Inside the Room: How UKAI's Responsible AI Working Group Is Building a Use Case Library for SMEs

On Tuesday 21 July, the UKAI Responsible AI Working Group swapped screens for a whiteboard. Hosted by the Federation of Small Businesses (FSB) at their Westminster office, the session brought together working group members, invited contributors, and FSB policy staff for an afternoon of hands-on, practical work: not another discussion about AI governance in the abstract, but an attempt to actually start building something SMEs can use.

From Direction to Delivery

The 21 July workshop was the group's first in-person session, building on the foundations laid at its inaugural meeting back in April. That first meeting, chaired by Matthew Holmes, set the group's strategic direction: rather than adding to the pile of high-level AI governance commentary, the Responsible AI Working Group would focus on producing practical, cross-sector tools for an audience that's too often overlooked in the AI governance conversation - SMEs, who typically have far less existing governance infrastructure than large enterprises.

Two core outputs were agreed at that first meeting: a use case library, covering governance, data readiness and impact measurement; and an accountability matrix, clarifying who's responsible for what in organisations where one person often wears several hats.

The FSB session was where that ambition became a working prototype.

What We Actually Built

Held as a closed "build sprint" - big tables, whiteboards, and real use cases on the table rather than hypotheticals - the session worked through two detailed foundational use cases in depth, contributed by Satheesh Pillai and Isabel Bathurst, testing a proposed structure against real organisational experience.

The use case format the group stress-tested covers categories most conversations about "responsible AI" skip entirely: not just what a tool does, but how governance actually got set up (often organically, through trial and error, rather than as a formal process); how impact was assessed (frequently anecdotal - did we win the pitch, were clients happy rather than measured); what the legal and IP position looks like when a document is AI-assisted; and where the open questions still are.

Some of what came out of the room:

• Governance is emergent, not designed. In both foundational use cases, controls grew out of active, everyday use rather than a top-down policy - a checkbox here, an informal chat with IT there. That's a realistic starting point for most SMEs, and the working group's job is to help formalise it without making it unworkable.
• "Quality" and "impact" are being judged informally. Neither organisation had a structured way of assessing whether their AI deployment was actually working. Success was read off proxies - did we win the business, are people happier rather than any deliberate measurement process.
• Constraints can drive better use, not worse. One case showed that tightly scoping what AI tools were allowed to touch (keeping client data in walled-off systems, for example) didn't just manage risk - it pushed the team toward more disciplined, more effective use.
• The open questions matter as much as the answers. The session surfaced genuinely hard, unresolved questions the working group will keep chewing on: How do you tell genuine innovation from AI outputs apart from pattern-matching or "algorithmic monoculture," where everyone converges on the same answers because they're using the same tools the same way? Who owns liability for an AI-assisted document when there's no settled UK legal framework yet? How do you train juniors when entry-level tasks are the first thing automated away?

Setting the Scene: FSB's Own Research

The session opened with a short presentation from Iliana Pearce, Policy Advisor at FSB, sharing early findings from FSB's own research into small business confidence and AI adoption - a fitting jumping-off point for a room about to spend two and a half hours getting into the practical realities of responsible AI use.

That research has since been published in full. FSB's report, The Confidence Code, went live, and it's a valuable companion piece to the working group's own output: where the Use Case Observatory is building the practical, on-the-ground detail of how individual SMEs are actually using AI, FSB's report speaks to the wider confidence and adoption picture across the small business community. Well worth a read alongside this piece.

What's Next

Attendees are now completing their own use case entries against the shared structure, due by 19 August, before the group reconvenes to refine the format, develop supporting commentary, and start identifying the patterns that repeat across use cases - patterns that were deliberately left out of this first pass so the group could establish a solid structure before layering on interpretation.

The ambition is for this foundational group to do the groundwork that makes it easy and genuinely worthwhile for the wider SME community to contribute their own use cases once the library opens up more broadly.

Thank you to FSB for hosting the session at their Westminster office, and to Matthew Holmes for chairing, along with everyone who gave up an afternoon to build something real rather than talk about it. This is exactly the kind of practical, member-led work UKAI's working groups exist to produce.

Interested in getting involved in the Responsible AI Working Group, or want to know more about the Use Case Observatory? Get in touch with the UKAI team.