AI and Learning: How LLMs Are Changing the Way We Learn at Work

AI Is Changing How Fast I Learn: What Happens When Knowledge Can Be Applied Almost Immediately?

One of the biggest changes AI has made to my work isn’t how quickly I can get an answer.
It’s how quickly I can turn that answer into something useful.

At times, using an LLM feels a little like the film Limitless — minus Bradley Cooper and the drugs.
Not because it suddenly gives you unlimited knowledge, but because the distance between:

“I don’t know how to do this”

and

“I understand it, I’ve applied it and I’m building on it”

has become incredibly short.

That change has made me think quite a lot about how I learn, what I actually need to remember, and whether AI is making me more capable or simply better at retrieving answers. I’m increasingly convinced it can be the former — if I use it properly.

Learning in the context of real work

I see this constantly in Power BI. I can hit a problem, learn a concept or technique I haven’t used before, apply it immediately to a real reporting issue, test the result and keep moving. That is very different from looking up random information and hoping some of it sticks.

The learning has context.

There is a real problem to solve, a reason to understand the answer and an immediate opportunity to put that knowledge to use. As an information designer, that makes intuitive sense to me. Information is easier to understand and remember when it forms part of a coherent narrative rather than existing as isolated facts.
In my case, the “story” is the work itself.
There is a problem.
There is a reason I need the answer.
There is an immediate consequence if I apply it correctly.

That makes the learning useful straight away.

The learning loop has become much faster

Traditionally, professional learning often follows something like this:

Learn → practise → eventually apply

With an LLM embedded in my day-to-day work, it increasingly looks more like:

Encounter problem → learn → apply → test → iterate

That loop can happen in minutes. And I think that is one of the genuinely significant changes these tools create for experienced professionals. I don’t need to stop the project, disappear into documentation for half a day, take a course and then return later hoping I can remember enough to continue. I can learn what I need in the context of the actual problem and immediately see whether my understanding works.

That does not make the underlying knowledge less important.

If anything, it makes understanding more important, because I need to be able to judge whether the answer I’m getting is sensible.

There is a trade-off

The downside is that the cycle can now be so fast that the finer details don’t always embed first time.
I might understand what I’m doing and why, but later need reminding where a particular setting lives or exactly how I completed one step.

I noticed this recently in the web version of Power BI. I needed to change the formatting for a measure.
Once I had found the correct route, the job was straightforward. Later that day, I needed the same option again and couldn’t immediately remember exactly where Microsoft had put it.

My first instinct was:

“Shouldn’t I know this now?”

But I’m increasingly convinced that is the wrong test. There is a difference between knowledge I need to understand and knowledge I simply need to be able to retrieve.

What do I actually need to remember?

For me, the important questions are things like:

• Why am I doing this?
• When should I use this technique?
• What does the result mean?
• How do I recognise when something is wrong?
• How does this fit into the wider project?

Those are the principles I want to retain.

The exact location of a setting in an interface? I’m comfortable looking that up again.

We have always done this.

Before LLMs, it might have meant:

• searching Google
• reading documentation
• checking notes
• asking a colleague
• finding an old Stack Overflow answer
• digging through a previous project

LLMs have simply made that retrieval dramatically faster.
The difference is that the interruption to the work can now be much smaller.

AI is also unlocking work that had stagnated

There is another side to this that I did not expect.
AI is not only helping me learn faster.
It is helping me move projects that had become stuck.
Over one weekend, I revisited parts of my personal website that had barely moved for years.
The problem wasn’t that I didn’t have the skills to work on it.
I had too many things I wanted to change, too many possible starting points, and no real momentum.
So I turned the website into a problem to work through with AI.

What should I prioritise?
What should I tackle first?
What could this layout look like?
What was actually worth fixing?

Some suggestions were wrong.

Some layouts gave me something I immediately knew I didn’t like.

But even that was useful.

Having something concrete to react to triggered my own judgement and design instincts.
The blank page was gone. Instead of trying to generate the perfect answer from nothing, I had something I could criticise, refine or reject. Before I knew it, the list was getting shorter and the site was much closer to where I had wanted it to be for years.

That was a bigger win than simply saying:

“AI saved me time.”

It created momentum.

Sometimes the wrong answer is still useful

This is something I’m finding particularly interesting in creative work. AI does not always need to produce the correct answer to be valuable. Sometimes its job is simply to turn an abstract problem into something concrete enough for me to respond to.

A suggested layout might be wrong.
A piece of copy might not sound like me.
A plan might prioritise the wrong thing.

But once I can see it, my own experience kicks in.
I can say:
“That’s not right.”
“This needs more emphasis.”
“That section should move.”
“I would never phrase it like that.”

That reaction is useful.

For years, designers have used sketches, mood boards, wireframes and rough concepts for exactly the same reason. You do not always need the first idea to be good. You need something that gets the thinking moving.

AI does not remove the need for expertise

The more I use these tools, the less convinced I am by the idea that AI simply replaces expertise.
In many situations, the opposite seems true. The better I understand the problem, the more useful the tool becomes.

I can recognise when an answer is wrong.
I can refine the prompt because I understand what information is missing.
I can reject features I do not need.
I can visualise the desired outcome before it exists.
I can judge whether the final result actually solves the problem.

That is particularly noticeable when working with code.

I may not have hard-coded something in years, but if I understand what I want the interface to do, how the data is structured and what the user experience should feel like, AI can help me bridge the implementation gap far faster than starting from scratch.

The tool accelerates execution.

Experience still provides direction.

The strange feeling of being ahead and behind at the same time

There is one psychological effect I keep noticing.
The faster I learn what is possible, the more aware I become of everything I have not explored yet.
Agents.
Automation.
Coding.
Data analysis.
Content generation.
Robotics.
Autonomous systems.

Every new capability seems to reveal another layer underneath it. So strangely, AI can make me feel both further ahead and further behind at the same time. I can do things today that would have taken me far longer only a few years ago. At the same time, the horizon of what is possible keeps expanding.
I suspect a lot of professionals are experiencing some version of this. The temptation is to try to learn everything. I don’t think that is realistic.

Maybe the advantage is not knowing everything

The more useful question may be:

How quickly can I learn what matters, apply it and adapt when the tools change?

That feels much more achievable.
I do not need to memorise every interface.
I do not need to become an expert in every new model.
I do not need to chase every new AI capability.

But I do want to keep improving my ability to:

• recognise useful opportunities
• understand the underlying principle
• apply new knowledge to real work
• judge whether the result is good
• retrieve the finer detail when I need it
• keep moving rather than getting stuck

That is probably the part of AI I find most exciting professionally.
It does not remove the need for experience, judgement or understanding.
It can dramatically shorten the distance between identifying a problem, learning what you need, testing an approach and actually doing something with it.
Maybe the competitive advantage isn’t knowing everything.
Maybe it’s becoming better at learning, applying and adapting while the tools around us continue to change.