Use this article to identify conversations that a personal AI agent started in your web Messenger, find them in the Inbox, route them with audiences, and report on them. It's for teammates who manage Fin, the Inbox, or reporting. The feature is available on all plans.
What is personal agent detection?
Personal agent detection is a feature that rates every new web Messenger conversation with a Personal agent involved attribute: Yes, Likely, or No. The rating tells you how likely it is that a personal agent started the conversation instead of a person.
Personal agents are AI assistants that act for a person. They open your website, open the Messenger, and chat with Fin the way a person would.
Why use personal agent detection?
Personal agent detection helps you separate these conversations from your human ones. You can:
See which conversations a personal agent probably started.
Find those conversations in Inbox views and Inbox search.
Use audiences to send likely personal agents to their own procedures.
Review how Fin handles them in the Personal agents report, including resolution and escalation rates.
Note: Nothing changes for your customers. They never see the rating, and the Messenger works as before. Fin only behaves differently if you set up an audience or procedure that uses the rating.
How does Fin rate a conversation?
When a visitor starts a new conversation in the web Messenger, Fin rates it. Fin saves the rating on the conversation in the read-only Personal agent involved attribute.
Rating | When Fin assigns it |
Yes | Fin is confident that a personal agent started the conversation. |
Likely | Fin thinks a personal agent probably started the conversation. |
No | The default rating, when neither of the above applies. |
Fin sets the rating when the conversation is created, before Fin's first reply, so it's ready when audiences are matched.
How to change your personal agent detection setting
The Personal agent detection setting controls if Intercom detects whether a personal agent, rather than a person, is contacting you through the Messenger. Intercom saves the result on each new conversation so Fin, Workflows, and reports can use it. By default, detection is enabled.
Go to Settings > Security > Data.
Scroll to the bottom of the page to the Personal agent detection section.
Open the dropdown and select Enabled or Disabled.
Click Save.
Enabled: Intercom detects personal agents and saves the rating on each new conversation.
Disabled: No detection runs. Ratings already saved on past conversations are kept.
How to use the Personal agent involved rating
Use the rating in the Inbox
Teammates see the Personal agent involved attribute in the sidebar of a conversation. Teammates can't edit it.
You can use the attribute in Inbox views, for example to build a view of Likely and Yes conversations.
You can also use the attribute in Inbox search to find those conversations.
Use the rating in audiences
You can build Fin audiences from the Personal agent involved attribute and apply them to Fin. For example, you can set up a procedure that only runs for Likely or Yes conversations by selecting that audience in the procedure's audience targeting.
Use the rating in reporting
You can filter any report by Personal agent involved, or use it to break down any chart. Reporting also includes a ready-made Personal agents report template, which is filtered to Likely and Yes conversations. It contains:
Fin resolution rate
Fin escalation rate (conversations that Fin routed to your team)
New conversations
Open conversations
New conversations compared with conversations replied to, over time
Note: The Personal agent involved rating is an estimate, not proof. Keep these limitations in mind:
The rating is an estimate. Expect some people rated Likely and some personal agents rated No.
Some personal agents can't be detected yet.
Only conversations started in the web Messenger are rated. Conversations from the iOS and Android SDKs (software development kits), email, and other channels aren't rated. Voice isn't supported yet.
Only new conversations are rated. Fin sets the rating once, when the conversation is created, and it doesn't change later.
Older conversations aren't rated. Only conversations created after November 13th, 2026 have the attribute (there's no backfill).
FAQs
Can customers see the Personal agent involved rating?
Can customers see the Personal agent involved rating?
Customers can't see the Personal agent involved rating. The rating is only visible to your teammates, and the Messenger behaves exactly as before.
Can I edit or clear the Personal agent involved attribute on a conversation?
Can I edit or clear the Personal agent involved attribute on a conversation?
You can't edit or clear the Personal agent involved attribute on a conversation. The attribute is read-only.
Why is the Personal agent involved attribute missing from a conversation?
Why is the Personal agent involved attribute missing from a conversation?
The Personal agent involved attribute is missing from a conversation if it didn't start in the web Messenger, if it was created before November 13th, 2026 when this was enabled, or if Personal agent detection was set to Disabled when the conversation started. Fin doesn't rate conversations from the iOS and Android SDKs, email, or other channels, and it doesn't backfill older conversations.
Why was a person rated Likely or Yes, or a personal agent rated No?
Why was a person rated Likely or Yes, or a personal agent rated No?
A person can be rated Likely or Yes, and a personal agent can be rated No, because the rating is an estimate. It isn't proof that a personal agent did or didn't start the conversation.
How do I turn off personal agent detection?
How do I turn off personal agent detection?
To turn off personal agent detection, go to Settings > Security > Data, scroll to the bottom of the page, set Personal agent detection to Disabled, and click Save. No detection runs while the setting is Disabled, and ratings already saved on past conversations are kept.
Need more help? Get support from our Community Forum
Find answers and get help from Intercom Support and Community Experts