All use cases

Onboard / Activate

Co-browsing guided takeover

Some setups can't be talked through — the customer is lost three menus deep, and the fastest way out is to ask "mind if I show you?" and do the first one together.

Next best thingBehavioral triggerIn-appChatE-commerce & RetailSubscription

What it does

There is a kind of stuck that no amount of explaining fixes. The customer is mid-setup — building a first filter, configuring a workspace, finding a buried setting — and every "tap the gear icon, then…" makes it worse because they cannot see what you see. Walls of instructions are where onboarding and complex setup quietly fail, and the customer either gives up or opens a ticket. With the customer's explicit approval, the agent takes the wheel of their own session and demonstrates the step live, narrating as it goes, then hands control straight back so the customer does the next one themselves. Showing beats telling at exactly the moment telling has stopped working.

How it works

  1. 1
    Trigger. an in-session signal that the customer is stuck on a multi-step setup — repeated failed attempts at the same screen, a help question that maps to a known hard flow, an explicit "can you just do it for me?".
  2. 2
    Decision. the agent confirms this flow is on the safe, approved co-browsing list (read-only or low-risk configuration, never anything money-moving), checks what the customer has already tried from memory, and asks for explicit consent to take over before touching anything.
  3. 3
    Action. with consent granted, the agent drives the customer's own session — moving to the right screen, completing the first instance of the step while narrating each move in chat — then returns control and invites the customer to repeat it.
  4. 4
    Follow-through. the takeover ends the moment the demo is done or the customer revokes consent; what the agent showed is written to memory so the next session skips the part they now know; the session is logged for audit.
Example
CustomerI've been trying to set up my first filter for ten minutes and I keep ending up somewhere else.
AgentEasiest if I show you — may I take over your screen for a moment? You'll see exactly what I tap, and I'll hand it straight back.
CustomerYes, go ahead.
AgentWatch the top-right — opening Filters, choosing the field, setting the rule, and saving. Done. That's one. Your turn: try adding a second the same way, and tell me if anything looks different.

Configuration

How the agent is wired for this use case.

Triggera setup_stuck in-session signal (repeated failed attempts on a known multi-step flow, or an explicit help request) raised in the chat or in-app widget.
Tools & actions
  • In-app co-browsing / session-control surface · request consent, drive the customer's own session within the approved flow, and release control
  • Product knowledge base · load the canonical steps for the flow being demonstrated and narrate each action
  • Customer memory · read what the customer has already tried; write back which flows were demonstrated
  • Messaging channel · run the chat narration alongside the takeover and confirm hand-back
Autonomythe takeover is consent-gated end to end — the agent acts only after explicit approval, only inside the pre-approved flow allow-list, and returns control on completion or on revocation; no money-moving, account-deletion, or irreversible action is ever on the co-browsing list, and those stay confirmation-gated in the product's own flow.
Channelsin-app · chat
Escalationconsent withheld or revoked, a flow outside the approved list, a payment or account-sensitive action, or a demo that fails twice hands off to a human with the session context.

What you need

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

Signals

in-session stuck indicators (repeated failed attempts on a hard flow, explicit help requests) tied to the screens where co-browsing is allowed.

Data

the allow-list of flows safe to demonstrate, the canonical steps for each, the customer's progress and prior attempts from memory, consent state.

Guardrails

explicit per-session consent before any takeover and instant hand-back on revocation; a strict allow-list that excludes every money-moving, destructive, or account-sensitive action; full session logging for audit; the customer keeps the ability to stop at any moment.

Metrics it moves

  • activation-rateup, by unblocking the multi-step setups where customers otherwise abandon onboarding
  • time-to-valuedown, the hardest flow is shown once instead of explained five times
  • ticket-deflectionup, a live demonstration resolves the "I can't find it" contact in the conversation
  • csatup, because being shown is less frustrating than being told when you're already lost

Related use cases

Onboard / Activate

In-app activation companion

the deep-link-and-nudge onboarding sibling, where the agent guides rather than operates the UI

Upsell & Cross-sell

In-app proactive merchandising block

another in-product surface the agent acts on directly

Prevent / Care

Operator copilot — human-agent assist

the assisted-resolution sibling for when a human takes the harder flow

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