Barking and Dagenham AI Search and Webchat

This information was sourced from "Barking and Dagenham AI Search and Webchat" by AI.GOV.UK, https://ai.gov.uk/knowledge-hub/tools/barking-and-dagenham-ai-search-and-webchat/?directTo=library. Licensed under Open Government Licence v3.0.

The challenge

Outside of the top interaction types, residents find it cumbersome to interact with the council's website. Furthermore, traditional search functions return an overwhelming number of results, making it difficult for users to find relevant information. This leaves residents feeling frustrated, leading them to abandon the self-service option in favour of calling the contact centre or giving up on the interaction entirely.

The solution

An LLM-powered tool built with Azure AI Search (formerly known as Azure Cognitive Search) that embeds the core, non-ephemeral content on the council's website. It uses Retrieval-Augmented Generation (RAG) to answer plain language queries with concise summaries and direct service links. Users can interact via search or chat interfaces. Telephony and WhatsApp channels are planned next.

The results

• The AI Search interface is handling 18,000 sessions per month. The first-time hit rate on search is high, with low levels of negative feedback.
• The AI Chatbot interface is currently serving 4,450 conversations per month, with a 93% success rate based on user feedback (thumbs up / down) at the end of each conversation.
• The combined impact of the tool across both interfaces has been a reduction in general enquiries and calls into the Council.

Learnings / insights

This approach has created a simpler, more efficient way for residents to access information and be directed to services. It has become a key part of the Council's digital resident engagement strategy. However, the Council still receives tens of thousands of telephone calls each month across its services. The addition of telephony and WhatsApp channels later this year is expected to significantly enhance the experience for residents.