AI-assisted conversation management
Have replies read, sorted and summarised as they arrive.
The AI agent can read replies as they arrive, work out what the person meant, and either draft a response or send one — while you keep an eye on the whole thing from a board.
Turning it on
Configured per campaign, in the sequence settings.
- Switch on AI response.
- Choose the mode:
- Assist — the agent suggests a reply, you approve it
- Autopilot — the agent replies on its own
The conversation goal
The most important setting. It tells the agent what outcome to drive toward — book a meeting, qualify the lead, send a resource.
Be explicit and action-oriented, include any link it needs, and give it constraints. For example:
- Book a meeting using this calendar link
- Only push for a booking if the lead shows interest
- If they are not interested, do not insist
That last line is what separates an agent that behaves like a person from one that behaves like a bot.
Company and offer context
Tell it what your company does, what your value proposition is, and how to handle the objections you hear most.
This is the section that keeps answers relevant, keeps messaging consistent, and stops the agent inventing things. Treat it as briefing a new colleague on the product.
What it does with that
It works out intent — interest, objection, question — adapts its tone, pushes toward the goal when that is warranted, and backs off or redirects when the person is not a fit. In practice it can answer questions, qualify, handle objections and propose meetings.
Watching it work
Two columns are added to the campaign's pipeline:
- AI conversation — conversations the agent is handling
- To review — the lead replied and it needs your attention
Each card carries an AI insight classifying intent — Interested, Meeting intent, Objection, Later — plus a progress indicator and a status. Enough to see who is worth your own time first.
You are still in it
Even on Autopilot you can read every conversation and step in at any point. The division that works is the obvious one: the agent absorbs volume, you take the edge cases and the decisions.