1:1 AI training vs online courses vs corporate workshops

Most people who message me have already tried to fix this themselves. They bought a Udemy course and got three modules in. Or they sat through a half-day workshop at a client's office and came out with a PDF they never opened.

Then they ask me what the difference is between 1:1 AI training, an online AI course, and a corporate workshop. Underneath that question is a better one: why has nothing stuck yet?

Here's the honest version. All three work. They just work for different people, and most buyers pick the wrong one because they compare on price instead of on what they need to change.

I sell one of these three. I'll tell you where the other two beat me.

The short version

	Online AI course	Corporate workshop	1:1 AI training
Typical cost	£0 to ~£30/month	£475 (half-day) to £3,200+ (multi-day), per group	£199 to £699 per person
Who it's for	Anyone who wants the concepts	A team that needs a shared starting point	One person who needs their own work to change
Built around	A generic curriculum	The group's average	Your actual files, tools and bottleneck
The main risk	You don't finish it	Everyone nods, nobody changes	Cost per head is high
Best at	Foundations, cheaply	Shared vocabulary, permission, governance	Getting one specific thing working

Everything below is the reasoning behind that table, with sources, because almost nothing else written on this question has any.

Online AI courses: brilliant, free, and mostly unfinished

Start with the strongest case against me: the best AI courses now cost nothing.

Anthropic runs a free course catalogue that includes an "AI Fluency for Small Businesses" track: nine lessons, under an hour of video, no charge. OpenAI Academy is free too. In the UK, the government's AI Skills Boost programme is giving free AI foundations training to every adult, aiming at 10 million workers by 2030, including at least 2 million SME employees. Some of those courses take under 20 minutes.

The model-makers' own courses tend to be more current than the paid third-party ones, which is a genuinely awkward fact for anyone selling AI training. If your goal is to understand what a large language model is and why it hallucinates, pay nobody. Go and do one.

Now the problem.

Katy Jordan looked at completion rates across 221 online courses and found a median completion rate of 12.6%. The range ran from 0.7% to 52.1%.

MIT and Harvard then tracked 5.63 million learners across 12.67 million course registrations on edX over six years. Completion didn't improve over that period. It fell. By 2017-18, 3.13% of participants finished. Among people who had actually paid for the verified track, 46% finished. Fewer than half, and they'd handed over money.

That last number is the one that matters for you. Paying doesn't make you finish. Buying the course is not the same as learning the thing, and a lot of people reading this have a Udemy library that proves it.

There's a fair objection here, so let me make it for you: plenty of people enrol on a free course with no intention of finishing, which drags the average down. True. When HarvardX asked learners upfront what they were there for, 22% of those who said they intended to complete actually did, against 6% of the browsers. Better. Still barely one in five, and that's across nine courses back in 2013-14, so treat it as a direction rather than a law.

And then there's the shelf life. ChatGPT's underlying model has changed several times in the last year alone. Vanderbilt's prompt engineering course, one of the most popular on Coursera with nearly 700,000 enrolments, was built in 2023. The patterns it teaches are fine. The chat window it teaches them in is not the one you're looking at.

Buy an online course if: you want the foundations, you have the discipline, and the thing you're missing is understanding.

Corporate workshops: the cheapest way to move a group

I run these. I still think they're the wrong purchase for most people who ask me about them, and the reason isn't price.

On price, workshops win outright and it isn't close. Indexify publishes £475+VAT for a half-day for up to 10 people. Elansio does a half-day for up to 20 at £980+VAT. NobleProg starts at £3,200+VAT for a private group, though that's for a 14-hour course rather than an afternoon.

Do the per-head maths honestly, though, because the providers won't. That £980 half-day is £49 a head if you fill all 20 seats. If eight people turn up, it's £123 a head, and the price doesn't drop. Group pricing flatters itself by assuming a full room.

Worth knowing before you shop: of the eight UK providers I checked, five publish no price at all. If you can't find out what a workshop costs without a sales call, that's a finding, not an oversight.

What a workshop actually buys you is a group that heard the same thing on the same day. That's a real product and 1:1 training cannot produce it. If you need eight people to agree on what they may and may not paste into a chatbot, put them in a room. If you need your team to believe that using AI is sanctioned rather than sneaky, a visible session run by the boss does that in an afternoon.

What it doesn't do is survive contact with Monday.

