AI Business Strategy

Everyone uses AI. Few organisations profit from it.

By Sam Solaimani, Professor of Digital Technology, Innovation & Operations Management at Nyenrode Business University

AI is developing fast, and most organisations now use it in some way. People who work with it are usually positive. They write faster, find things sooner and have a first draft in minutes. 

The organisation as a whole often sees much less. Costs do not go down, customers do not notice a difference, and the results stay the same. Study after study this year shows this gap between what employees gain and what the organisation gains. 

Why is that, and what can leaders do about it? 

Where AI fits and where it does not 

AI can be right for every business, but not for every problem inside a business. That is the honest starting point. The technology is powerful, but power is not the same as fit. 

It helps to understand that AI has what researchers call a jagged frontier. On some tasks it is remarkable, well beyond what most people expect. On others that look almost identical, it is surprisingly poor, and the line between the two is not obvious from the outside. You cannot tell in advance, from a brochure or a demo, exactly where AI will help you and where it will let you down. 

In practice the value is specific. AI tends to do well where there is repetition, a lot of data, and some tolerance for the occasional mistake. Think of summarising documents, drafting first versions, sorting requests or spotting patterns a person would miss. Used in the right place, it frees people for the parts of the job that need a human. 

The mistake is treating AI as a strategy in itself. It is a tool that supports a strategy. The businesses that get value start from something they need to fix and ask whether AI helps. The ones that struggle start with the technology and go looking for somewhere to put it. 

AI for the sake of AI 

A lot of adoption is driven by fear of missing out rather than real need. The boardroom asks what the company is doing about AI, someone has to answer, and projects get started so there is something to point to. 

We see this in our own research. In the AI Trend Research 2026 by Berenschot and Nyenrode Business University, with 379 respondents, just over half say they feel pressure to use AI. We also see more ad hoc use of AI and less use that is based on an AI policy. So organisations do more with AI, but they steer it less. 

That is the wrong reason to start. When AI is bought as a top-down statement rather than aimed at a real task, nobody has defined what success looks like, so nobody can tell whether it worked. The pilot gets praised, the budget gets spent, and six months later the process runs exactly as it did before. 

The same is now happening with AI agents. These are systems that can carry out several steps of a task by themselves. Gartner expects that more than 40% of agent projects will be stopped by the end of 2027, because they cost too much, deliver too little or carry too much risk. Gartner also thinks that only about 130 of the thousands of suppliers really offer agents. The others sell existing software under a new name. 

Start with the problem, and know what it costs 

So my advice is to set technology aside for a moment and start with the problem. What is slow, expensive or error-prone today, and what would it be worth to fix? Describe the problem in plain terms and put a number on it. How much time does it take, how often does it go wrong, what does it cost?  

That conversation filters out most of the bad ideas. Quite often the honest answer is that a better process, cleaner data or simpler software would solve it at a fraction of the cost. A good adviser, inside or outside the organisation, will tell you so. 

It also helps to look beyond saving costs. If you only ask what is slow or expensive, you end up doing the old process a bit cheaper. Organisations that get real value from AI often go a step further. They ask how the work itself should be organised now that AI is available. They also use AI to improve quality and to grow. 

There is one more question that few organisations ask. What does the AI cost each time we use it? Until recently you paid a fixed price per user. More and more, you pay for use. Every question and every answer is counted in so-called tokens, and AI agents use many more tokens than a simple chat. Uber reportedly spent its 2026 budget for AI coding tools in four months.  

Many companies cannot see during the month what their AI is costing them. 

A high bill is not always bad news. It can mean that people use AI a lot because it works. But you only know that if you can compare the cost of handling one case with what that case is worth. 

What leaders should talk about 

Less about the technology and more about the conditions for success. A few topics matter most. 

Start with the problem and the measure of success. If nobody can say what good looks like in numbers, the project is not ready. I would be careful with plans that count on fewer jobs. So far, organisations expect far more job cuts from AI than they actually carry out. 

Talk about data honestly. AI is only as good as the information it works from. If the data is messy, scattered or out of date, sorting it out is usually the larger part of the work. 

Talk about people. The technology rarely fails on its own. Projects fail because staff were not involved, not trained or not convinced. Pushing a tool at staff and hoping does not work. People change how they work when they are involved, supported and shown what is in it for them. That also asks something of leaders. They have to use AI themselves and show that they stand behind the change. 

And talk honestly about what the technology still cannot do. These systems still give confident answers that are simply wrong. There are real questions about security and privacy when company information goes into tools the business does not control. There are questions about bias when AI is used in decisions about people, and about who is responsible when something goes wrong. For every use, decide where a person needs to check the result. 

The headline point is simple. Digital transformation is more about transformation than about digital, and the same is true of AI. The technology is the easy part. The value comes from the problem you choose, the data you feed it, the people who have to use it, and how you organise the work and the responsibilities around it.  

Organisations that start with the problem, know what AI costs them and help their people to change will get far more out of AI than those that buy it as an end in itself. 

Related Articles

Back to top button