Docs/Human Handoff/Conversation Context

Conversation Context

When an agent claims a conversation, here's exactly what they have in front of them — and what they don't.

The message transcript

The claim screen (the WhatsApp Inbox, or the Email/Instagram/Messenger inbox tab under Chatbot → Deploy) shows the full conversation from the start: every AI reply, every visitor message, and any earlier agent replies, each labeled AI or Agent. On WhatsApp specifically, images render inline and voice notes play back with their transcription shown underneath. Nothing is summarized or truncated — it's the raw history.

Contact details

A sidebar panel shows whatever's been captured so far:

  • Name
  • Phone number
  • Email address
  • Detected language

These come from however the visitor was captured — a lead-capture form on the website widget, or auto-captured from the channel profile (e.g. the WhatsApp phone number) when there's no form to fill in.

Zoho CRM context, if connected

If the conversation's lead has synced to Zoho CRM and the agent has an active Zoho CRM integration, the sidebar adds a CRM panel showing:

  • Module (e.g. Leads, Contacts)
  • Owner
  • Stage
  • Status
  • Notes pulled from Zoho, with author and timestamp
  • A direct "View" link into Zoho CRM (only shown on integrations connected after the org ID started being captured)

This reads fields already cached on the lead record from the last sync — it does not call Zoho live on page load, so it stays fast, but it can be stale until the next sync runs.

What's not there

⚠️
No AI-generated summary. The conversation table has an unused handoff_summary column, but nothing in the codebase writes to it — agents read the transcript itself rather than a condensed version of it.
⚠️
No stated handoff reason, in the Inbox view. On WhatsApp, Messenger and Instagram, the AI's own stated reason and priority for the handoff (e.g. "billing dispute", priority high) are saved onto the auto-created lead's custom fields, but the Inbox conversation screen doesn't surface them — you'd need to open the lead record separately to see them.

Taking action

From the same sidebar, the agent has up to three buttons depending on the conversation's current state:

  • Claim conversation — visible while the conversation is human_requested or waiting_for_agent.
  • Mark resolved — visible once it's human_active.
  • Resume AI — visible once it's resolved, hands the conversation back to the AI.
💡
Because claiming is atomic, it's safe to have multiple teammates watching the same Inbox — only one of them will succeed in claiming a given conversation, and the rest get a clear "already claimed" response instead of overwriting each other.

Was this page helpful?