AI Adoption Is More Than Handing Out Licences
Buying AI tools creates access, but it does not change how work gets done on its own. Value appears when teams identify concrete problems, redesign processes and prepare people to use AI with judgement.
Companies are investing more and more in AI.
Perhaps they are investing too much in the technology and too little in what is needed for it to produce results.
A KPMG survey in the United States shows an interesting imbalance: executives are almost twice as likely to increase investment in new technology as they are to invest in worker training.
That helps explain an increasingly visible paradox.
Companies buy ChatGPT, Copilot, agents and other AI tools.
Then, many people continue to use them to:
• summarise meetings
• improve emails
• create presentations
• research information
• ask what present they should buy for Secret Santa
All useful. But far from AI’s potential.
There is another risk: when AI is introduced only because it is fashionable, or without a concrete problem to solve, the company may see no relevant impact and conclude that it does not work.
I do not argue that every process benefits from AI. But that conclusion cannot be drawn when what has been assessed is merely making a tool available, without concrete objectives.
Adopting AI is not simply making a tool available.
There are earlier questions:
Where can AI really create value in this company?
Which processes or tasks can be redesigned, automated or simply eliminated?
And what solution makes sense in each case: an existing tool, an automation or a small custom-built application?
AI adoption therefore stops being only a technology issue. It also becomes a question of organisation and work design.
Training becomes part of the solution, not an isolated session.
It should not serve only to teach prompts or show tools.
It should help people understand:
• which tasks they can delegate to AI
• how to improve workflows
• when automation makes sense
• how to build small tools, analyses or dashboards
• how to validate what AI produces
The sequence cannot be:
buy technology → distribute licences → wait for productivity
It needs to look more like this:
identify opportunities → redesign work → prepare people → measure results → improve
There is no need to start with the whole company at the same time.
It makes more sense to start with departments that have repetitive work, where results are easier to measure.
Not because the rest does not matter. But because this makes it simpler to understand what worked, correct what failed and only then expand.
People often ask me: “What is the best AI? What should I buy for my company?”
And I always give that annoying answer that does not seem to answer anything: “it depends”.
It depends on the problem they want to solve and on the change they are willing to make in their work.