Fautons
9 min readAI trainingL&D

AI in learning and development: the two jobs L&D now has

AI in learning and development: the two jobs L&D now has

What is AI in learning and development?

AI in learning and development means two things. The first is AI tools inside L&D: drafting course content, personalising learning and analysing skills gaps. The second, and now the more pressing, is L&D's job of getting the whole workforce to use AI well at work. Most L&D teams are asked to do both at once.

Most articles on AI in L&D cover only the first meaning. They list tools that turn a document into a course, a script into a video or a policy into a quiz. That part is real and useful, and it gets a fair hearing below. But if you run L&D, the second meaning is probably the one landing on your desk, so it gets most of this post.

Here is why it lands on you. Staff are already using ChatGPT, Microsoft Copilot or Claude, with or without a plan. Someone has to make that safe and useful, and the job tends to fall to whoever owns learning.

How far L&D teams have got

Further than most other functions. LinkedIn's 2025 Workplace Learning Report, based on a survey of 937 L&D and HR professionals and 679 learners, found that 71% of L&D professionals are exploring, experimenting with or integrating AI into their work.

Synthesia, which sells AI video software, surveyed 421 L&D professionals in October and November 2025. 87% said they already use AI: 57% actively and 30% in early pilots. The most common uses were voice and text-to-speech (63%), drafting content and quizzes (60%), video (52%) and translation (38%). Synthesia says its sample leans towards large companies and early adopters, so read these as the upper end rather than the typical team.

The blockers in the same survey are telling. Security came first at 58%, then accuracy at 52%. Next came lack of internal expertise at 46%. When asked what support they wanted, 67% said training in AI skills and design, and 63% said guidance on workflows and on measuring impact. So the people responsible for teaching skills are asking for skills themselves. That is the first thing to fix.

Where AI helps across the learning design cycle

The easiest way to see where AI fits is to walk through the stages most L&D teams already work to: analysis, design, development, delivery and evaluation. At every stage, AI is good at producing a first version. A person still owns the decision about whether it is right.

StageWhat AI does wellWhat a person still owns
AnalysisSummarising survey comments, exit interviews and performance themes into a first list of skills gapsDeciding which gaps matter to the business and which are noise
DesignDrafting learning outcomes, a session outline and a few ways to structure a programmeChoosing the approach that fits the audience, the time available and the culture
DevelopmentFirst drafts of scripts, scenarios, quiz questions, voice-over and translationsChecking every fact, especially in compliance and safety content, and cutting what is generic
DeliveryPractice partners for role-play, quick answers to learners' questions, reminders between sessionsRunning the room, reading the group and handling the questions AI gets wrong
EvaluationSorting open feedback and pulling themes from follow-up conversationsJudging whether people actually work differently, which no tool can see for you

The pattern holds across the table. AI saves the most time where the work is drafting and summarising, and the least where the work is judgement. Teams get into trouble when they let the speed of the first column push out the second.

The risks of AI in L&D

Accuracy is the obvious risk, and it matters more in learning than almost anywhere else. A model can write a confident, well-structured module that is wrong. If that module is about health and safety, data protection or a regulated process, you have trained people to make a mistake. Every fact in AI-drafted content needs checking against the source, by someone who knows the subject.

Data is the second risk. Skills-gap analysis is tempting to automate, but performance reviews and employee records are personal data. Agree with your data protection lead what may go into which tool before anyone pastes anything in. We cover the everyday rules in how to use AI at work without creating a problem.

The third risk is quieter: more learning, not better learning. When a course takes an afternoon instead of a month, it is easy to produce ten of them. Volume was never the problem L&D had. A shorter catalogue that people use beats a long one built in a week.

The second job: teaching everyone else to use AI

This is where L&D earns its place in an AI rollout. The basics are now free. Since January 2026, every adult in the UK can take free online AI courses through the government's AI Skills Hub, and some take under 20 minutes. Courses checked against Skills England's AI foundation skills for work benchmark cover understanding the risks of AI, writing prompts, reading simple AI dashboards and automating routine tasks.

Point staff at them. They are a sensible floor. But a 20-minute course teaches a person what the tools do. It doesn't change how a team works, and that is what the organisation is paying for. Closing that gap is L&D's job, and a few things make the difference:

  • Teach by role, on real tasks. A finance team and a sales team use the same tool for different work, so build sessions around each team's own documents and deadlines.
  • Space it out. A single day fades fast, which is the problem we describe in why one-off AI training workshops don't stick. Several short sessions with practice in between hold up better.
  • Teach the rules alongside the skills. Your approved tools, your AI policy and what must never go into a public tool belong in the training, not in a separate document nobody reads.
  • Grow champions. The people who start showing colleagues their workflows are the ones to back, as we set out in how to build an AI champions programme.
  • Measure behaviour, not attendance. Take a baseline first, then look at how people work a few weeks later. Our guide to measuring AI training success covers how.

