Investors are committing one of the biggest strategic mistakes in twenty years. They are laying off workers to accelerate AI adoption. As if AI alone could replace human competence. As if performance came from pure automation.
Except every serious study shows the exact opposite. Maximum value lies neither in humans alone, nor in AI alone, but in the human+AI duo. What I call the centaur model.
The mirage of total replacement
The logic dominating boardrooms right now is seductive in its simplicity: replace employees with AI agents, reduce payroll, increase margins, show impressive metrics to shareholders. On an Excel spreadsheet, it works perfectly.
In operational reality, it is a disaster in the making.
AI moves fast. Humans understand correctly. And in the real world, understanding correctly creates far more competitive advantage than moving fast. An AI agent can process a thousand support tickets in an hour. But it does not detect that a strategic client is about to leave. It does not sense that a particular phrasing in an email conceals deep frustration. It does not adapt its approach when context shifts subtly.
Companies that have massively laid off staff to automate their customer service are learning this the hard way. Customer satisfaction drops. Churn increases. Escalations explode. Because customers do not want a fast answer — they want a right answer.
The centaur model: when human and AI multiply each other
The centaur model idea comes from chess. In 1998, Garry Kasparov, defeated by Deep Blue, invented the concept of "centaur chess": a human + a machine systematically beat a machine alone AND a human alone. Not because each is better in their domain — but because the combination creates something neither can produce on its own.
In the field, the most spectacular gains I observe come from exactly this model:
Humans retain judgment, nuance, and meaning. Strategic decisions, negotiation, crisis management, exploratory creativity — these activities rely on contextual intelligence that AI does not possess and will not possess in the medium term.
AI handles repetitive workload, research, and structuring. Document synthesis, first draft generation, massive data analysis, automated monitoring, categorization — AI excels at tasks with low cognitive value-add but high time cost.
Workflows redesigned so that each amplifies the other. This is the key. It is not enough to hand an AI tool to an employee. You must redesign the workflow so that human steps and AI steps chain together fluidly, with clear control points.
A consultant augmented by AI does not work faster. They work better. They arrive at a meeting with a 200-page synthesis already digested by AI. They ask the right questions because they had time to think instead of compiling data. They produce sharper recommendations because AI showed them patterns they would not have seen alone.
What investors cannot see in their spreadsheets
The problem with purely financial reasoning is that it fails to capture second-order effects.
Loss of tacit knowledge. When you lay off 30% of your workforce, you do not lose 30% of labor capacity. You lose years of business knowledge, client relationships, and understanding of operational subtleties. This knowledge is not documented anywhere. And AI cannot recreate it — it never had access to that information in the first place.
Cultural fragility. Layoff survivors are not motivated. They are terrified. Creativity, risk-taking, initiative — everything that makes an organization innovate — vanishes when people fear for their jobs. AI does not compensate for a climate of distrust.
Degraded supervision quality. Generative AI needs supervision. By competent humans. Fewer humans means less supervision means more errors slipping through. It is a vicious cycle that accelerates with each round of layoffs.
Reputational risk. Companies known for mass layoffs in favor of AI attract fewer top talents. Yet in 18 months, when the limits of pure automation become obvious, these companies will need to rehire — in a market where nobody wants to join them.
The future of work is not autonomous. It is symbiotic.
The organizations that will win will not be those that automate the most. They will be those that design the right human+AI tandems in their workflows.
Concretely, this means:
Map your activities. For each process, identify what requires human judgment (keep), what involves raw processing (automate), and what benefits from human+AI collaboration (redesign).
Train, do not replace. Invest massively in AI upskilling for your existing teams. An employee trained to work with AI produces more value than an unsupervised AI agent. And costs less than a new hire in 18 months.
Measure the right metric. Not headcount reduction. Not inference cost. The value created per augmented human+AI tandem. Revenue per augmented employee. Decision quality. Customer satisfaction. Time-to-insight.
Those who bet on AI as a replacement destroy value. Those who bet on AI as augmentation create leverage. The difference between the two is measured in billions over a five-year horizon.
Is your organization replacing, or augmenting? The answer to that question will determine your competitive position for the next decade.