when Meta cut 8,000 employees Last April, it explicitly framed it as an investment in artificial intelligence, joining a growing list of highly profitable companies trading headcount in computing. Coinbase followed weeks later, Cutting about 14 percent of its workforcepointing to market volatility and the spread of artificial intelligence tools at the same time.
Direction of travel understood. The economics of AI are real, the competitive pressures are real, and at some point, every company with serious AI ambitions will have to make difficult capital allocation decisions in the face of these realities.
What is less clear is whether the approach taken by some companies is wise. There is, for example, a clear division between the two sides of the Atlantic in the speed and scale of this reallocation process, as American companies in particular seem willing to move more aggressively and reduce their size deeper and faster. The bets that are offered are high. And some of the risks associated with it are really hard to predict. But there are at least two of them that are completely predictable, and neither of them gets enough attention.
Efficiency trap
The first concerns what happens to organizations when you strip them of their capabilities in the name of efficiency. The word itself does a lot of work in these ads. Because when Zuckerberg talks about efficiency in the same sentence as investing in AI, he is using a word that carries, in regulatory terms, a specific and well-documented set of consequences that are clearly not just financial. It’s not always clear that it’s about efficiency either.
For example, research on how organizations respond to efficiency programs shows that when you remove slack and compress the number of forums for alignment and discussion, it does not make these organizations more adaptable. Instead, and contrary to what is expected, it makes them more fragile. The reason is structural. What looks like waste, including people and processes that don’t seem important, is often the system’s ability to self-correct. It’s where disruptions are caught early, where new ideas find testing grounds, and where the organization senses what’s changing before it becomes visible in the numbers. By removing them, the organization becomes faster in the short term and more blind in the medium term.
The irony for companies like Meta is notable. The innovation that investment in AI is supposed to unleash depends on the kind of organizational behavior that efficiency reductions tend to suppress: experimentation, initiative, and a willingness to raise problems early and openly. This is not soft stuff. Google’s own search, a long-standing People Analytics work that defined Characteristics of effective teams-I found it Psychological safetyThe degree to which people felt it was safe to take risks and speak up was the most important variable. It sits above talent, above resources, above everything else. It is, almost by definition, the first victim of a massive redundancy program that revolves around the idea that human labor is being replaced by machine power.
Signal problem
The second risk is deeper, and relates to something I’ve spent a lot of time researching in the context of power and organizational information flows. When leaders make decisions, the quality of those decisions depends on the quality of the information that reaches them. And this flow of information is not a neutral pipe. It is constantly shaped by what people say.
The research on this matter is consistent and realistic. Subordinates speak less openly with people over whom they have authority. Bad news slowly travels to the top. Ambiguous information is simplified and cleansed as it moves through the layers of management. Leaders end up having a clearer and more coherent picture of reality than what actually exists. In my book Power trapI’ve called this the clarity gap: the distance between what senior leaders think is happening and what is actually happening below them.
Layoffs do not cause this problem. It is already present in every organization with any degree of hierarchy. But large-scale layoff programs, especially those framed the way Meta was, are dramatically accelerating it. The moment employees realize that their role may be next in line to be replaced by AI, they start calculating what to say safely. They become more politically cautious, more cautious about stirring up trouble, and more likely to tell leaders what they think leaders want to hear. Opposition becomes costly. The experiment becomes risky. The bad news that leaders desperately need in order to make good AI investment decisions is starting to come a long way, or not arriving at all.
This is not just hypothetical. I have worked with senior executives in various sectors who, after extensive restructuring, found their organizations to become curiously quiet. This strategy ends without resistance, problems appear too late, and middle managers begin to improve on what their boss notices rather than what the organization needs. The restructuring achieved its financial goal, but managing the organization became more difficult.
What leaders indicate, whether they intend to or not
There is a third dynamic that is worth naming, because it combines the above. Under conditions of uncertainty, research shows that people become significantly more sensitive to signals from those in power, especially negative ones. A leader who expresses doubt creates more anxiety than the same doubt expressed by a colleague. A big advertisement that implies job insecurity becomes more difficult to spread and spreads more than the same message in a stable environment. Small errors are magnified.
This means that leaders who take a direct investing approach operate at a broader scale than they may realize. The language they use matters more now, not less. When efficiency and investment in AI appear in the same sentence, the inference employees draw is not a narrow financial one. It is existential. The behavioral consequences of greater risk aversion, filtering of information, and management of a rising workforce are precisely the conditions under which big technology bets fail.
None of this is to say that capital reallocation in and of itself is wrong. The structural shift from human capital to computing is real, and it will accelerate. The question for leaders is whether they manage it in a way that maintains the organizational conditions their investment in AI will need to succeed. Right now, a large number of them are not. This is a risk that will not appear in the capital allocation model but will eventually appear in the results.
Nick Kinley He is a leadership psychologist, executive coach, and book author Power trapwho has worked with national bank CEOs, national security chiefs and hedge fund heads, as well as royalty, criminals, politicians and children.
