Business Intelligence Subject Tagger

This information was sourced from "Business Intelligence Subject Tagger" by AI.GOV.UK, https://ai.gov.uk/knowledge-hub/tools/business-intelligence-subject-tagger/?directTo=library. Licensed under Open Government Licence v3.0.

The challenge

DBT colleagues had to manually read each piece of Business Intelligence in full to understand the subjects being raised by businesses. This manual process could take hours depending on volume, created delays in identifying emerging business issues, and prevented rapid response to urgent topics requiring immediate departmental attention.

The solution

The Business Intelligence Subject Tagger uses a Convolutional Neural Network with ensemble binary classification models to automatically suggest subject tags for business feedback. The system processes over 20,000 interactions through 12 predefined categories including Exports, Investment, Regulation, and Supply Chains. It runs nightly automated tagging using Apache Airflow, providing same-day availability of categorised feedback to enable rapid identification of emerging business trends and hot topics.

The results

The system processes over 20,000 business interactions automatically with tags available the day after entry into the CRM system. Performance varies significantly across categories, with precision ranging from 0.34 to 0.84 and recall from 0.60 to 0.92. The tool eliminated hours of manual reading required for subject identification and enables rapid triage of business feedback, though it has very low impact on direct decision-making and serves primarily as an initial sorting mechanism.