⚠️ Update: This is an older version of our AI-powered categorization feature.
We’ve launched a newer and more advanced system called Fin Categories for classifying conversations automatically.
The AI Category Detection beta is no longer available to new customers, but if you’ve already been using it, you can continue to do so.
Going forward, we recommend using Fin Categories, which offers better accuracy, flexible routing, and deeper reporting insights.
How to get access to Fin Categories?
Fin Categories is currently in beta. If you’d like to try it, just reach out to Zoe Sinnott at zoe.sinnott@intercom.io with your workspace ID.
Can I use both AI Category Detection and Fin Categories?
Yes - if you have access to AI Category Detection and Fin Categories, you can use them alongside one another for now. That said, we recommend:
Requesting access and using Fin Categories for all new AI-powered categorization use cases.
Planning to gradually move away from AI Category Detection, as this feature will be sunset in the future.
This helps ensure your setup is ready for the future as we focus improvements on Fin Categories.
When you use Fin with your helpdesk, you have the ability to control which topics Fin responds to and what it routes to your team, capturing all relevant context in the process. Fin can automatically categorize conversations by topic, sentiment, or any other relevant category and you can use this data to ensure that every conversation is given the right level of support.
How it works
The way AI category detection works is by defining a conversation attribute and describing each of the values in natural language. Then you can trigger an action in your Fin workflow to automatically classify this attribute at any point in the conversation.
Once classified, you can use this attribute to create different routing branches, pass on relevant context to your team, and improve your reporting.
Example use case
Suppose you want Fin to recognize when a customer’s sentiment is negative. You can set up a "Sentiment" attribute with values like "Positive," "Neutral," and "Negative." If a conversation is classified as “Negative,” Fin can then trigger an escalation or route the conversation directly to an agent.
By setting up automatic classification and defining conditions, you can ensure that Fin delivers a more contextual and responsive customer experience. You can also monitor the performance of your categories by generating reports and analyzing the distribution of messages across different categories in real time.
Setup
Set up AI category detection with Fin for Zendesk
Set up AI category detection with Fin for Zendesk
Step 1. Define ticket field in Zendesk
First, ensure you have a list-type ticket field in Zendesk that outlines the categories you want Fin to recognize (e.g., "Product Inquiry," "Complaint," "Feedback").
In your Zendesk admin settings, go to Objects and rules > Tickets > Fields, and create or edit a dropdown field.
Step 2: Sync the ticket field with Fin
Go to Settings > Integrations > Zendesk integration and in the Sync data from Zendesk section, click the + icon and select the attribute you want to add. This will sync the attribute to make it available for AI category detection.
Note: If the list-type attribute already exists in Fin but has been updated in Zendesk, you can re-sync it by selecting the attribute and clicking Re-sync. This will update the field in Fin with the latest options from Zendesk.
Step 3: Describe attribute list options
Once your ticket field has been synced, you'll need to add a description for each list option. It's important to be very clear with your details here, as the description is what helps Fin to understand and detect the correct attribute.
If you need to update descriptions, go to Settings > Integrations > Zendesk integration in your Fin workspace and scroll down to the "Sync data from Zendesk" section. Click on the attribute to modify the descriptions below each list option to fit your workflow requirements.
Note: Attribute descriptions must be a minimum of 10 characters.
Set up AI category detection with Fin for Salesforce
Set up AI category detection with Fin for Salesforce
Step 1. Define case field in Salesforce
First, ensure you have a picklist case field in Salesforce that outlines the categories you want Fin to recognize (e.g., "Product Inquiry," "Complaint," "Feedback").
From the Salesforce setup menu:
Go to Object Manager, then select the Case object.
In the Case object management page, click on Fields & Relationships.
Click the New button to begin creating a new field and choose the "Picklist" field type.
Enter the field label and other relevant properties.
Remember to add the new field to the appropriate profiles and page layouts and then save the case type.
Step 2: Sync the case field with Fin
Go to Settings > Integrations > Salesforce integration and in the "Pull data from Salesforce" section, click the + icon under Case fields and select the list-type attribute you want to add. This will sync the case field to make it available for AI category detection.
Step 3: Describe attribute list options
Once your case field has been synced, you'll need to add a description for each list option. It's important to be very clear with your details here, as the description is what helps Fin to understand and detect the correct attribute.
If you need to update descriptions, go to Settings > Integrations > Salesforce integration in your Fin workspace and scroll down to the "Pull data from Salesforce" section. Click on the attribute to modify the descriptions below each list option to fit your workflow requirements.
Note: Attribute list descriptions must be a minimum of 10 characters.
Enable AI category detection in your Fin workflow
Once you have synced your attribute data, you can add AI category detection to your Fin workflows, by selecting Add step and choosing the action AI category detection. This enables automatic categorization based on real-time conversation analysis.
Now use this attribute in your workflow by adding branches with conditions based on the AI category detected or include it in hand-off notes to agents.
Monitoring and reporting on AI category detection
To review how many conversations were generated for each category by the AI category detection feature, follow these steps:
Navigate to Analyze > Custom Report within your Fin workspace.
Add the AI category attributes to your report or chart filters.
View the generated report to analyze the message distribution across categories and assess the efficiency and relevance of AI categorization in your workflows.
FAQs
Why am I seeing an error when saving attribute descriptions, and how does it affect AI category detection?
Why am I seeing an error when saving attribute descriptions, and how does it affect AI category detection?
You may encounter a red error banner when saving attributes that include descriptions. While it might appear that the attributes are saved, this error can lead to downstream issues—such as failures in AI category detection within your workflows.
Cause:
The error typically occurs when the description for one or more attributes does not meet the required minimum of 10 characters.
Resolution:
Ensure that all attribute descriptions are at least 10 characters long. After updating each description to meet this minimum, you should be able to save without errors, and AI category detection should function as expected in your workflows.
If you're still encountering issues after updating your descriptions, try refreshing your workflow and saving again. If problems persist, feel free to contact our support team.
Can I exclude some Salesforce case types from AI category detection?
Can I exclude some Salesforce case types from AI category detection?
Here are some approaches to exclude Salesforce picklist types from AI category detection:
Exclude case types from use - If all case types are in a single picklist and can't be removed, add descriptive labels to the attribute descriptions in your Fin workspace such as "Do not use this case record type" for any types you wish to exclude - Fin can use these descriptions to identify and ignore such items during AI category detection.
Separate picklists by team - To strictly delineate case types for different departments, create separate picklists within Salesforce for each team. While this approach requires longer setup, it ensures greater control over available options used in AI category detection.