AI adopted everywhere, invisible in the accounts

Nearly seven in ten companies use AI. Almost none see the effect in their financials. That is the finding of a survey of nearly 6,000 executives at US, UK, German and Australian companies, conducted from November 2025 to January 2026 and published by the NBER as Firm Data on AI. In its March 2026 version, the study sums it up this way: “nine-in-ten reporting no impact on employment or productivity.” In other words, nine in ten executives see no effect of AI on their company’s employment or productivity over the past three years (PDF of the study). On October 19, the Institut Xenia, a new independent research center focused on AI adoption in European small, midsize and mid-market companies, made it the starting point of its first note, Le dilemme du dirigeant diligent (the diligent leader’s dilemma) (launch announcement). One limit to keep in mind: these are statements by executives, not audited accounts, and the sample includes no French companies.

What does this survey really measure?

It measures what executives can see from the corner office. That is its strength and its weakness. Its strength, because a CEO knows whether headcount has moved and whether revenue per employee has gone up. Its weakness, because a diffuse gain, a few hours freed up here and there, never reaches the CEO as a number.

The detail tells you more than the headline. According to the study, 69% of companies actively use AI. More than two-thirds of executives use it regularly, but for an average of an hour and a half a week, and 28% do not use it at all in their working week. More than 90% see no effect on headcount, and 89% see no effect on productivity measured as sales per employee.

Yet the same executives expect a lot from the next three years: productivity up 1.4%, output up 0.8%, headcount down 0.7%. The employees surveyed, for their part, expect employment to rise by 0.5%. So this is not a tool people consider useless. It is a tool from which they expect everything, later.

Why doesn’t an adopted tool show up in the accounts?

Because “using AI” doesn’t mean much. The European Central Bank’s survey of euro area companies, conducted in late 2025, makes the point clearly: 38% of companies are at an advanced stage, with significant or moderate use, and 33% are at a very occasional or experimental stage. Add the two together and you get a flattering adoption rate. Separate them and you see that a third of the companies “using AI” are mostly running trials.

A trial doesn’t change a process. It changes the way one person does one task, sometimes. The gain then stays with that person: an email drafted faster, a summary produced in ten minutes. No one reorganizes the department around the time saved, so no one finds it in a metric.

I ran a network of 38 training centers. I know how an innovation can be adopted everywhere on paper and change nothing in a center’s margin. The question was never whether the tool was installed. It was who had the authority to change the way people worked once the tool was in place.

Where does the value no one measures sit?

The Institut Xenia note puts forward a useful thesis: the value lies in the “last mile,” meaning the adaptation of AI to the data, procedures and exceptions specific to each company. A generic model can write. It doesn’t know how your customer service team handles a dispute with a distributor, or why a particular quote goes through two approvals.

The European data point the same way. A European Investment Bank working paper published in January 2026, covering more than 12,000 companies, estimates that AI adoption raises labor productivity by 4%, without reducing employment in the short term. The gain is concentrated in midsize and large companies, and the EIB stresses one point: it requires complementary investment in software, data and team training.

In other words, AI pays off when the company invests around it. When the company just hands out licenses, it gets licenses.

Why do good management instincts slow things down?

This is the heart of the Xenia note, and the title says it: the diligent leader’s dilemma. The instincts of sound management are well known. Wait for proof before investing. Limit risk. Centralize control. Applied to AI, they produce exactly what the NBER survey describes: broad, cautious adoption, with no effect.

It is not a question of character. No one in this story is timid or backward. It is a question of decision architecture. If proof is required before the project, and proof can only come from the project, the decision goes around in circles. If every use has to be approved at headquarters, the teams that know the operational exceptions build nothing. The CEO did the job. The decision system is what blocked the value.

The note offers a formula worth remembering: centralize the foundation, decentralize the building. The foundation is the shared data and the connection of tools to the company’s systems. The building is the use cases, designed by the people who do the work.

What can the CEO of a small, midsize or mid-market company do this quarter?

First, separate the two numbers everyone confuses. How many employees use AI, and how many processes have changed because of it. The first is counted in licenses. The second is counted in tasks eliminated, lead times shortened, margin. If you only know the first, you are probably in the 89%.

Next, choose one process, just one, where the gain will show up in a metric you already track: quote turnaround time, follow-up rate, month-end close time. Set the metric before you start, not after.

Then decide who decides. Name the person who has the authority to change that process once the tool is in place, with a budget and a deadline. Without that authority, the tool stays a trial.

Finally, look at the support budget. If the “training and data” line is empty next to the “licenses” line, the EIB figures make it fairly clear what is likely to happen.

The executives surveyed by the NBER expect the effect within the next three years. It may come. But it will not come from the calendar: it will come from the organizational decisions made between now and then.

Frequently asked questions