When the first general-purpose AI systems began behaving like reasoning partners rather than calculators, every existing leadership playbook quietly went out of date. Boards still discuss "digital transformation" as if it were a five-year initiative. Cabinet ministers still appoint "AI envoys" as if the technology were a foreign country to be negotiated with. Each of these framings shares a common error: they assume the new layer of intelligence sits outside the organization, waiting to be invited in.
It does not. It is already inside: in spreadsheets that suggest their own answers, in legal drafts that catch their own contradictions, in scientific assistants that propose the next experiment. The question for leaders is no longer whether to incorporate AI but how to govern a workforce that is permanently mixed.
The illusion of control
Traditional governance assumes that the people who write the rules also understand what they are governing. With AI, that assumption is now broken in two places. The technology evolves faster than committees can convene, and its capabilities emerge in ways even its creators do not fully anticipate. A regulator asked to assess a frontier model is in roughly the position of a 19th-century legislator asked to assess a steam engine, except that the engine doubles in horsepower every six months.
The temptation is to respond with new acronyms: AI offices, sovereign compute commissions, frontier-risk task forces. Most have been established in the last 18 months. Few have produced consequential decisions. The instinct is correct: leadership must engage with this. But the form is wrong. You cannot legislate around something you cannot see clearly.
The new vocabulary
There are now four words every senior leader should be able to define without help: compute, capability, alignment, and deployment. They are not technical jargon. They are the building blocks of any honest conversation about what the technology is, what it can do, where it is safe, and how it reaches people.
The shortest summary: compute is the fuel, capability is what the engine produces, alignment is whether the engine goes where you want, and deployment is the moment the engine actually enters the world. A leader who cannot articulate these four words is unfit to make strategic decisions about AI, not because the words are difficult, but because their absence guarantees that the leader is being briefed in metaphors, not substance.
Three traps
There are three traps senior decision-makers consistently fall into.
The first is the procurement trap. Treating AI as a vendor problem (which tool to buy, which provider to standardize on) misses the structural shift. The technology is not a SaaS subscription that integrates with your stack. It is a new kind of colleague whose abilities you do not yet understand and whose mistakes will be attributed to you.
The second is the ethics trap. Outsourcing the hard questions to an "AI ethics committee" that sits next to communications and legal. Ethics committees that do not have a stop button on actual deployment are theatre. The serious version of this work lives inside the operating decisions of product, research, and policy teams.
The third, and the most expensive, is the sovereignty trap. Believing that owning the model is the same as owning the outcome. A government can fund a national champion, a corporation can build proprietary infrastructure, and neither answers the question of what gets done with the capability. The Soviet Union owned its computers. It did not own the future of computing.
The case for cross-sectoral coalitions
The challenges that matter (climate, AI governance, inclusion, the next pandemic) share a structural feature. They cannot be solved by any single sector, region, or institution. Climate negotiation collapses if the largest emitters are missing; AI governance collapses if the frontier labs are missing; inclusion collapses if the affected communities are missing.
What this implies, in practice, is that the leaders who will matter most over the next ten years are those who can convene across the lines that used to separate their predecessors. A finance minister who can speak the language of a frontier lab. A frontier-lab CEO who can read a climate finance term sheet. A diaspora entrepreneur who can sit in a UN room without losing their voice. These are rare profiles today. They will be central tomorrow.
What leadership looks like now
A leader in the AI era is not the person who has the most decisive opinion. The most decisive opinions are usually a few months out of date. The leader is the person who can hold the question open long enough for the system around them to absorb a new piece of evidence, and then act before it is too late.
I have not yet made up my mind: the most underrated leadership statement of the next decade.WLS Editorial
The leaders who can deliver that sentence, and then go on to make the call when it matters, will be the ones the rest of us remember.
An editorial analysis from the World Leaders Society.
All analyses
