How UK government teams can get hands-on Claude and AI training in 2026

The push is real — and most of it hasn't landed yet

There's a genuine, funded ambition inside the UK civil service to get everyone using AI well. In October 2025 the Cabinet Office made its annual 'One Big Thing' campaign all about it, under the banner 'AI for All': the goal is for every civil servant to become a confident, responsible user of AI.

The reason it's a campaign and not a footnote is that the reality is a long way behind the ambition. The same programme cited that only 28% of civil servants felt confident using AI at work. So there's a large, official gap between where teams are and where they're being asked to get to — and closing that gap is a training problem, not a technology one.

What the government has actually put in place

It helps to know what already exists, because a lot of it goes unused. There are three things worth knowing about:

• A learning entitlement. The AI Playbook for the UK Government states that every civil servant can take up to 5 days of learning a year. That's real time, already budgeted, that a team can point at AI.
• Free foundational courses. There's a set of free online AI courses on Civil Service Learning — fundamentals, ethics, the business value, an overview of the tools. Good for grounding, and a sensible first step before anything hands-on.
• A shared vocabulary. The Government Campus AI prospectus sorts AI learning into Foundation, Working and Expert tiers. It's worth using, because it gives a scattered team a common way to talk about who needs what.

None of this requires new money or a business case. If you lead a team, you already have the time and the framework. What's usually missing is the part that turns awareness into capability.

Why free modules alone don't build capability

Here's the uncomfortable bit. A person can complete every foundational module and still not be able to use AI on their actual job. Modules are good at explaining what a large language model is and what not to paste into one. They're poor at building the habits that make AI useful under real conditions.

The skills that matter are judgement (knowing what to hand to AI and what to keep), workflow design (breaking a task into steps rather than hoping one prompt does it), and verification (knowing when an answer is safe to act on). We wrote about those in AI skills every professional needs in 2026. You don't get them from a video. You get them by doing your own work with AI, getting a few things slightly wrong, and adjusting — ideally alongside someone who already works this way.

What good hands-on training looks like for a public sector team

The difference between training that sticks and training that evaporates is whether people worked on their own tasks during it. For a government team, that means a few specific things:

• It's on your real work. People practise on the briefing, the consultation response, the case notes, the spreadsheet they'd have done anyway — not a toy exercise about a fictional company.
• It respects your data rules. Good training bakes in what can and can't go into a tool, and teaches the safe pattern, rather than treating data handling as an afterthought.
• It builds verification in. Public sector work has to be defensible. So the habit of checking an AI's output before acting on it isn't a nice-to-have; it's the point. The practical version is in what responsible AI use actually means.
• It maps to the tiers. Frontline staff need the Foundation and Working level — confident, responsible use on everyday tasks. Digital and data teams need the Expert level — building workflows, prompt systems and safe automation. Leaders need enough to decide what to adopt and how to govern it.

That's the shape of the AI and Claude training we run for public sector teams, delivered live online across the UK or in person in Birmingham and the West Midlands.

Claude, Copilot, and choosing what to train on

A fair question is which AI a team should actually learn. Many departments already have Microsoft Copilot through their M365 estate, so it's the tool sitting in front of most staff, and training people to use what they already have is the pragmatic starting point.

Claude is where a lot of the quality lives for the work civil servants actually do — careful drafting, summarising long documents, reasoning through analysis, working with source material without losing the thread. In practice the right answer for most teams isn't one tool; it's teaching people to use what they have well, and showing where a model like Claude does the harder thinking better. The skill is choosing the right tool for the task, which is really just judgement again. I went through the trade-offs properly in Copilot vs Claude vs ChatGPT for government teams.

How to actually get it going

If you want to move from 'we should do something about AI' to people who are genuinely more capable, a simple sequence works:

• Start with one team, not the whole department. A single group that gets properly good is more convincing than an org-wide module nobody remembers.
• Point the 5 days at it. Use the learning entitlement you already have rather than waiting for a new budget line, and see how to actually spend those 5 days on AI.
• Map it to the tiers. Decide who needs Foundation/Working and who needs Expert, so the training fits the people rather than the other way round.
• Pick a provider who'll meet you where you are. That means someone UK-based who'll complete your security and supplier due diligence and work to your procurement route, and who teaches on your real tasks. If you'd like to see how we do it, our AI training across the UK page has the wider picture, and the public sector page is the specific one.