Businesses have embraced AI. But is AI anxiety the reason it isn't showing up in the financial results?

Most businesses know they need to get started with AI, and many have already given their employees access to the tools. Yet making the technology matter in day-to-day work often takes longer than expected, and in my experience part of the explanation lies in something very human: how we react to uncertainty. It rarely looks like resistance. More often it shows up as good, sensible reasons to postpone the decision.

McKinsey's 2026 global survey (opens in a new window) shows that 89 per cent of responding organisations use AI regularly in at least one business function. At the same time, 44 per cent say AI is now being scaled across the organisation, up from 38 per cent the year before.

That says something important about where many businesses find themselves right now: the main problem is no longer getting access to AI, but getting the organisation to use it in a way that creates real value.

A psychological layer we may be overlooking

Part of the answer may be technological, and part of it is about strategy, processes and skills. But there is also a psychological layer that, in my experience, deserves far more attention if businesses want to accelerate.

When technology challenges our expertise

AI doesn't only affect how we work. It can also affect how we experience our own expertise, control and worth.

If a technology can suddenly write a first draft, analyse vast amounts of information, develop code, summarise complex documents or come up with qualified suggestions in seconds, it changes our view of what requires human expertise.

That can be fascinating, but also unsettling. What does it mean for me? Which skills become more important, and which become less valuable?

This isn't necessarily resistance to technology. It can just as well be a natural reaction to a situation where the rules of the game are not yet clear. Researchers talk about (opens in a new window) AI anxiety, and although the term probably doesn't resonate with many people yet, the interesting thing is that anxiety doesn't necessarily feel like anxiety. It can also show up as caution, a need for more information, or a wish to wait a little longer.

“We just need some clear guidelines first, once we've understood the possibilities and the technology better, once we've seen what others are doing, and once the solutions are free of teething problems.”

Statements like these can be perfectly rational, but they can also be a way of coping with uncertainty. Research on anxiety and decision-making (opens in a new window) suggests that anxiety can make us focus more on potential negative outcomes, interpret ambiguity more negatively and be more inclined to avoid situations where something could go wrong.

A recent article in *Organizational Dynamics* (opens in a new window) identifies two recurring forms of anxiety when employees encounter AI:

1. the fear of making mistakes with the systems
2. the fear of being replaced by them

The article links both concerns to psychological safety.

Our experience from working with clients points the same way: the employee who doesn't take up AI doesn't necessarily lack motivation. More often, the employee is trying to protect their professional expertise, or to avoid making mistakes with a technology they don't yet feel comfortable with.

Access to AI is not the same as adoption

An employee can have access to ChatGPT, Copilot or other AI tools without using them systematically, and a company can have an ambitious AI strategy without AI having moved anything significant in the business. It is therefore useful to distinguish between technical implementation, transformation and adoption.

• Technical implementation means the technology has been made available to employees.
• Transformation is something bigger. The organisation changes how it works and creates value, because AI becomes part of the way of working itself, or of the business model.
• Adoption is when we see people take the technology on board and integrate it into their work, so that the way they work actually changes.

When usage doesn't automatically create value

McKinsey's figures illustrate the gap between using AI and real business value. A full 80 per cent of those who use AI report better individual productivity, but only 37 per cent can point to an effect on the company's operating profit.

We should therefore be careful about concluding that an organisation is far along with AI simply because many employees use it. Instead, we should ask:

“What has changed in our employees' work because they use AI?”

Frameworks are necessary, but not sufficient

When new technology creates uncertainty, leaders' first reaction is often to try to reduce it with guidelines, training and governance, a selection of approved tools and clear rules for how they should be used.

That isn't superfluous, because clear frameworks and governance are essential for using AI responsibly. But our focus on frameworks and control can also give us a blind spot.

AI cannot be fully implemented through rules and instructions, because much of the value only emerges when people experiment with the technology and discover new uses that no guidance could have foreseen.

That places different demands on leadership, which must support curiosity, make room for experimentation and build psychological safety. Without it, people won't dare to try things out and learn from what doesn't work, when nobody has all the answers yet.

Employees often know where AI can make a difference

AI adoption shouldn't necessarily be a top-down exercise either. In fact, it lends itself to a bottom-up approach. An analysis from MIT Sloan (opens in a new window), based on more than 50 interviews with business experts, developers, trade unions, public authorities and researchers, among others, highlights that employees should be involved in defining the problems, developing the workflows, testing the technology and identifying where AI can actually create value.

The employee who works with a particular process every day knows about complexity, challenges and dependencies that neither the executive team nor the IT department can see.

My hypothesis is that letting each business area develop its own AI strategy, with leadership limiting itself to setting the framework from above, will have a big impact. What we see is that when employees start using AI together, across departments too, changes emerge that really show up in the financial results. And that requires open dialogue.

So perhaps the question should be:

“How do we get our employees involved in discovering where AI can make our business better?”

From fear to curiosity

As a leader in several different organisations, I have seen what a difference it makes whether a change is communicated as something the organisation has to implement, or as something we explore and shape together.

My view is that this applies even more to AI, because we don't know all the consequences, and because AI is not a traditional off-the-shelf system. That brings an inherent uncertainty that doesn't go away, however long we wait to get started.

Leadership's task is therefore not to have all the answers, but to create a framework and a culture where concern can be turned into curiosity and action, and where it is legitimate to experiment and learn by doing.

It may require more leadership than before, but above all it demands more of the leader's own capacity to drive innovation, set direction, prioritise and lead from the front. That is hard, and developing your own leadership skills often calls for new self-insight.

AI is already changing work

It is also worth remembering that we are no longer on the threshold of this development. A recent study based on Danish data (opens in a new window) shows that AI chatbots have already been widely adopted in the job functions most exposed to the technology. The researchers also find new AI-related tasks in areas such as content production, AI monitoring and AI integration, but as yet no major effects on wages or recorded working hours.

Danmarks Nationalbank's latest analysis (opens in a new window) shows signs of lower employment growth in companies that implement AI, but as yet no major changes in overall employment. The picture is more nuanced than a simple story of either job losses or efficiency gains, because what is changing is rather the content of the work.

That applies to us at Capacit too. We see that AI is changing not only technical skills, but also the demands on professional judgement, problem-solving, creativity and the ability to work with the technology.

In practice, this means that in many of our roles, employees now write far less code themselves and instead increasingly need to assess, challenge and quality-check what AI produces. It also means we have to keep revisiting and adjusting our own business model, because the value of what we deliver to our clients shifts in step with the technology.

AI adoption thus becomes, to a large extent, a matter of learning a new way to work and think.

What does this mean for the executive team?

If AI anxiety is part of the explanation for the slow move from technical implementation to real adoption, it isn't enough to place the responsibility on employees.

It is a leadership task to create the conditions that make it possible to move from uncertainty to learning, and from learning to use. But it also means leadership must accept that not every answer can be defined in advance.

The executive team should be clear about why the company is working with AI, which business problems it wants to solve, and how employees will be involved in finding the answers.

Perhaps one of the most fundamental things about AI is that we don't just have to learn to use a new technology; we also have to learn to think differently about our work. If progress isn't as fast as you'd like, ask yourselves: