This is not a time for violins though. We can explain.
At the time, the logic seemed perfectly reasonable. AI could help people explore ideas, practise skills, apply what they were learning and continue their development outside the workshop. So if some AI was useful, adding more of it should make the programme even better.
It didn’t.
The programme was by no means a disaster. In fact, it still achieved a 92% recommendation rate. Most people would probably be pretty happy with that.
We weren’t though.
We were used to seeing something closer to 99%, and we have very high standards for the programmes we put in front of people. So, whilst nobody was storming out demanding they never had to come back, we knew something had shifted.
It started to feel a little too much like an AI experiment and a little less like a development programme. The balance felt off.
And that raised a question which has changed the way we use AI ever since.
How much AI is actually useful in learning?
Our approach to AI-Augmented learning has never been about replacing human-led development.
It is actually quite the opposite.
We design programmes around human skills, human behaviour and the wonderfully inconvenient fact that people rarely behave exactly as a model says they should. A leadership programme still needs people challenging each other. Managers still need to practise difficult conversations with other humans. Teams still need opportunities to disagree, discuss, reflect and occasionally realise that the problem they thought they had is not actually the problem at all.
AI can do some remarkable things, but if 15 people have travelled somewhere to learn together and you spend half the day putting them behind individual laptop screens, something has probably gone awry.
Our original question was therefore fairly simple.
Could we use AI to make the time people spend learning together even more valuable?
And could we build something around the programme that helps the learning continue when the facilitator is no longer standing in front of them?
The answer to both is a definitive yes. But there is a rather important how attached.
One of our early AI-augmented programmes had quite a lot of AI in it. Too much in fact.
There were sections of more traditional facilitated learning, followed by periods where participants moved onto AI tools to work through exercises and application.
On paper, that sounds sensible. You learn something and use this amazing technology to help you apply it.
In reality, the transition felt clunky. Instead of the AI disappearing naturally into the learning experience, it became an event in itself.
Right, everybody, let’s get those laptops open! It’s AI time!
And once the technology becomes the most noticeable thing in the room, there is a danger that the learning starts playing second fiddle to it. The participant feedback reflected that. So, we analysed and changed it for the better.
I actually think that matters.
There is an understandable temptation at the moment for any company doing something with AI to present a beautifully polished version of events in which they had a visionary idea, built it perfectly first time and then strolled slowly into the sunset while everybody applauded.
That has not been our experience.
We have tried things. Some worked brilliantly and exactly how they were intended to. Others worked less brilliantly. But that experimentation gave us a much better idea of what actually belonged in the programme.
Which, now I think about it, is a fairly appropriate way for a learning company to approach this.
The conclusion we reached was not that AI had failed.
Some of the most useful parts of the programme were only possible because of AI. The problem was that we had started with the technology and looked for opportunities to use it.
Now we do almost the reverse. For every part of the learning experience, we can ask:
Should AI be used here?
Could AI be useful here?
Or would this be better without it?
That last question is surprisingly important because sometimes the right answer is to put phones away, to shut down laptops and to talk to another human sharing this experience with you.
If someone is exploring trust, giving difficult feedback, handling conflict or thinking about how their behaviour affects the people around them, introducing a chatbot into every moment does not automatically improve things.
Sometimes people need information, practice and another person to make the experience worthwhile.
Good learning design has always depended on knowing the difference and AI hasn’t changed that. It has simply given us another very powerful option.
One area we are particularly interested in is what happens outside formal learning sessions.
Think about the traditional programme workbook. You attend a leadership programme and you’re given a beautifully designed document containing the models, exercises, ideas and supporting information.
Then six weeks later you need to give somebody difficult feedback and you remember that there was that model that was in the workbook.
I think it was page 47, or maybe it was 74?
Also where is the workbook? Did I put it in my drawer, or did I even take it out of my bag?
The PDF may be somewhere in my inbox. There is probably a folder on SharePoint that theoretically contains it, although nobody has opened that for a while.
AI changes that interaction completely. Instead of searching through a manual, you can ask a question.
“What was the feedback model we learned?”
Then…
“I need to give feedback to someone on my team who keeps interrupting colleagues in meetings. Help me think through how I could approach it.”
