dug 
All this talk of reorganising departments, disbanding DSIT etc. is a bit confusing. Mind you, I was already confused…
When GDS was created, it led a massive cultural change with lasting positive outcomes across government departments — and arguably across governments worldwide. It was a coordinated, technologically and culturally relevant response to the maturing of internet technologies, and to the opportunity they offered: universally accessible government services.
GDS was the digital centre of government. Folk like Martha Lane Fox, Francis Maude, Mike Bracken, Tom Loosemore, Ben Terrett, Russell Davies, Lou Downe, Tim Paul, Andrew Greenway, Richard Pope and many others actually changed the world. And they did it from the actual centre of government — GDS sat in the Cabinet Office from 2011 to 2025, when it was moved to DSIT.
Except DSIT is not the centre of government.
Last month the Science, Innovation and Technology Committee said as much — concluding the move had hamstrung transformation plans and that the digital centre should have stayed in the Cabinet Office. Now the talk is that DSIT itself may not survive the reorganisation, with its functions folded into a business department, which would move the digital function further from the centre still.
I don’t know how this plays out, and it’s way above my pay grade, but here’s what I worry about:
I worry that clustering AI practitioners, policy people and researchers into a siloed ‘centre of excellence’ — an ivory tower — means all that effort and noise happens in the tower, not down where services are designed and delivered — where teams still sit with members of the public, and learn from watching them struggle. That’s where the magic happens, and where good judgement about any ultimately human-serving technology actually gets formed.
I worry we’re chasing shiny — that being seen to do cool stuff like the American tech bros matters more than knowing why. “AI adoption” is becoming the metric, and Goodhart’s law tells us what happens next: when the measure becomes the target, it stops measuring anything useful.
Adoption isn’t outcomes. It says nothing about whether services improved, or what happens to jobs, education, culture — the society on the receiving end.
We’re standing up teams, budgets and kit to become “AI-native” before we’ve asked what native means. We’re swallowing the pills before the clinical trials have even started.
Maybe there is no pattern from 2011 that can provide answers for 2026. In 2011, if you had the right taste and knew what to look for, there was a right tech horse to back.
But in 2026, could it be that there is simply no way of knowing which horse is the right one?