All use cases

Engage / Lifecycle

Personalized conversation starters

The customer opens the chat and it already knows them — "your order arrives Thursday, and the size you asked about is back in stock" — instead of a blank box and a "How can I help?".

Next best thingBehavioral triggerSystem triggerChatIn-appE-commerce & RetailHealth & TelehealthSubscriptionFintechTravelWellness (B2C apps)

What it does

The blank chat greeting wastes the most valuable second of the conversation. A customer who opens support or the in-app assistant almost always arrives with live context the business already has — an order in transit, a question they asked last week, a problem the system can see forming — yet the thread starts cold and makes them re-explain it from scratch. This opens every conversation pre-loaded with that context: a short, specific starter built from the customer's own recent activity and memory, so the experience feels like a business that remembers them from message one. It answers the obvious question before it is asked and turns a dead-end greeting into a reason to engage.

How it works

  1. 1
    Trigger. the customer opens the chat or in-app assistant; behavioral and system context is read at that moment — a recent order, a saved item back in stock, an open issue, a step left unfinished.
  2. 2
    Decision. the agent assembles the one or two most relevant facts from live state and persistent memory, decides which is worth leading with, and drops anything stale or low-value — a generic "welcome back" is a valid fallback when there is nothing specific worth saying. The opener clears a quality judge before it shows.
  3. 3
    Action. the chat renders a personalized starter — the order ETA, the back-in-stock answer, the next step — with a one-tap way to act, instead of an empty input box.
  4. 4
    Follow-through. if the context goes stale before the customer returns (the order arrives, the issue resolves), the starter is recomputed or dropped rather than shown wrong; opens, taps, and replies write back, and lift is measured against a plain-greeting control.
Example
AgentWelcome back. Two things you might want: your order is on track to arrive Thursday, and the jacket you asked about last week is back in your size. Anything I can help with on either?
CustomerOh nice — can you hold the jacket while I decide?
AgentI can keep it in your basket and flag it if stock gets low. Want me to do that, and ping you Thursday when the order's out for delivery?

Configuration

How the agent is wired for this use case.

Triggera chat_opened event from the messaging surface, which prompts a context read across the customer's live state and memory.
Tools & actions
  • Customer 360 / CRM · read recent orders, open cases, lifecycle stage, and channel consent
  • Order / fulfilment system · read delivery status and ETA for an in-transit order
  • Catalog / inventory service · check whether a saved or asked-about item is back in stock or has changed
  • Customer memory · recall the customer's recent questions and unfinished steps to ground the opener
  • Messaging channel · render the personalized starter with one-tap actions in the chat or in-app surface
Autonomythe starter is composed and shown unattended under judge gating on relevance, freshness, and tone; any action the customer then takes (place a hold, change an order, move money) follows that action's own policy class and confirmation, in the client's secure flow.
Channelschat · in-app
Escalationif the live context signals a problem outside self-serve scope — a payment dispute, a complaint, a sensitive account issue — the opener routes the customer to the right human path rather than guessing.

What you need

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

Signals

a chat-open event and the live state to read at that moment — order status, stock changes on watched items, open cases, unfinished steps.

Data

recent order and account history, saved/asked-about items, lifecycle stage, channel consent, and persistent memory of recent questions.

Guardrails

judge gating on every opener for relevance and freshness; recompute or drop a starter when its context goes stale so it is never shown wrong; a graceful generic fallback when nothing specific is worth saying; the client keeps owning the chat surface.

Metrics it moves

  • dau-mauup, because a chat that opens with something worth tapping pulls customers back into the conversation
  • click-through-rateup on the starter versus a blank greeting, since it leads with a relevant, actionable fact
  • contact-ratedown on the obvious questions the opener answers before they are asked
  • csatup, from the felt experience of a business that already knows the context

Related use cases

Engage / Lifecycle

Proactive bubbles and personalized push

the event-triggered in-app and push sibling, versus this on-open opener

Engage / Lifecycle

Proactive health-coach companion

the ongoing memory-led companion this opener feeds into

Convert

In-session hesitation rescue

the live in-session intervention once the conversation is underway

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