Detect when personal AI agents start a Messenger conversation

Intercom added a Personal agent involved attribute that flags new web Messenger conversations likely started by a personal AI agent, meaning an assistant chatting with Fin on someone's behalf.

Fin rates each new conversation Yes, Likely, or No when it's created, before Fin's first reply. Teammates see the rating in the conversation sidebar but can't edit it. You can build Inbox views and searches around Likely and Yes conversations, or use the attribute in Fin audiences so a procedure only runs for those conversations.

Reports can be filtered or broken down by the attribute. There's also a ready-made Personal agents report, filtered to Likely and Yes conversations, showing Fin resolution rate, Fin escalation rate, new and open conversations, and new conversations compared with replies over time.

It's available on all plans, and detection is on by default. You can turn it off under Settings, Security, Data. Customers never see the rating, and Fin answers the same way unless you set up an audience or procedure that uses it.

The rating is an estimate, so some people will be rated Likely and some agents will be rated No. Only new web Messenger conversations created after November 13, 2026 are rated. iOS and Android SDK, email, and other channels aren't rated, voice isn't supported yet, and older conversations aren't backfilled.