How to capture FIN AI response text in Workflows for SendGrid email forwarding? | Community
Skip to main content
Question

How to capture FIN AI response text in Workflows for SendGrid email forwarding?

  • September 17, 2026
  • 1 reply
  • 42 views

We connected a POST API to SendGrid within our email workflow to capture FIN's response to users and store it as audit evidence in forwarded emails. However, we haven't found any parameter or attribute that exposes FIN's response text.

Is there a way to create a custom attribute to pass this value, or do you recommend an alternative method to capture it? ​@Nathan Sudds 

1 reply

Conor
Super User ✨
Forum|alt.badge.img+6
  • Super User ✨
  • September 18, 2026

Hey ​@Adrian Vargas ! We didn’t get to this workflow on the call yesterday (this isn’t really procedures and data connectors related, so I recommend joining Monday’s office hours as it’s more generally about all of Intercom and Fin), but my speculation based on your description would be that you’re pushing the conversation to SendGrid before Fin responds (that’s a guess, so if that’s wrong, tell me where the workflow is set up). 

If you share a little more about the specifics of your need here, then that could help narrow down the options, but an alternative would be to remove that part of the email workflow entirely, and instead subscribe to the conversation.operator.replied webhook topic ("Fin/Bot replies to a conversation" — topic list). The payload's conversation_part.body is Fin's reply text, and the author object tells you it's Fin:

  • author.from_ai_agent: true — came from the AI Agent
  • author.is_ai_answer: true — this is Fin's final answer, not a workflow bot message

Three things that will bite you on an audit trail specifically:

  • Expect duplicates. Bot replies fire on both conversation.operator.replied and conversation.admin.replied. Intercom's guidance is to treat the operator topic as primary and deduplicate on the conversation part ID.
  • Quick replies arrive separately, as an extra event with part_type: "quick_reply", and need enabling at workspace level.

If you need the whole thread rather than each reply as it happens, call GET /conversations/{id} and use conversation_parts, taking parts where part_type is comment. Two tips for audit evidence: pass display_as=plaintext so you get clean text instead of HTML for the forwarded email, and note the API returns a maximum of 500 parts (the most recent 500 on longer conversations).

For audit purposes I'd also store conversation_part.id and created_at alongside the body (the ID gives you deduplication and the timestamp gives you ordering). ai_agent.resolution_state and content_sources are worth capturing too: they evidence what Fin based the answer on, which could be a question an auditor actually asks.

Conor