Now the programme content is doing something. It is helping somebody apply learning to an actual situation at the moment they need it.
They could rehearse the conversation, use it to challenge their first approach, potentially ask for another version and come back after the conversation and reflect on what happened.
That starts to look less like a digital workbook and more like ongoing access to the programme itself. And that, for me, is where AI becomes genuinely interesting for L&D.
Most leadership development programmes have a practical limitation.
Eventually, everybody goes home, the facilitator packs up and leaves, the coaching sessions come to an end and the reality of work starts happening again.
And that is normally when the difficult bit begins, because knowing a coaching model in a workshop environment that is focused on it is one thing, however using it with an employee who has just told you they are furious about being passed over for promotion is another.
AI gives learners somewhere to go between those formal learning moments. Not to replace their manager or a coach, and not to replace judgement, but to ask…
“How could I use what I learned here?”
“What might I be missing?”
“Can we practise this?”
“Explain that model again, but use my situation.”
That dramatically increases the number of opportunities a person has to engage with the learning. It also puts much more control into their hands.
Someone might want to read the material, someone else might want to talk it through, another might want a short summary before a meeting, maybe you even prefer to listen on a commute.
The underlying learning has not changed but the way someone interacts with it can. That is a much more interesting use of AI than simply producing course materials faster.
There is another lesson here… an AI-Augmented programme should not require a client to adopt whichever shiny platform happens to be fashionable that month.
Different organisations have different approved systems, security requirements, policies and levels of AI maturity.
One client prefers Gemini, another Copilot, another likes Claude.
Another may want a much more controlled environment where the AI only works from approved source material. That is why we are increasingly interested in designing the learning resource itself to work across different AI environments rather than forcing everybody into one platform.
The technology will keep changing anyway. Building a learning strategy around this week’s favourite AI tool feels like building your house on a very springy trampoline.
The programme needs to survive the technology that is constantly moving underneath it.
This might be the most important lesson we have learned so far.
When AI augmentation is working well, participants should not constantly be thinking about the fact they are on an AI-Augmented programme.
They should be thinking about the difference that the programme is there to create. That might be them becoming a better manager, or having a more useful conversation or solving a problem with their team.
AI should appear when it helps but then immediately get out of the learner’s way.
That means there may be parts of a programme where we actively encourage people to use it, there may be optional opportunities where someone can choose whether it would help them, and there will be moments where we deliberately leave it out.
That is not being cautious about AI… I think it is being more ambitious about learning.
Because the objective is not to prove how sophisticated the technology is. The objective is behaviour change boosted by the technology.
I suspect we will see plenty of “AI-powered”, “AI-enabled” and “AI-first” learning propositions over the next few years.
Some will be brilliant and some will be existing digital learning with a chatbot delicately balanced on top. So, I would ask a few fairly simple questions.
What does the AI allow the learner to do that they could not do before?
Where has it deliberately not been used?
How does it help people apply learning to their real work?
And if you removed the phrase “powered by AI” from the proposal, would the programme still sound like good learning?
That last one is quite revealing because AI cannot rescue weak programme design.
If the learning is generic, disconnected from the organisation and unlikely to change behaviour, generating it faster does not make it better. You simply arrive at mediocre learning more efficiently… which is a bit like flying to the place you’d least like to visit on Earth rather than driving there.
AI-Augmented learning is going to keep changing. We are definitely not writing this blog and expecting to never have to update it and redefine what great learning looks like with AI.
The tools are developing ridiculously quickly. The things that feel experimental today will probably seem completely ordinary before long. So, I do not think the sensible position for an L&D provider is to claim it has solved all of this.
We are not claiming this, but what we do have though is valuable experience.
We have already started integrating AI into real development programmes. We have seen where it adds something and we have seen where it starts taking something away, and we are adjusting our approach as a result.
Perhaps that is the useful distinction. The question is no longer whether AI belongs in learning because of course it can.
The more interesting question is where.
Because the most sophisticated AI-Augmented programme probably will not be the one with the most AI in it. It will be the one that knows exactly when to use it.
And, just as importantly, when to not use it.
Interested in exploring how AI could augment an existing leadership or management programme? Talk to us about what that could look like for your learners.
Thanks
Alex & The Excel Team