In fast-moving customer channels, every message gets attention. That’s a good thing when a user reports an issue — but not when the message is “Good morning” or “Thanks, it’s working now" or "Look forward to catching up".

ClearFeed has used AI from the beginning to filter out some of the noise in Slack Channels. Today we are releasing the ability for customers to control this AI themselves - to identify important customer requests based on their own needs. This blog dives into this feature and describes how it works and how to set it up.

‍

The Problem with “Every Message is a Request”

A naïve way to monitor Slack channels with customer queries is to treat everything as a "Request". While this makes sure that every customer message gets aresponse - it leads to fatigue in customer facing teams - particularly as the number of channels increases. When automatically converting Slack messages to Tickets - it also creates a lot of noise for the Support teams and leads to Ticket sprawl.

To solve this - ClearFeed has employed multiple mechanisms to identify important messages from the very initial release.

• A default filter uses AI to ignore messages that belong entirely within the following categories - Greetings, Acknowledgements, Appreciation and Resolved.
• For Customer Support use cases - treating only messages from customers as Requests that need a response (and can potentially create tickets)
• Finally - multiple related messages can be merged into the same Request to reduce the number of Requests (with the option to Split these Requests later)

However, until now, our AI Filtering was not customizable by end users.

‍

Customizable AI Filtering

With our new AI Filtering feature, ClearFeed allows users to categorize messages to ignore - beyond the default set of categories that are pre-configured in ClearFeed. With the customization, users can

• Add new categories of messages they would like to ignore
• Tweak existing ones to better fit their communication patterns

This flexibility lets you decide what makes it into Triage channels and, in combination with automatic ticketing, what gets converted into tickets.

‍

How it works

ClearFeed uses GPT-4.1-mini by default to check if incoming messages fall entirely into one of the categories that is deemed as "Ignorable". The classification is done in real time and suppresses request creation for messages that fall entirely into Categories that should be ignored. The AI Prompt contains the category names and examples. For example - the default prompt has the following categories and examples as indicated below:

• Greetings — “Good morning,” “Happy birthday”
• Acknowledgments — “Got it,” “Ok thanks”
• Appreciation — “Great work,” “Thanks for helping out”
• Resolved — “It’s working now,” “This is solved”

To customize the AI Filter - users can make the copy of the default prompt listed in our Documentation - and add additional categories and examples for them. They can also remove some of these categories. The customized prompt can then be uploaded to ClearFeed (for now users need to contact ClearFeed Support for this).

⚠️ Because ignoring genuine requests is fraught with risk - lot of care needs to be taken to define narrow categories and very specific examples. The system is built in a way that if any part of an incoming message does not fit into one of the "Ignorable" categories - ClearFeed will create a Request out of it. This makes sure that any important customer messages are not dropped.

‍

AI Filtering, Automatic Ticketing & Automated Responses

Many users use ClearFeed to file tickets automatically in external systems like Zendesk, Intercom, FreshDesk, SalesForce, Clickup or Linear. In such cases, by default, ClearFeed does not filter messages and every new message is auto-converted to a ticket. However users have a choice of:

• Enabling AI Filtering to prevent ticket creation for unimportant messages.
• With this release - they can further refine the AI Filtering settings to suit their requirements.

For users who have setup automated Out-Of-Office responses (or Virtual Agent) - AI Filtering also controls messages on which automated responses are posted. For example - users may want to customize the Filter to avoid posting formal OOO responses on informal chat messages.

‍

Conclusion

AI Filtering is a small but powerful mechanism. By reducing noise, your team spends less time on “non-requests” and more time resolving real issues. The result is:

• Cleaner Triage channels
• Fewer unnecessary tickets
• Faster response times for the messages that count
• Better customer experience with automated responses only where appropriate

We look forward to users trying out the powerful new building block that is part of the ClearFeed product now. Please contact our Support team if you have any questions regarding the same.