10.12.2026
Author - Enrique Dans
If, as Mustafa Suleyman says, AI will be wiping out most white collar jobs in the next 12 months, we should try to clarify what leadership becomes once that machinery exists.
In my opinion, at optimized companies, leadership will stop being mainly about making more decisions than everyone else, since many operational decisions will increasingly be made or recommended by systems that understand the company’s ontology, can simulate consequences through world models, and can learn from feedback. Essentially, the leader’s role moves up one level.
Today, executives spend enormous amounts of time resolving trade-offs, setting priorities, escalating decisions and coordinating. In an optimizable company, many of these local decisions will be handled by learning systems, so the leader’s job becomes to define the objectives, constraints, and trade-offs that these systems will be allowed to optimize.
This is a much more demanding and sophisticated form of leadership, since ambiguity can no longer be hidden inside slogans and empty declarations. You can repeat as much as you can that “customer comes first,” “we need to grow responsibly,” “we want to create long-term value,” and so on, but all that will be meaningless and insufficient unless the organization defines and specifies what all these ideas mean when objectives start to conflict.
When we balance margin versus growth, efficiency versus resilience, personalization versus privacy, speed versus quality, etc., a learning system can optimize aggressively against the objective it has been given. Locally rational loops can pull the company in incompatible directions unless objectives remain coherent at firm level. And making decisions on which sacrifices are acceptable will likely be a profoundly human and strategic responsibility.
The CEO starts to resemble a constitutional designer: not deciding every case personally, but defining what may be automated, what the organization is trying to maximize, what it must never violate, where humans retain veto power, which trade-offs require escalation, etc.
Once a company has its ontology, feedback loops and learning systems working properly, every new operation and transaction becomes additional experience and improves future performance and this can potentially create a new kind of winner-takes-all dynamic: a strong reinforcement effect with a better company model, better decisions, better outcomes, more useful feedback and a better model of the company. Once that loop begins compounding, late adopters may find themselves competing not against better software, but against organizations that have accumulated (and owned, not rented) years of proprietary experience.
Machines can optimize inside a defined goal, but humans remain responsible for redefining the objective itself when circumstances change. And that includes entering a new market, responding to some social or political shock, abandoning a product line, changing risk appetite or deciding that a metric no longer reflects the company’s values.
Leadership becomes the ability to know when the optimization target itself is wrong.
