Caddy

Supporting customer service agents with fast, precise information

The challenge

In 2022-23 alone, Citizen Advice with its national organisation and network of 250 independent local charities across England and Wales, helped over 2.5 million people with one-to-one advice. Highly trained advisors have to tackle a broad range of topics, with people often seeking help on complex interconnected issues. Thus the need to make it easier for the advisors to quickly find and share information with the people seeking help from the charity.

The solution

Developed in collaboration with Citizens Advice Stockport, Oldham, Rochdale & Trafford (CASORT), Caddy is a human-in-the-loop copilot for front-line contact centres. Caddy is designed to support advisors by providing quick, accurate, and validated information. It accesses public and internal training and guidance, utilizing semantic routing for topic-specific consideration. Generated content is always reviewed and validated by advisors and supervisors before being presented. Caddy is accessed through existing chat platforms, minimising training needs for advisors.

The results

  • Initial results from a randomized control trial showed 80% of Caddy's responses were of high enough quality to be passed directly to advisors without revisions.
  • Messages were generated, validated, and returned in approximately 4 minutes, significantly faster than human research and typing.
  • Advisors with access to Caddy were twice as likely to report confidence in giving advice and 1.5 times more likely to resolve client issues.
  • Rolled out to 6 local Citizens Advice offices with plans for further expansion and government pilots.

Learnings / insights

  • Building in "humans-in-the-loop" mitigates AI hallucination and risk.
  • Focusing on ease of use and integration with existing workflows is important for adoption.
  • Measuring impact through randomized control trials provides valuable insights for improvement.

Details

Organisation name
Department for Science, Innovation and Technology (DSIT) / Incubator for Artificial Intelligence (i.AI) / Citizens Advice Stockport, Oldham, Rochdale & Trafford (CASORT)
Government body
UK Government
User group
Wider public sector
Use case type
Reusable
Type of technology
Generative AI
Phase
Beta
Impact
Improved efficiency

Links

Get in touch

Email ai-knowledge-hub@dsit.gov.uk to:

  • find out how to use or collaborate on this tool
  • talk to us about your own tool

Content created: 9 May 2025 | Last updated: 23 May 2025