Why companies want to deploy AI… and whether they really should

Most conversations about artificial intelligence start very similarly. A phone call, an email, an online meeting and one key phrase: “We want to implement AI.”

At this stage, everything looks promising. There is energy. There is interest. Sometimes there is even a preliminary budget. The company wants to be modern. It wants to move forward. It wants to make the technology “work.” But then something happens that repeats surprisingly often. All it takes is a few questions, a few hours of a workshop or one meeting on the shop floor and… suddenly the picture starts to blur.

Very often it turns out that behind the idea of implementing AI is not the desire to solve a problem, but ambition, the desire to catch up with the market, to be trendy. To show off in a meeting with the board of directors or the owner of the company. And this is where the space opens up for the topic I would like to tell you about. Learn 5 reasons why AI implementation can fail.

There’s a moment in AI discussions that I call “scratch the paint off.” On the surface, everything is right. Underneath – things are sometimes different.

When I ask:

…Silence often falls. Not because someone doesn’t want to answer. It’s just because these questions are being asked for the first time. And very quickly it comes out that artificial intelligence was chosen as a solution before anyone had named the problem well. In fact, someone wanted to implement AI, not overcome the difficulties.

Artificial intelligence does not operate in a vacuum. It does not “guess” reality. It reconstructs it based on data. But not random data:

And it is the data that is the most expensive, difficult and underestimated piece of the whole puzzle.

I remember a project where a company wanted to use AI to optimize energy consumption. The goal made sense. Electricity bills can eat up a fair amount of margin. Finding 10-20% savings is a real opportunity to save costs.

The problem was that the company had meters. “Problem? – You ask. – I guess that’s an advantage!” Apparently so, but the meter itself without context is just a source of numbers.

Artificial intelligence, therefore, “saw” energy consumption, but did not have access to information about it,

It’s a bit like trying to judge the quality of a movie based on the soundtrack alone.

In such situations, I very often encourage people to take a step back. Instead of teaching algorithms, let’s first show the reality. Let’s visualize the data. Let’s connect them. Let’s make sense of them.

Sometimes a simple energy management system can open your eyes more than the most advanced AI model. Because suddenly you can see:

Only when this picture is clear does the artificial intelligence have something to work on. Before that… it can only guess.

There is another pattern that I see very often. AI is sometimes treated as a way to leapfrog several stages of digital transformation, as a cure for every evil:

This way of thinking may seem tempting. But don’t go down that road, because you might get lost or collide with a wall. I would like to see it spoken out more often that artificial intelligence doesn’t fix chaos. It scales it back.

It also happens that after deeper analysis we come to the conclusion that AI is not needed at all. Even when we have the right data and have defined the problem well. Why? Because:

And then there is always the boundary question: do you really want to pay more for a certain solution just because it is called “AI”?

Sometimes the answer is “yes.” More often, however, the reflection is that technology is there for effect, not narrative.

The further we get into conversations about AI, the clearer one thing becomes: the biggest challenge is not the technology.

Artificial intelligence touches areas that for years have been the domain of humans: analysis, decision-making, optimization. This raises resistance, creates uncertainty and fears that cannot be dispelled with a slick presentation to the board.

That’s why I very often stop projects at the stage of asking whether the organization is ready for implementation. Because if people don’t want to use the system, even the best algorithm will become an expensive curiosity at best.

Every project, sooner or later, reaches the desk of someone who asks, how much does it cost and when will it pay for itself?

This confrontation can sometimes be like a cold shower. It is important to remember that AI is not just implementation. It’s years of maintenance. New competencies, new infrastructure, responsibilities.

If the turnaround starts to draw after five years, many projects simply do not go beyond the planning phase. And this is a rational decision.

This sentence sounds perverse, but it is very true.

Digital maturity for an organization is not about implementing every new technology. It lies in being able to consciously withdraw from it if it doesn’t make business sense. Artificial intelligence should not be the beginning of the journey, but the culmination of it.

The foundation always remains:

Without them, AI remains just a pretty and trendy buzzword.