Politics

Copy article

New report published on UK Government AI Transformation

ended 23. March 2026

A new AI report states that two thirds of senior civil servants want a say in how AI reshapes their work. Only 29% have been asked.

The Fabian Society and FDA union surveyed over 2,000 civil service managers and the results quietly demolish the "blob blocking progress" narrative. 72% have either already introduced AI into their teams or want to. 55% say their organisation wastes resources. These aren't resisters. They're people watching failure from the inside and asking for tools to fix it.

But almost nobody is consulting them. Ministers talk about rewiring the state while the people who actually run it are left guessing what's coming and why. The report warns that concern levels (79% report at least some worry) aren't evidence of resistance. Only 3% could cite something that had actually gone wrong. Over half offered hypothetical fears. That's what happens when you skip the conversation: people fill silence with worst cases.

The Fabian Society is direct. It believes this is about insufficient clarity, consultation and communication, not about harm. They explicitly warn against copying the US "project chainsaw" model of cuts first, questions never. Their five recommendations (performance bonuses for delivery, access to approved tools, protected training time, departments keeping their own savings, and digital platforms for staff input) aren't radical. They're the minimum conditions for transformation that doesn't collapse on contact with reality.

We'd like your views:

  • If 72% of civil servants are willing to build with AI, what exactly is the bottleneck, and who owns removing it?
  • Should departments that generate AI driven savings be allowed to keep and reinvest them, or does Treasury clawback kill the incentive to improve?
  • The report says concerns are driven by communication gaps, not evidence of harm. Is that a fair reading, or does it undercount risks that haven't surfaced yet?
  • What does meaningful staff consultation on AI actually look like in a department of thousands and how do you stop it becoming another delay mechanism?
  • The Fabian Society warns against the US model. Is there a version of "move fast" that genuinely works with the workforce rather than around it?

4 responses from the Newspage community

Copy all

Star Quote
Copy

Colette Mason
AI Ethics Consultant at Clever Clogs AI
72% of civil servants want to build with AI. Only 29% have been asked. Meanwhile, vendors are selling "fully autonomous" workflows to the same departments that haven't consulted the people doing the work.
The gap between what marketers promise and what workforces need is where AI transformation actually fails. You can't sell "no humans required" to an organisation where the humans are the ones who know where the processes break. Civil servants aren't blocking change. They're watching it arrive without training, tools, or a conversation. The Fabian Society found that most concern isn't based on evidence of harm, Rather, it's based on silence. When nobody explains the plan, people write their own. Removing humans from the loop isn't efficiency. It's flying blind towards the ground with shinier software as the soon-to-fail parachute.
Copy

Editor
Nouran Moustafa
Practice Principal & IFA at Roxton Wealth
The bottleneck is clearly not willingness, it is leadership, trust and incentives. If most civil servants are open to using AI but barely any have been properly consulted, then the blockage sits at the top. You cannot modernise the state by talking at the workforce instead of with it. On savings, yes, departments should be allowed to keep and reinvest at least some of what they generate. If Treasury claws everything back, you kill the incentive to improve and turn transformation into theatre. I do think communication gaps are driving a lot of the anxiety, but that does not mean the risks are imaginary. In government, hypothetical risk still matters because the stakes are high. The answer is not to slow everything down, it is to involve staff early, pilot properly, train seriously and put clear guardrails in place. Real reform is not “move fast and break things”. It is move smart, bring people with you, and build something that actually works in real life.
Copy

Editor
Mitali Deypurkaystha
AI Governance Strategist & CMO at AISensa
72% support but just 29% consulted shows a top-down transformation trap taking hold across government. This isn’t resistance. It’s what happens when change is imposed rather than built with the people doing the job. I see this repeatedly in many businesses: the larger the organisation, the bigger the problem. It’s exactly why I insist on mapping real workflows with frontline staff before any AI and automation is introduced. That approach delivers real-world efficiency and buy-in from the very people who will use these tools, because they are shaped by reality, not assumptions. Ignore that, and AI risks becoming another reform done to the workforce, not with it, eroding both delivery and trust.
Copy

Rohit Parmar-Mistry
Founder at Pattrn Data
The bottleneck is not willingness. It is the plumbing: access to approved tools, data permissions, and a delivery culture that rewards solving problems rather than avoiding blame. If two thirds of senior staff want a say and only a minority have been asked, you get the worst of both worlds: leaders announce “transformation” while teams treat AI as an unofficial side project.

Ownership has to be explicit. Central government should set clear guardrails (security, privacy, procurement frameworks, model and vendor evaluation) and departments should own delivery outcomes, with named SROs for AI-enabled service improvements.

On savings: Treasury clawback is a strong anti-incentive. Letting departments keep a portion of verified productivity gains to reinvest in skills, data clean-up and tooling would create momentum.

Consultation should be structured: short, time-boxed staff input cycles tied to specific workflows, plus publishing what was adopted and why.