On 25 November 2025, ABN AMRO chief executive Marguerite Bérard announced that the bank would cut more than 5,200 full-time positions by 2028. That is more than a fifth of the workforce. Customer services, operations, administrative teams and the departments responsible for anti-money-laundering controls are expected to shrink by around 35 per cent. AI, Bérard stated, would take over a large part of the manual work.

The Amsterdam stock exchange closed higher that day. The trade unions called it a shockwave.

That contrast says something. Not about who is right, but about the fact that the same news reaches two completely different realities. One reality is about returns, cost ratios and market positions. The other is about people coming home in the evening and telling their partners that their job is disappearing. And that AI is the reason.

But is AI truly the reason?

That is the question I want to ask, not to be cynical about ABN AMRO’s choices, but because the answer genuinely matters. For the people losing their jobs. For the people who remain. And for leaders who want to be more than a conduit for a strategy that was already decided.

Bérard wants the cost-to-income ratio to fall to 55 cents for every euro earned. She wants ABN AMRO to grow into one of the five largest private banks in Europe. Those are legitimate strategic ambitions. But ambitions are not evidence that AI can fulfil them.

The research data say something different from the press release. MIT research from 2025, based on 150 interviews, a survey of 350 employees and the analysis of 300 public AI deployments, found that only around five per cent of AI pilot programmes succeed in achieving rapid revenue growth. An analysis by the RAND Corporation from the same year found that more than 80 per cent of AI projects fail to deliver what was expected of them. Gartner found that only 28 per cent of AI use cases in infrastructure and operations fully succeed and meet ROI expectations. In 2025, 42 per cent of organisations in North America and Europe abandoned the majority of their AI initiatives, twice as many as the year before.

These are not the figures of a sceptic. They are the findings of the most widely consulted research institutions in this field.

On one side, a bank saying that AI is taking over thousands of jobs. On the other, research saying that most AI projects are nowhere near delivering what they promise.

That gap is not a minor detail. It is the heart of what makes this question so difficult for leaders who want to be honest.

Because two stories are possible. The first is that ABN AMRO is genuinely undergoing a far-reaching technological transformation, in which AI takes over tasks currently performed by people, and that the headcount reduction is the honest consequence of that change. The second is that the bank wants to reduce its costs to become more profitable, that AI provides the narrative that makes the choice sound less like a choice, and that the technology is currently nowhere near doing what is being claimed.

Both stories can be true at the same time. They probably are. And that is precisely the point.

CNV negotiator Arthur Bot described the reorganisation at its announcement as “a mass redundancy, driven primarily by profit growth.” Bjorn Cumps, professor at the Vlerick Management School, put it carefully in the same week: “But it remains to be seen whether ABN AMRO can actually deliver this.” Neither remark is anti-technology. They are precise. They ask for a distinction that is absent from the bank’s own communications: the distinction between what AI can already do and what executives hope it will eventually be able to do.

When that distinction is not made, something emerges that I call an AI alibi. The technology becomes the reason for decisions that were already taken for other reasons. It softens the visible accountability of the board, but it amplifies the existential uncertainty felt by employees. If the cause lies beyond everyone’s reach, in the spirit of the age, in the inevitable march of algorithms, where is influence still possible? Who is then accountable?

Uncertainty about your job is hard enough. Uncertainty about who is responsible is unbearable.

Let me go a step further, because I think a structural pattern is becoming visible that reaches well beyond ABN AMRO alone. In sectors under pressure you see a retreating movement: more control, more risk aversion, more emphasis on measurable outcomes. AI fits that pattern well. Everything becomes data, models and dashboards. The softer questions about meaningful work, loyalty and trust recede into the background. Not because they are unimportant, but because they are hard to quantify and therefore easy to leave off the strategic agenda.

That is a problem. Not in a moral sense, but in a practical one. Because the employee who loses their job to a system that is not yet fully working, who then hears that they will be well supported in finding other work, who then watches their former colleagues do twice the work with half the team, draws conclusions. About what the organisation truly values. About whether they would ever again go the extra mile for it. Those conclusions are difficult to reverse, even if the technology eventually delivers what was promised.

There is also a symbolic dimension to all of this that cannot go unnamed. AI is not only a set of models and systems. It is also a signal. A board that says “we are future-proof” and points to AI to demonstrate it shows that it dares to digitalise, dares to renew, dares to cut. On financial markets, that is read as decisiveness. Within the organisation, it is also read as distance. As the sense that the people who have spent years managing customer relationships, processing files and keeping processes running have been traded in for a promise not yet redeemed.

That does something to loyalty. To engagement. To the willingness to go the extra mile for an organisation that has just told you that you can be replaced by software.

