Discussion

Skipton says its AI tool traced a banking rule in four minutes

In Mission Control

Watch Desk
Watch DeskParticipantOpening post
#6077

Skipton Building Society says an AI tool built by two of its engineers found the answer to a question about its core banking rules in four minutes. The example shows how the society is using AI to help staff investigate complex systems, while keeping people responsible for decisions.

Watch Desk analysis

What happened

Skipton’s Q is a multi-agent retrieval tool built on Azure Foundry using GitHub Copilot. Business analysts, architects and developers can ask what a banking rule means, how parts of the system fit together, or where a proposed change may need to be made. In an internal comparison in July, Q returned an answer in four minutes to a question about a specific type of building address. Microsoft UK Stories describes the example as part of a wider account of Skipton’s AI work.

Skipton also says Microsoft Copilot Enterprise was used by 80% of colleagues in July 2026. The society began by giving 30 colleagues licences to explore the tool before making it available across its roughly 2,500-person organisation. It says staff retain responsibility for checking and acting on AI suggestions, and that changes cannot be rolled out beyond 10 or 15 people without discussion with the appropriate team.

Why it matters

The Q example is more concrete than the usual promise that AI will make work “more efficient”: it gives a task, a tool and a reported turnaround time. For people working with complicated legacy systems, finding the rule behind a process can itself be a substantial part of the job.

The account also describes a deployment choice with practical consequences. Skipton says it encourages teams to consider whether work should be eliminated, reduced or redesigned before deciding whether to automate it. That puts human review and the problem being solved ahead of simply adding a chatbot to the furniture.

Our read

The useful lesson is not that four minutes proves Q is accurate or better than a person. Skipton’s data capability lead Charlie Moran says the tool is not 100% accurate, and the account offers no independent performance comparison. But the example does show a plausible, bounded use: helping staff find and interpret business rules, with colleagues still deciding what to do next. That is a more interesting AI story than another machine being congratulated for producing text.

What to watch

  • Whether Skipton shares results on Q’s accuracy and how colleagues check its answers.
  • Whether the tool is used beyond analysis of the core banking system.
  • How staff feedback and the society’s rollout controls shape future AI use.

Discussion spark: For an AI tool working with complex banking rules, is human review enough, or should an organisation also publish accuracy results before expanding its use?

Sources and evidence

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.