There is a paradox in the middle of most AI programmes. The volume of AI in the organisation grows every year, while the effect of AI on the company’s performance does not grow with it.
It happens quietly, and for an understandable reason: adoption spreads out. One unit automates its reporting. Another pilots something in customer service. A third uses generative AI in marketing. A fourth builds forecasting models. Individual employees use general-purpose assistants in their own work, without anyone asking them to.
Each of these can be entirely sensible on its own. The whole can still be strategically incoherent. And when it is, the organisation has no answer to the questions that matter: which use cases are the important ones, where should we invest more, what should we scale, and what should we stop.
The amount of AI grows. The organisation’s ability to make choices about AI does not.
The measurement trap
The reason this goes unnoticed is that the easy things to measure are the things that move fastest. How many pilots are running. How many use cases have been identified. How many employees use an AI tool. How many training sessions have been held.
These numbers rise reliably, and they feel like progress. What they measure is activity.
The harder questions are the ones that carry the answer: what in the company’s operations has actually changed, which process now works materially better, where has decision-making improved, how has the value the customer experiences changed, and what can the company now do that it could not do before.
What this means for a board
The distinction matters most in the boardroom, because the wrong question produces a comfortable answer.
“How many AI projects do we have running?” is a question about activity. “Which of these could materially affect our performance or our competitive position?” is a question about prioritisation. “What evidence do we have that the impact is starting to materialise?” is a question about management.
The first sign of strategic AI is not a large volume of activity. It is the ability to tell meaningful activity apart from the rest.
Not everything has to be strategic
This needs one important boundary, because the argument is easily overcorrected.
A company should not try to make a strategic initiative out of every use of AI. An employee condensing text, finance automating a single step, marketing drafting content faster, customer service searching information — these can be very sensible uses. They do not need to change the company’s competitive position, and it is no failure that they don’t.
The problem starts only when a large number of such use cases is presented as evidence that the company has an AI strategy.
The job of a strategy is not to make everything strategic. It is to separate what matters from what does not.
This piece draws on Strateginen tekoäly — hallituksen ja johdon kysymykset (Aamu Editions, 2026).