Product update1 min read
AI agents now hand off to a human with the full conversation
When a bot escalates, the agent picking it up gets the whole thread, the bot's reasoning and a one-line summary of what the customer actually wants, so nobody has to ask them to repeat themselves.
What's changing
- Full transcript and bot reasoning attached to every escalation
- One-line summary of the customer's intent at the top of the thread
- Escalation rules per intent, sentiment or number of failed turns
- Available on every plan that includes AI agents
The worst moment in an automated conversation is the handoff. A customer explains their problem to a bot, the bot gives up, and a human opens the thread and asks them to explain it again. This release removes that moment.
What the agent sees now
An escalated conversation arrives in the inbox with a context panel already filled in: a one-line summary of what the customer is trying to do, the intents the bot matched, the data it collected, and the point at which it decided it could not finish. The transcript is there in full, but the agent rarely needs to read it.
Deciding when to escalate
Escalation is a rule you write rather than a fixed threshold. You can hand off on a matched intent, on detected frustration, after a set number of turns without progress, or on an explicit request from the customer. Rules are per-flow, so a billing query can escalate on the first uncertain turn while a delivery lookup keeps trying.
It is on by default for every account with AI agents included. Existing flows keep working as they are configure escalation rules on a flow to turn it on for that flow.