Skip to main content

Use AI-powered suggestions to improve Fin

Get AI-powered recommendations to improve Fin's content and performance.

Updated over a week ago

Suggestions is an AI-powered feature that recommends specific actions to help teammates improve Fin performance. It identifies gaps in knowledge, unclear responses, and proposes updates to ensure Fin delivers better answers—faster.

  • Know what to fix and how – Suggestions highlight where Fin struggled and recommend clear, specific content updates.

  • Skip the manual QA – Suggestions scan unresolved Fin conversations, compare them to human replies, and surface what to fix—no transcript digging needed.

  • Fix what matters most – Each suggestion is ranked by impact so you can prioritize the fixes that improve the most conversations.

  • Stay in control – Edit, accept, or reject any suggestion before it goes live—so changes happen on your terms.

Note: We’re working on additional suggestion types like Fin Tasks and Guidance, but these aren’t available yet.

How to access Suggestions

Through the Optimize dashboard

For those needing extra context when reviewing suggestions, go to Analyze > Optimize to identify high-impact actions you can take for topics driving volume, handling time, or poor CX.

Access these Suggestions in the AI Topics table under the "Suggestions" column.

The overlay displays:

  • Number of edits required

  • Creation date

  • Source conversations used to generate the suggestion

Through the Suggestions page

For teammates who manage content, go to Train > Suggestions for a focused to-do list for optimizing content.

The Suggestions page provides:

  • A prioritized view of suggestions based on impact to Fin’s performance

  • Filter options by AI topic and content type

  • Sort options by:

    • Number of source conversations

    • Creation date

How to use Suggestions

Suggestions are generated by analyzing:

  • Failed Fin responses (e.g. escalations or poor-quality replies) and comparing them to successful human replies to similar questions.

  • Teammate-handled responses to check whether there are gaps in your knowledge base.

Suggestions identify the likely root cause and recommend one or more actions.

Types of suggestions

Action

Goals

Availability

Add new content

  • Fill content gaps

Articles

Snippets

Edit existing content

  • Fill content gaps

  • Ensure content is up to date and relevant

  • Improve content quality

Articles

Snippets

Remove or merge content

  • Remove redundant content

  • Remove/merge duplicate content

  • Improve underperforming content

Articles (coming soon)

Snippets (coming soon)

Note: Suggestions only work with snippets created in your Fin workspace. Synced or read-only content is not currently supported.

Reviewing suggestions

You can review all suggestions before enabling them for Fin. Each suggestion includes:

  • A summary explanation

  • Creation date

  • Source conversations

  • Review actions required

Click the conversation icon in the suggestion card to view the source conversations. This helps you understand why and how the suggestion was generated.

Review options:

  • New content: Accept or reject the snippet

  • Edits: Scroll through multiple changes including:

    • Red text (suggested removals)

    • Green text (suggested additions)

After accepting a suggestion, snippets are immediately added to Fin's available content.

Tip: You can edit content directly before accepting or rejecting a suggestion.

Coming soon

  • Suggestions for duplicate content (merge/removal).

  • Suggestions for underperforming/redundant content (merge/removal/update).

  • Suggestions for Fin Tasks and Guidance.

FAQs

How often are suggestions created?

Suggestions are triggered daily or weekly, based on:

  • Volume: High number of conversations where a question and answer can be found.

  • Topic activity: Regular queries (1+ a day) on the same topic for at least 7 days.

  • Spikes: Rapid increases in related queries over 4 days.

What’s filtered out when generating suggestions?

  • Conversations without teammate responses

  • Abandoned conversations

  • Conversations where a teammate repeated the same answer as Fin

  • Conversations that mainly focus on a feature request or bug reporting

  • Existing content available to Fin (including from your external sources)

What’s the difference between suggestions in Train vs Analyze?

They are the same suggestions, but:

  • Train: Gives a prioritized task list for knowledge managers.

  • Analyze: More data-driven, ideal for support leaders and teams.

Are there any limitations for AI-powered suggestions?

  • Suggestions are only generated for conversations that have an AI topic assigned.

  • No option to fast-track or manually flag individual conversations for suggestions.

  • Low-volume customers (with fewer conversations) may receive fewer or no suggestions.

  • Suggestions are only available for snippets (not from external sources).

  • Suggestions are currently limited to English-language content only.

Did this answer your question?