Flagging a bad reply

Report an incorrect answer from the agent in one click — it alerts your team and feeds back into improving the agent.

When the agent writes something incorrect — a wrong price, false information, a tone that does not fit — you can report it in one click. Flagging alerts your team and feeds into improving the agent.

Flag a message

  1. Open the conversation in the inbox.
  2. Find a message written by the agent. It carries an AI Assistant label above it.
  3. Hover over that message without clicking. A small flag appears on the left.
  4. Click the flag. The reply is reported.
Hovering an agent message reveals a flag on the left, circled in red here.
Close-up of the flag — the button to click.

Turn flags into fixes

A flag records that an answer was wrong. Fixing the cause happens in the AI Center:

  • Wrong prices or stock usually mean the catalogue is stale — check the product catalogue.
  • Wrong policy answers, such as delivery or returns, belong in the agent's policies.
  • A tone or wording problem belongs in Agent behaviour, under Always do and Never do.

FAQ and troubleshooting

I cannot see the flag

Check that you are hovering a message from the agent rather than from the customer or from a teammate. Only messages labelled AI Assistant can be flagged, and the flag is hidden until you hover.

Does the customer know I flagged a reply?

No. Flagging is internal. The message stays in the conversation exactly as the customer already saw it.

Should I flag or take over?

Do both when it matters. Flag the message so the cause gets fixed, then turn AI Autopilot off and correct the information with the customer yourself.

What's next?

Some conversations need a human before you even notice. That is escalation.

Escalation

Was this article helpful?

Share this article

Inbox

AI Autopilot

Let the agent answer on its own, and take back a single conversation without stopping it everywhere else.

Inbox

Escalation

What happens when the agent hands a conversation to a human — the yellow note, the silent agent, and how to pick it up.

AI Center

Agent behaviour

Pick a personality, set the tone, choose the languages and script, and set the guardrails with Always do, Never do and escalation rules.