Defra has launched Ask the Toolkit, a public-facing AI assistant that answers questions from the department’s AI guidance and links users to the pages it used. Its most useful feature may be knowing when a search, formula or script would do the job better than AI.
Watch Desk analysis
What happened
The assistant is part of Defra’s AI digital toolkit, which helps people consider how to use AI at work. UKAuthority reports that it sits alongside a separate tool, “Check if AI is right for your idea”, and can explain how Defra’s AI Capability and Enablement team assesses proposals.
The report gives practical examples from the toolkit’s pilots, including checking expenses claims against policy. It says the assistant identifies poor fits too: tasks that need one right answer every time, or depend on data that is difficult to access.
Its safeguards are aimed at a particularly awkward failure mode: confidently getting official policy wrong. Defra software developer Farhana Abdul Alim says the service checks answers before showing them. It removes links that do not point to pages in the service, and only presents a quotation when its exact wording appears on the source page. Otherwise, it gives a plain explanation and a link. Answers can also direct users to a person, and the service has distinct messages for cases such as needing more detail, being unable to answer, or encountering an error.
Why it matters
This is a concrete example of AI being used to help people navigate public-sector guidance, with the consequences of a wrong answer treated as a design problem rather than a small-print afterthought. Showing the source behind an answer gives users something useful to check; saying when a simpler tool is enough is a welcome refusal to make every task look like an AI problem.
The assistant is not a substitute for professional judgement. That is Defra’s stated approach, and the advice for users is to check the source and consider the context. Sensible words; their value depends on whether the tool’s answers and hand-offs work in practice.
Our read
The strongest design choice is restraint: preserve exact policy wording, show the guidance behind an answer and make room for “I can’t answer” or “talk to a person”. In public services, a less impressive answer can be the more trustworthy one. The useful test now is whether those safeguards hold up when real users ask messy questions, not just the tidy ones in a demonstration.
What to watch
- Whether users can reliably trace answers to the relevant toolkit guidance.
- How often the assistant needs to hand questions to a person or say it cannot answer.
- Whether Defra shares what it learns before applying the approach to other services.
Discussion spark: For AI that explains official guidance, should the default be to answer only when it can reproduce a source-backed rule, or is a cautious plain-language explanation useful enough to justify the extra risk?
Sources and evidence
- Source update (18 November 2026, 00:00 UTC)
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.