Anthropic CEO Dario Amodei has reportedly predicted that the first one-person billion-dollar company could emerge before the end of 2026, enabled by AI tools that radically reduce the need for employees across software development, marketing, analysis, and operations. (The Times)

Is he right? Maybe. Is he exaggerating? I believe so, but the more important question is not whether one founder can build a billion-dollar company with AI. The more important question is why the future of work is being defined so heavily by the people building the technology, rather than by the people who will have to live with its consequences.

To dismiss the current dominance of tech leaders as mere arrogance is too easy. There is structural logic behind it. The builders do have arguments that deserve to be taken seriously before we explain why they are insufficient.

First, there is the argument of technical necessity. AI is complex. Governing requires knowledge of machine learning, scaling laws, infrastructure, model behavior, and security. A poorly designed regulation could be ineffective, technically impossible, or even dangerous. How can economists, union representatives, teachers, or public-sector leaders regulate systems they do not understand?

Second, there is the argument of geopolitical speed. In a world where AI capability is increasingly treated as a national-security imperative, deliberation can look like delay. If democratic societies pause to consult labor unions, educators, social scientists, and public institutions, we risk falling behind authoritarian competitors that face no such constraints. In this view, speed is not a luxury but survival.

This is the hardest point to answer. The threat model is real: adversaries don’t wait for consensus. I believe framing it as a choice between speed and democracy is a false dichotomy. The real question is an engineering one; how do we build governance structures that are both legitimate and fast enough to matter? We don’t solve this by abandoning deliberation; we solve it by designing institutions with shorter feedback loops, clearer mandates, and the ability to act decisively once consultation is complete. Speed without direction is dangerous. Direction without speed is obsolete, and we need both.

Third, there is faith in market efficiency. If AI creates productivity, then companies will adopt it. If it displaces workers, the market will eventually create new roles. (Exactly how that will happen is an open question). Intervention, proponents argue, risks distorting the very signals that tell us where value is being created.

Markets are efficient at price discovery and capital allocation. They are not designed to distribute gains equitably or maintain social stability. History shows that automation generates enormous productivity gains, but those gains are not distributed fairly without policy, institutions, bargaining power, taxation, education, and social protection.

These arguments are not trivial. They rest on a category error; namely treating AI as a primarily technical challenge, when it is also a social, economic, and political one.

Our goal should be to ensure that innovation serves society rather than merely concentrating power among a narrow group of founders, investors, and platform companies. If AI reshapes work, education, healthcare, government services, and public trust, then the conversation must include more than model builders and venture capitalists. It must include the nurses, our teachers, social workers, coders, designers, small-business owners, public-sector leaders, unions, economists, and your ‘ordinary’ workers.

Technical expertise is essential. But expertise is multidimensional. Knowing how to build a model is not the same as knowing how to manage its effect on a workforce.

Consider cybersecurity. We don’t let vulnerability researchers alone decide disclosure policy. The CVE ecosystem exists because we recognized that finding a flaw and governing its impact are different problems. Bug bounty programs are a form of distributed oversight. NIST frameworks are technocrats and policymakers working together. We accept that technical knowledge must be combined with regulation, ethics, and public-interest institutions. And the same should be true for AI.

Builders understand how to automate. But economists understand labor-market shocks. Teachers understand how people learn. Union leaders understand bargaining power. Public-sector leaders understand institutional capacity. Social workers understand vulnerability. Small-business owners understand what happens when technology changes competition overnight.

These are not “soft” perspectives. They are critical data points for a functioning society. So yes, we should listen to Amodei, Altman, Hassabis, and the other builders of frontier AI. However, listening is not the same as handing them the steering wheel.

AI may be too important to leave to people who don’t understand the technology, but it’s far too important to leave only to those who build it.