United KingdomPublic sector2021 – ongoingInformation Commissioner's Office
Registration inbox classification and address-change replies
The ICO classified approximately 23,000 emails monthly and automatically directed address-change requests to a self-service form
Problem
The ICO's registration team manually read and answered high volumes of enquiries. Simple address-change requests took time that officers could spend on complex cases.
Approach
ICS.AI and Microsoft developed a Naive Bayes classifier for the registration inbox. It identified address-change requests and automatically sent a form link when confidence reached the specified threshold. The system entered production in May 2021 and did not change registration details itself.
Results
Takeaway
Automation covered routing to a form rather than changing a customer's record. The confidence threshold described when a reply was sent, not the classifier's measured accuracy.
Categories
- Industry
- Public sector
- Business function
- Customer service
- Innovation type
- Process innovation
- AI technology
- Predictive analytics and ML
- Adoption
- Partner co-development
- Evidence
- Company's own claim
Sources
Source summaries
The ICO's transparency record describes a classifier that identifies address-change requests and sends self-service links. It records production use since May 2021 and approximately 23,000 emails classified monthly.
Full details are in the original sources.