More than once, I have seen transformation plans stall for the same reason: the human factor had been neglected. The story that follows draws on several of them.
It concerns a transformation that looked purely technological: AI. And yet the company believed it could do without bringing its people along. It paid dearly for it.
An ambitious project… and top secret
A large engineering company had launched an AI project focused on its operations. The goal: to automate a large share of its technical studies.
The engineers whose jobs were about to change had not been involved. “Deliberately kept out” would be closer to the truth. Management feared they would dig in their heels, or leave before the tool was ready. The project would therefore be run from headquarters, in the strictest confidence.
It leaked, of course. Neglected people quickly turn into first-class opponents.
Act I: when fear breeds distrust
A few hypotheses needed to be tested in the field. So a headquarters delegation travelled to two subsidiaries. Secrecy being paramount, it met only the head of operations and the IT manager, without telling them what it was about. Officially, this was an international benchmark.
Unfortunately, every relationship obeys a principle of balance. Build it to your sole advantage, and the other party will find a way to restore the balance. Only you will not be the one setting the terms. The local managers chose loyalty to their teams over loyalty to a headquarters that did not trust them. They talked about the meeting, the odd questions, the unclear purpose. Within days, the whole subsidiary was talking about the secret project. No one knew what it was for, and the unknown is by nature a potential enemy. The toothpaste was out of the tube…
Act II: when lack of attention looks like ill intent
A second decision made matters worse: AI investment was reserved for the central engineering department. Yet a proposal was on the table: to train all employees in AI, starting by equipping them with the advanced version of an off-the-shelf AI assistant. There was no shortage of arguments against it: a tight budget, country teams set to shrink, a return on investment that was hard to demonstrate.
Let’s do the maths anyway. Take a company with 10,000 employees. The measure costs around €4 million, the payroll of about forty positions. A productivity gain of 0.4% is therefore enough for it to pay for itself within the year. In Europe, productivity grows by roughly 1% a year over the long term: the effort is minimal.
But beyond the missed opportunity, the decision destroyed a major source of value: employee engagement. It amounted to a double denial of intelligence. Denial, first, of their ability to learn. The average age in large European companies is around 42: these employees still have more than twenty years of working life ahead of them. Twenty years without AI training? That is a losing bet every time. Denial, second, of their ability to read the situation… as if, kept in the dark, they would follow along without seeing or understanding anything. Yet one message was understood immediately: AI is headquarters’ business, and the company has no intention of letting you play a part in it.
Act III: the system always finds its balance
Deprived of official tools, some employees did not wait for headquarters to make up its mind. They used their own tools, sometimes on personal accounts, exposing the company to risk every day despite its firewalls. The project that management wanted to keep under wraps produced exactly what it feared: uses of AI beyond any control. And for many talented people, especially the youngest, a move to more switched-on companies, no pun intended!
When the law restores the balance… without repairing anything
Yet the training that was set aside is a legal obligation. Since February 2025, Article 4 of the AI Act has required every company whose staff use AI (and they do, whether it was planned or not) to develop their AI literacy. The Digital Omnibus regulation has softened the wording without removing the obligation, and enforcement began in August 2026.
When the employer fails to provide it, the legislator steps in to restore the balance. But the law repairs nothing, neither loyalty nor engagement. Employees know very well the difference between what they are granted because the law requires it and what the company chooses to give them because it cares about them. The first is received as an entitlement. The second creates a debt… of gratitude.
Why neglecting the human factor costs more with AI
According to BCG (October 2024), 74% of companies struggle to achieve tangible value from AI. Some 70% of the challenges stem from people and processes, 20% from technology, and only 10% from algorithms. Budgets, meanwhile, all too often follow the reverse order.
With AI, the bill is heavier, because the investments are colossal and because it affects skilled professions. AI also unsettles what the French sociologist Renaud Sainsaulieu called identity at work. Letting an engineer guess that a machine will tomorrow do what makes them proud today undermines that identity. It therefore becomes crucial to involve them in the change, so that they decide for themselves how the tool can help them make better use of their expertise. Otherwise, they will conclude that they are the ones who will be assisting the tool.
What gets a plan moving again
So what principles should guide an AI plan back on track? The same as for any transformation that changes the content of work:
Involving early the people whose jobs will change. A small group of recognised experts, brought together as soon as use cases are chosen, changes the whole dynamic. Their awkward questions are often the right ones. And they spread the word positively among their colleagues.
Briefing managers before sending in the consultants. Once informed, managers act as relays. Kept out of the loop, they stop doing so and become an echo chamber.
Opening the tool to everyone. Training and equipping everyone costs little compared with an AI programme. With a “train the trainers” approach, internal champions train their colleagues, and the whole company gains in pride. And the tinkerers of Act III become allies.
Treating the risk of departures with pragmatism. If it calls for a retention plan, HR will know how to build one, provided it has a seat in the programme’s governance from day one. Secrecy, for its part, accelerates the very departures it hopes to prevent.
Saying what you know… and what you do not know yet. On jobs, silence is always read in the worst possible way. You have to accept communicating as you go. “We believe 20% of working time could be freed up, but we do not yet know how roles will change. We will decide with you, and we will come back to you step by step”: teams accept uncertainty, as long as they are kept informed.
Engagement, the keystone of the AI revolution
Provided, too, that teams believe what they are told. This is prepared upstream, through attention: listening, knowing what people are going through and what worries them. That is what makes management’s intentions credible. Then comes the truth. You cannot always say everything, nor at any moment. But you can almost always say far more than you think. Telling the truth is a way of showing trust. And trust breeds trust: it is from trust that engagement is born.
And as long as it takes women and men to run companies, engagement will create value. The value of AI depends on it too, because a technology never embeds itself: it takes people who believe in it, who test it, and who change the way they work to build it into their processes. Without them, the best AI strategy stays on paper. Or, to borrow a phrase often attributed to Peter Drucker, “culture eats strategy for breakfast”.