And it is precisely here that the task lies for leaders who want to be more than executors of a reorganisation plan.

What that concretely requires, first and foremost, is honesty about the nature of the choice. Make an explicit distinction as a leader between cost reduction, strategic reorientation and genuine technology-driven change. If the primary driver is the cost-to-income ratio, say so. If AI will eventually take over tasks but the technology is not yet fully proven, say that too. That takes courage, because it places the accountability visibly back with the board rather than with an algorithm.

Connect the technology to the concrete work. Which tasks are disappearing precisely? Which are shifting? What new forms of expertise are needed? What changes in the dignity of the work that remains? These questions are not soft. They are strategic. Organisations that successfully integrate AI, McKinsey research shows, are twice as likely to have redesigned end-to-end work processes before selecting technology. The technology follows the intention. Not the other way around.

And use the reorganisation as an occasion for a serious conversation about culture. What kind of organisation does ABN AMRO want to be when more work is done by systems? What role will judgement, dissent and the capacity to draw moral boundaries play? What is the value of the employee who notices something the model has missed? Those questions disappear from view when technology is presented as the self-evident answer.

There are also things better avoided. Do not use AI as a label for conventional cost savings. Do not continue with experiments without clear purpose, accountability and evaluation. And avoid the assumption that acquiring technology is the same as integrating it into daily work. Gartner reports that 85 per cent of AI projects fail due to poor data quality or the absence of relevant data. The infrastructure, the processes and the governance need to be in order before large-scale rollout makes sense. Without that foundation, an organisation primarily increases its complexity and pressure, without lasting value to show for it.

Real change does not happen in systems. It happens in roles, relationships and routines.

There is a risk I want to name because it is rarely said out loud. When employees experience technology being used as an excuse for decisions that were already made, trust erodes quickly. Not as a one-off reaction to bad news, but as structural damage to the relationship between organisation and people. Research on psychological safety, including the work of Amy Edmondson, shows that employees who do not feel safe to voice their judgement stop signalling problems. In an environment that relies increasingly on AI systems, that is particularly dangerous. The employee who sees that the model is missing something needs to be able to say so.

There is also the risk of erosion of professional identity, a risk rarely named in reorganisation plans but one that weighs heavily in practice. When support professionals hear primarily that their work will eventually be taken over by systems, their engagement falls before the technology has even started working. Not because they are unwilling to cooperate, but because the narrative tells them that their expertise, their experience and their judgement are interchangeable. That does something to people. It affects the energy with which they do their work in the months and years that follow. Organisations that underestimate this pay a price that does not appear in the reorganisation plan but does appear in the results.

There is also the risk of overstated promises. If the organisation is counting on savings that are not realised in practice, a gap opens in the narrative. The promise of AI then becomes a source of disappointment, for executives, for shareholders and for the employees who were made redundant for the sake of technology that did not yet do what was claimed.

A story larger than reality is eventually caught by reality.

What all of this asks of leaders is not complicated in theory, but demanding in practice. It asks the willingness to say: this decision is ours. Not the spirit of the age’s. Not the algorithm’s. Ours. With all the consequences that entails.

That is not weakness. It is precisely the strength people expect from leaders at the moment their world tilts.

ABN AMRO had already cut around 1,500 jobs in 2025 by the time the bank reported in February 2026. That is nearly 30 per cent of the planned reduction in a single year. The reorganisation is underway. The question is no longer whether it will proceed. The question is whether the people caught up in it, and the people left behind, will meet a leader who tells them with open eyes exactly why. And who, when the answer does not fully hold up, has the courage to say so.

Notes for those who wish to read further:

1. MIT Project NANDA, The GenAI Divide (2025). Research based on 150 interviews, a survey of 350 employees and the analysis of 300 public AI deployments. Finding: only around five per cent of generative AI pilot programmes achieve measurable revenue growth.
2. RAND Corporation, AI Project Outcomes Analysis (2025). Analysis of AI implementations in large organisations. Finding: more than 80 per cent of AI projects fail to deliver their intended business value.
3. Gartner, AI Use Cases in Infrastructure and Operations (2026). Survey of 782 infrastructure and operations leaders. Finding: only 28 per cent of AI use cases fully succeed and meet expected ROI.
4. Amy C. Edmondson, The Fearless Organization (2018, Wiley). On psychological safety as a precondition for learning organisations, and why people stop signalling problems when they do not feel heard.
5. Ajay Agrawal, Joshua Gans & Avi Goldfarb, Power and Prediction (2022, Harvard Business Review Press). On how AI changes the nature of decision-making, and why the organisational consequences outweigh the technological ones.