AI Adoption Is Not the Problem. Thinking Is.

Part 1 of a Series.

AI Adoption Is Not the Problem. Thinking Is.

Last week, I sat in on a discussion that, at first glance, looked like things were working exactly as they should.

A team had just reviewed a summary of customer feedback generated using AI, and the output was everything you would expect from a well-trained professional. It was structured, balanced, and written in a tone that felt measured and appropriate, the kind of summary that makes meetings move faster because no one feels the need to question it too much.

Naturally, everyone nodded, agreed on the key themes, and prepared to move on to the next agenda item.

But something about it stayed with me.

Not because the summary was wrong, but because it felt a little too smooth, as though some of the edges had been quietly rounded off in the process.

Earlier, I had glanced through the raw feedback, and one particular comment stood out. It was not just negative, it was frustrated in a way that usually signals something worth paying attention to. It was the kind of comment that, in a different setting, might have changed the direction of the conversation.

In the summary, it appeared as “minor dissatisfaction.”

Technically, that was not incorrect.

But it was not the same thing either, and in business, those small differences are often where decisions begin to shift.

When Nothing Looks Wrong, But Something Is

The more I reflect on moments like this, the more I realise that what we are seeing today is not really a problem of technology, nor is it a matter of people using tools incorrectly.

It is something quieter than that, and perhaps more important. The way we work has changed, but our habits around thinking have not quite caught up.

A few years ago, most of the effort in work like this would have gone into producing the first draft, whether that meant summarising feedback, building a report, or preparing slides for a meeting. There was a natural pause built into the process because creating something from scratch takes time, and that time often allowed space for reflection, clarification, and sometimes even disagreement.

Now, that first draft appears almost instantly. And when something appears complete, especially when it looks structured and well-written, we tend to treat it as complete, even if we have not fully interrogated what sits beneath it.

The Data Is Clear. The Decision Is Not.

I have seen similar patterns play out in different contexts.

In one session, a manager was reviewing a dashboard that clearly showed a decline in sales, and while the data itself was not in question, the interpretation of that data quickly became fragmented. Within minutes, different teams began attributing the decline to different causes, with marketing focusing on pricing, sales pointing to product issues, and operations questioning distribution.

The numbers were clear. The decision was not.

In another case, someone had compiled a report by pulling together data from multiple Excel sheets, and on the surface, everything looked aligned. Totals matched, categories were consistent, and there were no visible errors that would immediately raise concern.

But one of the sheets had been updated more recently than the others, which meant that while the report was internally consistent, it was not entirely reflective of the current situation.

Again, nothing looked obviously wrong.

And yet, something important was off.

These are not dramatic failures. They are small, almost invisible shifts that happen in everyday work, and precisely because they do not feel like mistakes, they are easy to overlook.

The Shift We Didn’t Notice

This is why I have started to think that the main challenge we are facing today is no longer about AI adoption. That part, for the most part, has already happened.

People are using AI, sometimes formally within approved systems, and sometimes informally in ways that simply help them get through their workload more efficiently.

The real challenge is what happens after the answer appears.

AI changes something subtle but significant in the way work is done, because it allows us to arrive at an answer before we have fully understood the question. While that can be incredibly useful, it also creates a new responsibility, one that is easy to miss if we focus only on speed and efficiency.

Before, much of the effort was spent producing output.

Now, the effort needs to shift towards evaluating it.

That shift is not always intuitive.

When something is presented clearly, written confidently, and organised neatly, it reduces our instinct to question it, because it signals completion. It gives us a sense that the thinking has already been done, when in reality, what we are looking at is often just the starting point.

The Work Has Moved

In many of the training sessions I have been part of, this is where the real difficulty begins to show.

It is not that people do not understand the tools, or that they are unable to generate outputs.

It is that they are not always equipped with a clear way of interrogating those outputs, especially when the outputs appear credible at first glance.

What seems to help is not a more complex tool, or a more advanced technique, but a simple change in habit.

A pause.

Before accepting what is in front of us, whether it comes from AI, a dashboard, or a report, it helps to ask a few quiet questions.

1. What context might be missing here?

2. What assumptions are embedded in this output?

3. What decision am I about to make based on this?

4. And what would I want to validate before acting on it?

These questions do not slow things down in a meaningful way.

If anything, they prevent the kind of rework and misalignment that often comes from acting too quickly on something that only appears complete.

Because ultimately, the goal was never just to get answers faster. The goal has always been to make better decisions.

AI does not remove the need for thinking. If anything, it makes the quality of our thinking more visible.

And that is where the real advantage will lie going forward, not with those who can generate outputs the fastest, but with those who are able to pause, question, and connect those outputs to the decisions that actually matter.

Because in the end, the tool is rarely the problem.

The thinking usually is.