United KingdomPublic sector2026 – ongoingHM Revenue & Customs
Customer feedback classification with the Organisational Listening Tool
HMRC reported 185 distinct users of its machine-learning customer feedback tool
Problem
HMRC service owners needed to identify recurring problems in feedback from digital services and contact channels. Fixed feedback categories could miss emerging issues.
Approach
HMRC built classifiers to assign feedback to known themes and used BERTopic clustering to identify additional topics. A self-hosted language model generated topic labels. Results appeared in an R Shiny dashboard that service owners could filter, with user feedback informing revisions.
Results
Takeaway
Classification supported tracking of known issues, while clustering exposed additional themes. User counts established internal use without measuring improvements in customer experience.
Categories
- Size
- Large enterprise
- Industry
- Public sector
- Business function
- Customer service
- Innovation type
- Process innovation
- AI technology
- Predictive analytics and MLGenerative AI
- Adoption
- Built in-house
- Evidence
- Company's own claim
Sources
Source summaries
HMRC described supervised classification and topic modelling of customer feedback. It reported 185 distinct users in the preceding 30 days and explained how service owners reviewed the results.
Full details are in the original sources.