The best study I can find on this asked training professionals across 150 organisations how much training people actually apply. Their answer: 62% of employees apply it immediately, 44% at six months, 34% a year later. It's a survey of estimates rather than a hard measurement, and I'd rather tell you that than pretend otherwise. But look at the shape of it. Everyone leaves the room fired up, and the decay starts that week. Nothing in a one-off workshop is designed to stop it.

While we're here: you will see a statistic everywhere claiming only 10% of training transfers to the job. Don't repeat it. It traces back to a 1982 article in which the author opened with a rhetorical question about what training directors estimate, cited no data, and has been formally debunked at least three times since. Half the AI training industry is quoting a made-up number at you.

Buy a workshop if: you have a group, you need a shared baseline, and the gap you're closing is permission and vocabulary.

1:1 training: the only format that touches your actual work

Here's the structural thing a course and a workshop both have in common. Neither of them can look at your inbox.

A course is generic by construction. A workshop is built around the group's average, which is nobody. Both teach you the tool. Only one format sits with the file you're stuck on.

That's the whole argument, and the research quietly backs it up. The big meta-analysis of training transfer pooled 89 studies and split skills into two kinds. "Closed" skills have a right answer and a fixed procedure. "Open" skills need judgement. For open skills, whether training transfers depends far more on what surrounds it than on the training itself: the correlation between a supportive work environment and actual transfer was .26 for open skills against .04 for closed ones.

Working out which parts of your job an autocomplete machine should touch is about as open as a skill gets. Which means the format that wins is the one that leaves you with support and momentum after the session ends, not the one with the best slides during it.

In practice, you walk out of a 90-minute session holding something that runs. Recent ones have produced a working CRM, an automated weekly numbers report, and a workflow that turns meeting notes into an action list. In one session a client suggested we try building a Pokémon card collectibles shop as a test, so we did, live, on the call.

The honest cost comparison: £199 for a Power Hour, £699 for a bespoke 90-minute session. Against a workshop at £50 to £120 a head, that looks expensive, and per head it is. But a workshop is priced per person in the room, and a 1:1 is priced against the one workflow you're going to stop doing by hand.

If that workflow is two hours a week, the maths does itself.

Buy 1:1 training if: you're one person, you already pay for ChatGPT or Claude, and you're getting almost nothing back for it.

Which one you should actually buy

Look at what the UK data actually shows. Only 21% of UK workers feel confident using AI at work. And on ONS figures, micro businesses with fewer than 10 people have a 24.3% AI adoption rate against 44.4% for businesses over 250. Free courses have been abundant for three years, and those numbers are where they are anyway. Information was never the thing in short supply.

So pick on the change you need, not the price:

• You need to understand AI. Free course. Anthropic's or OpenAI's. Don't pay anyone, including me.
• You need a group to start from the same place. Workshop. Cheapest thing per head that exists, and it does something 1:1 can't.
• You need your own work to change this month. 1:1. It's the only one of the three that opens your actual files.
• You're a company or a department, not a small team. That's a different product and I don't sell it here. My sister business, Exponential Partners, does.

Most owner-managers I speak to have already done the free course. That's why they're messaging me.

Common questions

Is a group AI workshop or 1:1 training better for a small team? For 3 to 10 people who need a shared baseline, a workshop is better value and better suited. For an owner-manager who needs their own workflows rebuilt, 1:1 is the only format that does it. Plenty of small businesses do both, in that order.

Do online AI courses actually work? For foundations, yes, and the free ones from Anthropic and OpenAI are excellent. The issue is completion. One study of 221 online courses found a median completion rate of 12.6%. Separately, on MIT and Harvard's edX data, even the students who paid for the certificate track finished less than half the time.

How much does AI training cost in the UK? Online courses: free to about £30 a month. Group workshops: roughly £475 to £3,200 per engagement, depending on length and group size, though most UK providers don't publish prices at all. 1:1 training: £199 for an hour, £699 for a bespoke 90 minutes.

Does one-off AI training stick? Not on its own. Application of training decays over the year following it. Whatever format you choose, the thing that makes it stick is having something built during the session that you then use on the Monday.

If you're not sure

Message me and tell me what you're actually trying to fix. If the answer is a free course, I'll say so. If it's a workshop for your team, I'll say that too, and I run those.

If it's your own work that needs to change, book a session and bring the thing you're stuck on.