users), the plan and feature catalog, the user's progress so far, notification preferences and consent.
Guardrails
frequency caps so the companion never feels like a campaign engine, judge review of every unprompted message, quiet hours, and holdout measurement so claimed lift is proven.
Metrics it moves
- activation-rateup, by walking each newcomer to first value instead of hoping they find it.
- feature-adoptionup, because deep-linked nudges put the unused feature in front of the right user at the right moment.
- time-to-valuedown, the companion compresses the wandering phase of week one.
- churndown, users who activate early are the users who stay.
Related use cases
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