← Back to writing

AI Training for Employees: A Practical Guide for UK Teams

On this page

AI training for employees works when people put AI to work on real tasks, feed it their own material and judge the results against their own standards. That’s how I’ve used AI tools almost every day for 3.5 years, and it’s how I train people to use them.

This guide is for UK firms without a training department, where the people who need to learn AI are also the people doing the work.

What should AI training for employees cover?

Good AI training for employees covers three things: giving the AI rich context, judging what comes back, and building something reusable for a task people do every week.

Context comes first. I tell people to think of the AI as a colleague. A search box gets keywords, and a colleague gets the full picture and the relevant files.

The quickest way to give that picture is voice typing. You talk through the problem, the goal, any constraints and what you don’t want, and the AI organises it. Your thoughts don’t need to come out tidy, and for many of my students, watching messy spoken notes turn into something clear is the moment it clicks.

Judgement comes next. AI produces work that sounds convincing very quickly, but plausible isn’t always accurate, useful or excellent, so people need to check every output against their own standard.

The third part is something reusable. In Claude that’s a Project, and ChatGPT has its own projects too. As I put it in my training, a chat is where your expertise gets spent, and a Project is where it gets stored.

What goes wrong when employees use AI without training?

People without training tend to give AI very short, generic instructions and leave it to fill in the gaps. It does that quite well, but the results come out generic.

That’s the risky part. Vague instructions now produce competent, generic work, which is harder to spot than rubbish, and someone who doesn’t use AI much may not know what good looks like well enough to tell the difference.

The other problem is that untrained staff don’t know how much AI can do. They assume some tasks aren’t suitable for AI, when it can help with a surprisingly large number of them.

Do employees need prompt engineering training?

Employees need to communicate well and think clearly about the problem they’re trying to solve, which is a simpler skill than “prompt engineering” makes it sound. Prompt templates and the word “engineering” make it seem far more complicated than it is.

If you can brief a capable colleague, you can brief an AI. Good training builds that habit.

So I teach a simple method: start naturally, use your voice to give context, continue the conversation, and keep ownership. The AI helps you explore and structure the work, and you verify, judge and decide.

Want your team using AI on their real work, with a method they can keep using? On a free 30-minute call, we’ll go through the tasks that take up most of your team’s week and work out where AI could save the most time. Book your discovery call.

Who should go first when you train employees in AI?

In a small firm, train everyone together. A team of 10 fits in one session, and everyone leaves using the same Projects and the same way of working.

In a bigger firm, start with a mixed group. Include the people who are most enthusiastic, because they’ll pass that enthusiasm on to their colleagues, and include at least one leader, because staff make time for what the boss uses.

Most importantly, include someone who’s an expert in the work itself. Judgement is what separates useful AI output from competent, generic output, and the people who know what good looks like in their job are best placed to spot the difference.

At Quest Professional, the executive principal, Claire Granados, took part in the session herself and came away with new approaches to her own work.

Whoever goes first should bring a task they do every week, so the first Project they build saves them time straight away.

Are free AI courses enough for employees?

Free courses are good for learning the theory, and on their own they aren’t enough. Every UK adult can take the free AI Skills Boost courses backed by the government, and Anthropic and OpenAI both publish their own learning material.

Anthropic’s Claude Academy has free courses, including AI Fluency: Framework and foundations, which runs to 14 lessons and about 4 hours. OpenAI Academy has courses for knowledge workers, builders and leaders.

Material from Anthropic or OpenAI gives people valuable theory, but nobody learns AI by reading about it. They have to put it into practice on their own work, regularly, and the holy grail of training is hands-on work on their own tasks.

What should employees have built by the end of AI training?

Each person should leave with something reusable built around a task they do regularly, usually a Project, and a more fluent way of working with AI. A Project holds what you know, and a Skill holds how you do something, so a step-by-step process they repeat can become a Skill. In my session for Quest Professional’s trainers, the aim was a Claude Project built around a recurring teaching task.

Building one starts with gathering what people already have:

  • two or three examples of the task done well, including their own
  • the standard it’s judged against, such as a checklist or rubric
  • the source material, if there are facts the AI could get wrong
  • one sample of their own writing, for the voice
  • the constraints: anything that must never appear

My rule here is simple: don’t ask AI to invent the standard, bring it one. Then the Project’s instructions say what it creates, who it’s for, what good looks like, the shape of the output and what to avoid.

Finally, people test it on a realistic task, check it against their standard and fix the right layer. A one-off problem gets fixed in the conversation, and a problem likely to come back gets fixed in the instructions or the source material.

How do employees keep using AI after training?

The habit that matters most is noticing repetitive manual tasks. When one comes up, open your AI tool, describe the task and what you want to achieve, and ask how it can help.

Doing that regularly is how people discover the full range of what these tools can do. It also breaks the habit of assuming AI can’t help with tasks it handles well.

The first week matters too. I ask everyone to use their Project for one real task, inspect the result, make one improvement and use it again.

How much does AI training for employees cost?

A hands-on team workshop with me costs £3,500 for a half day and £5,500 for a full day, for up to 10 people. For other UK providers’ prices, the free options and the staff time to budget for, see my guide to AI training costs in the UK.

If you’re still deciding which tool to train your team on, my comparison of Claude vs ChatGPT for work covers how I use each one. I teach both.

My AI Independence Foundations workshop teaches this method on your team’s real tasks, in Claude or ChatGPT, and everyone leaves with something they’ve built. See the programme details on my AI training for companies page.

Frequently asked questions

Where can I find AI training for my workplace?

You can start with the free AI Skills Boost courses and Anthropic’s Claude Academy, look at OpenAI Academy, or book a trainer to run a hands-on session on your team’s own work. I offer workshops online or in person in London and across the UK.

Is there free AI training available in the UK?

Yes. The government’s AI Skills Boost courses are free for every UK adult, and Anthropic’s Claude Academy also offers free courses.

What training do I need to work with AI?

You need to learn to give AI rich context, judge what it produces against your own standard, and build reusable Projects for tasks you repeat. No technical background is needed for everyday work.

How long does AI training for employees take?

My workshops run as a 4-hour half day or a full day. After that, people improve by using AI on their real tasks every week.

Book a free 30-minute discovery call, and we’ll work out what your team would build on the day and which programme fits.

Sources

Read on 4 and 6 November 2026: