Recommendations reviews the conversations Fin can’t answer and sends to your team, then provides weekly recommendations to fix gaps in Fin’s content, data, and actions.
By analyzing real customer conversations, it highlights the highest-impact opportunities to improve automation and increase Fin’s resolution rate faster.
Key benefits
See what’s blocking outcomes by identifying content gaps, customer data gaps, and action gaps.
Prioritize the highest impact fixes using consistent, conversation-based impact scores.
Work from real customer conversations that directly informed each recommendation.
Move from escalation to automation with clearer guidance on when to create content, data connectors, or procedures.
Note: To access Recommendations and other AI-driven insights, you’ll need the Pro add-on.
How Recommendations works
Recommendations brings all AI-powered improvement recommendations into a single experience, making it easier to understand, filter, and act on recommendations.
To get started, go to Fin AI Agent > Analyze > Recommendations. This provides you with a list of recommendations grouped by the type of gap that prevented Fin from resolving a conversation.
Content gaps
Content gap recommendations highlight where Fin couldn’t answer because help content was missing, unclear, duplicated, or contradictory.
For each content gap recommendation, you can:
Review AI-generated recommendations to create or edit existing content.
See the exact conversations that triggered the recommendation.
Update public articles or snippets directly to improve future outcomes.
Tip: Learn more about using AI-powered content recommendations to improve Fin.
Customer data gaps
Customer data gap recommendations appear when Fin needed information from an external system that wasn’t available, such as order status or account details.
Each customer data gap recommendation clearly shows:
A guide which outlines the API and data needed to build connectors so Fin can resolve conversations automatically. The examples provided should be reviewed and adapted by an engineer.
How to create the automation using a data connector.
The customer queries that would get answered with this customer data.
The implementation effort, which is based on technical complexity, infrastructure and dependency requirements, security and compliance considerations, data operations complexity, business logic requirements, performance and scalability needs, testing and validation complexity, and risk assessment.
Sample API documentation (for reference only). This sample schema illustrates the type of request/response structure you might implement. Your actual endpoints, parameters, and auth flows may differ, so treat this as a blueprint for the integration patterns required to resolve these queries.
An escalation guidance recommendation you could implement quickly as a temporary way to route complex queries to your team and maintain a smooth customer experience, while the full automation is being built.
Tip: Learn more about Fin's escalation guidance and rules.
Action gaps
Action gap recommendations identify where Fin needed to take an action in another system, like updating a workflow or canceling an order.
Each action gap recommendation clearly shows:
A guide which outlines the API and data needed to build connectors or tasks so Fin can resolve conversations automatically. The examples provided should be reviewed and adapted by an engineer.
How to create the automation using data connectors and Fin Tasks or Procedures, depending on your setup.
The customer queries that would get answered with this action.
The implementation effort, which is based on technical complexity, infrastructure and dependency requirements, security and compliance considerations, data operations complexity, business logic requirements, performance and scalability needs, testing and validation complexity, and risk assessment.
Sample API Documentation (for reference only). This sample schema illustrates the type of request/response structure you might implement. Your actual endpoints, parameters, and auth flows may differ, so treat this as a blueprint for the integration patterns required to resolve these queries.
An escalation guidance recommendation you could implement quickly as a temporary way to route complex queries to your team and maintain a smooth customer experience, while the full automation is being built.
Analyzing the data
How impact is calculated
Recommendations use a consistent impact model across all recommendation types where impact is based on:
The number of related conversations.
The time period those conversations cover.
This means impact reflects real, historical demand, so it’s easier to judge which fixes are worth prioritizing.
Note: By default, recommendations are sorted from highest to lowest impact, with the option to sort by date (newest to oldest) when reviewing longer time ranges.
View conversations that informed recommendations
If you open the side drawer in the top right, you’ll find the conversations that directly informed a recommendation.
This makes it easier to:
Understand the exact customer questions behind a recommendation.
Validate whether the fix will meaningfully improve outcomes.
Share concrete examples with teammates when planning changes.
Filter and sort recommendations
Recommendations includes powerful filters so different teams can focus on what matters most.
You can filter by:
Reason: to select a recommendation type (content, customer data, action).
Date range: to review recommendations week by week.
Topic: using AI-generated topics like Billing or Pricing.
Impact: to focus on the biggest wins first.
Accept or reject a recommendation
Once you click Accept, Mark as done, or Reject a recommendation, it’s removed from the list and won’t reappear again.
Recommendations settings
Click on the settings icon at the top of the recommendations page to segment content recommendations by Fin audiences you’ve set up.
When you segment content recommendations by audience, each selected audience receives its own set of tailored recommendations. This ensures that recommendations stay relevant and specific to each audience or brand. If you select “Everyone,” the recommendations will be more general rather than audience-specific.
Note:
Segmentation currently only works for content recommendations.
When you save, your current content recommendations will be cleared. New segmented content recommendations will be generated, which may take a few hours.
FAQs
What’s changed from the previous recommendations dashboard?
What’s changed from the previous recommendations dashboard?
Recommendations Fin has been simplified and refocused to make improvements easier to act on.
Key changes include:
A single recommendations experience, replacing separate Train and Analyze views.
Clearer automation guidance, instead of prioritizing escalation guidance.
Removal of “Investigation needed” recommendations, which were vague and hard to act on.
Exclusive focus on resolution rate, with other metrics planned for future iterations.
What conversations are included in recommendations?
What conversations are included in recommendations?
Recommendations analyzes only meaningful support conversations:
Inbound conversations
Written by a customer
With at least two responses from Fin or a teammate
It excludes pending conversations and automated workflow messages.
How often are recommendations generated?
How often are recommendations generated?
Content gap recommendations are generated weekly, with additional triggers for high volume, sustained activity, or sudden spikes.
Customer data and action gap recommendations are generated weekly.
recommendations are reviewed and worked on week by week using date range filters.
What happened to the Unresolved Questions report?
What happened to the Unresolved Questions report?
The Unresolved Questions report has been replaced by recommendations. Recommendations not only shows where Fin couldn’t resolve conversations, but also provides clear, actionable guidance on how to fix those issues and improve resolution rate.
Need more help? Get support from our Community Forum
Find answers and get help from Intercom Support and Community Experts