Platform · Live Assist
The agent hands off. Your team picks up with everything it needs.
Escalations arrive in your Inbox with the full conversation, the escalation trigger, and the KB passages the agent retrieved. When your team takes over, it can ask for a suggested reply grounded in the same knowledge base, and send only what it chooses to.
Handoffs that give your team a running start.
Full transcript in the handoff
Every escalated conversation arrives in your Inbox with the complete transcript — not a summary, not a ticket description. Your team sees exactly what the customer said and exactly how the agent responded.
Escalation trigger surfaced
The Inbox shows why the agent escalated — low confidence signal, sentiment threshold, explicit escalation intent, or a specific action the agent couldn't take. Your team knows what they're walking into before they say hello.
KB passages included
The knowledge base chunks the agent retrieved are attached to the handoff. Your team can see what the agent was working from — and quickly identify if a wrong escalation came from a KB gap.
In the Inbox: help when your team takes over.
These are built into the dashboard and tested in code. They have not yet been used by a customer on a live conversation, so we will not quote a result for them.
Suggested replies
On an escalated conversation, your team can ask for a drafted reply grounded in the same knowledge base the agent uses, with the passages it drew on shown beside it. The draft never sends itself: your teammate reviews, edits and sends it, or discards it.
Knowledge panel
A panel in the conversation view shows the passages most relevant to the visitor's latest message, so your team doesn't have to go search. It loads when you open the conversation and on refresh; it does not update as you type.
Reply policy nudges
Owners and admins set phrases your team should never use and wording every reply must contain, such as a recording notice. The reply box shows a notice when a draft breaks either. It is advisory: the teammate can still send, it applies to replies your team types and not to the AI agent, and it matches exact phrases, not meaning.
Approvals, not action buttons
Custom actions run through the agent, and destructive ones wait for a teammate to approve or decline them in the dashboard. There is no inbox button that lets a human run an action directly.
Better handoffs from day one.
Clean escalations with full context are in the product today. If you want to see how the handoff looks in practice, we're happy to walk through it with your data.