Clientpulse

@nikolay_golos Great practice! Used to do the same. I think beyond a certain scale, it becomes really difficult to keep up. From talking to different teams, I’ve seen that once you hit about 1k tickets per month, manual review becomes too time-consuming, and quantitative approaches become more useful.

@aleksanadr_lavrinenko Awesome to hear that even a simple version made a big impact on support! 🙌
Right now, we’re going deeper into analytics across multiple feedback sources, not just support tickets. I see much greater value in aggregating all customer feedback in one place.

That said, we’re also exploring a few interesting use cases like syncing tags back into helpdesk after clustering to help prioritize responses, and automatically generating tasks based on new topics, emerging bugs, or feature requests. 

However, we’re not focusing on chatbot automation, since there are already many solutions in that space.

@hina_siddiqui Great question!

We define a topic as a unique issue or request, formulated like a task name for development or resolution. If multiple users report the same problem, their feedback will be assigned to the same topic.

At the same time, related topics can be grouped into subcategories—for example, all issues related to notifications can be grouped together to track overall trend. That’s why we built a flexible taxonomy that can be adjusted as needed.

These rules are handled at the prompting level, and as LLMs improve, accuracy gets even better.