Is it really that surprising that the Department for Business and Trade’s AI pilot resulted in a vague conclusion with no evidence for “productivity” gains?

Pilots like this – broad, un-targeted, and poorly defined – lack significance. They always remind me of what Ben Terrett said his line at GDS was: no innovation until everything works. This isn’t to say innovation shouldn’t mean try shiny tools around the edges, but rather, use that impulse to uncover where the real pain lies, prove what works (not what’s fashionable), and build from there.

In my role at HMRC, for instance, I manually log two separate timesheets across two systems, record attendance manually too, and still have to post HR documents by mail. Two and a half years in, I couldn’t tell you how to navigate our internal systems for guidance without help. That’s the texture of everyday inefficiency within the civil service – the context in which “AI productivity” lands.

Instead of buying into blanket technology fixes, maybe we should start from the socio-side of the work itself: understanding the friction civil servants face, design from the ground up, and measure progress by satisfaction and quality, not by output per hour.

After all, a modern civil service isn’t built on AI – it’s built on human-beings with a collective purpose, supported by systems that actually work.

DBT Pilot review: assets.publishing.servi…

Link to Ben terret’s post: linkedin.com/posts/bent…