History and assignment for Fin recommendations | Community
Skip to main content
Submitted

History and assignment for Fin recommendations

Related products:Knowledge & Help CenterFin & AI
  • Oct 20, 2026 7:15 AM UTC
  • 5 replies
  • 29 views

Sam Wilson
Forum|alt.badge.img

Problem:

Once a content recommendation is accepted or rejected, there's no history to show what was suggested, what was decided, by whom, and when. We also can't assign a recommendation to a teammate to work on.

Why it matters:
We need to show which changes were made to Fin's knowledge and why, and to tell whether a rejected suggestion keeps coming back. Without assignment, recommendations rely on one person checking them, which doesn't scale across a support team.

What we'd like:

  • A history view of every recommendation, with its outcome, who decided, when, and a link to the resulting change.
  • The ability to assign a recommendation to a teammate, with a status, the way a ticket works.

5 replies

Forum|alt.badge.img+3
  • Active User
  • Oct 26, 2026 6:13 PM UTC

The audit history and assignment part makes a lot of sense, especially when several people are maintaining the knowledge base that an AI agent relies on.

 

I'm curious: when Fin gives an incorrect or outdated answer to a customer, how does your team trace it back to the underlying knowledge or recommendation and make sure the issue doesn't happen again? Is that handled through manual conversation reviews today?


Sam Wilson
Forum|alt.badge.img
  • Author
  • Connector
  • Oct 27, 2026 2:03 PM UTC

Thanks Pravesh, good to hear this resonates.

Honestly, we don't trace it today, which is a big part of why I asked. Someone either accepts a recommendation or sees a problem with it and leaves it out. There's no follow-up and no record of why, so it's just gone. We do review conversations, partly through our own review system and partly by hand, and that catches some wrong answers and some gaps Fin doesn't flag. But it's hard to manage at volume, and it isn't connected to the recommendations at all. A history and assignment view would join those up.


Forum|alt.badge.img+3
  • Active User
  • Oct 27, 2026 2:22 PM UTC

Thanks, Sam that helps me understand the problem much better. It sounds like reviewing conversations and tracking the knowledge decisions behind them are currently disconnected, making it difficult to learn from mistakes consistently at scale.

 

I'm building ChatShield to flag potentially risky support conversations for human review, including issues in both human agent and AI replies. It doesn't currently replace the recommendation history workflow you're describing, but your explanation is really useful for understanding where conversation monitoring fits into the bigger picture.

 

When your team finds an incorrect answer during a review, how do you currently make sure the underlying knowledge gap gets fixed so Fin doesn't repeat the same mistake? Is that process still mostly manual?


Sam Wilson
Forum|alt.badge.img
  • Author
  • Connector
  • Oct 27, 2026 3:02 PM UTC

Yes, mostly manual until recently. It was spread across the team and done when someone had time. I'm now on a short project to improve Fin, which includes our own review of every conversation to find wrong answers and knowledge gaps. I'd still like Fin's recommendations working alongside that, which is why history and assignment matter to us.


Forum|alt.badge.img+3
  • Active User
  • Oct 27, 2026 3:12 PM UTC

That makes sense, Sam. Reviewing every conversation should help uncover wrong answers and knowledge gaps, while history and assignment would make it easier to ensure those issues are actually addressed.

 

I'm building ChatShield, which flags potentially risky conversations across Intercom, including replies from human agents and AI, for human review. It isn't designed to replace your recommendation history workflow, but I'm curious: does your review process also flag things like unsupported promises, policy violations, and privacy risks, or is it primarily focused on factual inaccuracies and knowledge gaps?

 

I'd be interested to hear your perspective, especially since you're working on reviewing every conversation.