What to train first, by team

Each team needs a different first task, one it does every week and can check easily. These are the starting points we would pick, with a fuller guide for each:

A first 90 days for an L&D team

You don't need a strategy deck to start. A quarter is enough to get your own team fluent and run a first pilot for the business.

  • Weeks 1 to 4: your own team first. Pick one real course you need to build anyway and build it with AI, checking every output. You can't teach what you haven't done.
  • In the same month, agree the data rules with your data protection and IT leads: which tools are approved and what may go into them.
  • Weeks 5 to 8: one pilot group. Choose a team with a clear, repetitive workload and a manager who wants it. Baseline how they work, then run hands-on sessions on their own tasks.
  • Weeks 9 to 12: review honestly. What do people do differently now? What did they stop using? Use the answer to plan the next cohorts, and write down what didn't work as well as what did.

If you are planning the wider rollout, our guide to rolling out AI across your team picks up where this leaves off.

Buying AI training as an L&D manager

At some point you will decide whether to build the training in-house or bring someone in, and what to pay. Build it yourself if your team already uses AI well on its own work. Bring someone in if it doesn't, because you can't teach what you haven't done. Either way, ask any provider three things. Will the sessions use our own tasks and documents? Will they teach our AI policy and approved tools? What will we be able to show four weeks later?

We've written separately on what corporate AI training costs and how to choose an AI training provider, with a checklist you can take into a first call.

Where Fautons fits

We run live, hands-on AI training for UK teams, and L&D is usually the team we plan it with. We scope each programme with L&D and team leads first, including your approved tools and your AI policy. People then practise on their own work in cohorts of about 10 to 15. Afterwards you get recordings and materials, follow-up office hours, internal champions to carry it on, and a wrap-up report for L&D on who took part and what they built.

With a training partner, we co-delivered a six-week programme for around 20 senior product managers and designers at a Fortune 500 tech company, on their own product work. It is the spaced, hands-on shape this post argues for. If you are working out how your organisation should learn AI, see our AI training across the UK or book a discovery call and we'll talk it through.

Frequently asked questions

How is AI used in learning and development?

In two ways. L&D teams use AI in their own work to draft course content and quizzes, create voice-over and video, translate material and summarise feedback. And L&D increasingly leads how the rest of the organisation learns to use AI, through role-based training, AI policy and champions. In Synthesia's late 2025 survey of 421 L&D professionals, the most common uses were voice and text-to-speech (63%), drafting content and quizzes (60%), video (52%) and translation (38%).

Will AI replace L&D professionals?

It replaces some of the work, mostly the first drafts. It doesn't replace deciding what people need to learn, checking that content is right, running a room or judging whether behaviour changed. Those parts of the job matter more now, because AI has made poor content quick to produce and the organisation needs someone to teach it to use AI well.

Are the free government AI courses enough for our staff?

They are a good floor. Since January 2026 every UK adult can take free AI courses through the government's AI Skills Hub, and some take under 20 minutes. They teach individuals what the tools do. They don't train a team to use AI on its own work under its own rules, and that is where most of the value is. Use the free courses for the basics and hands-on, role-based training for the change.

What AI skills do L&D teams need first?

Writing prompts that hold up on real work, checking AI output for accuracy, knowing what data may go into which tool, and building one real piece of learning end to end with AI. In Synthesia's late 2025 survey, 67% of L&D professionals said they wanted training in AI skills and design, so this is a common gap rather than a sign a team is behind.

How should L&D measure whether AI training worked?

Take a baseline before the training: how many people use AI on real work, how long key tasks take and a few samples of work. Then look again four to six weeks later for changed behaviour, not just completion rates or satisfaction scores. Kirkpatrick's four levels are a useful frame, and the value sits in levels three and four.

Sources

More from our Blog

November 18, 20269 min read

AI for small business: how to train a team of 10 to 20

Most guides to AI for small business are lists of tools. The tools are the cheap part. This is how a business of 10 to 20 people gets everyone using AI well on real work, in about a month.

AI trainingSmall business
Read the article
November 13, 20269 min read

AI for procurement: what your team should learn first

Most guides to AI for procurement describe what a platform can do. This one is about the people: the four jobs AI already does well for a buying team, what to teach first, and a one-day workshop plan built on your own documents.

AI trainingProcurement
Read the article
October 18, 20268 min read

How to roll out AI across your team: an adoption plan that sticks

Most AI rollouts stall because they're run like a software purchase when they're really a change programme. Here's a five-step plan for People and L&D leaders: start with a reason, pilot on real work, set guardrails, train people properly and measure weekly usage.

AI adoptionAI training
Read the article