Access is not a benefit
The same AI models reached Denmark and the United States in the same years and produced opposite effects on the job market. What separates the countries that gain from artificial intelligence from those that do is not compute, capital, or geography. It is leadership.
Click on the image to buy
the book on Amazon
This book was written for the more than 190 governments that will not build a frontier model and must still decide how this technology enters their societies. It turns the evidence assembled by the UN Scientific Panel into sequenced national action: a ladder for locating where your country stands, seven pillars of work, and a first move for Monday morning.
The executive summary is free, runs to a few pages, and carries the whole shape of the argument.
Neither the book nor the executive summary requires any technical background.
What the evidence actually shows
On July 6, 2026, governments convened in Geneva for the first time to discuss how artificial intelligence should be governed. Five days earlier, the Independent International Scientific Panel on AI, forty scientists from every region, published the first global independent scientific assessment of the technology. Its preliminary report is careful, sober, and quietly alarming.
Three findings from that report define the moment. Adoption has moved faster than any general-purpose technology in history: more than a billion people now use conversational AI every week. Capability is improving faster than institutions can measure it, with task complexity driving systems to double every few months, which means the technology will transform itself several times within a single electoral term. And capacity is concentrated: two countries hold roughly ninety percent of the specialized computing power behind the world’s leading AI supercomputers. For nearly every other nation, including wealthy ones, this technology arrives as an import.
There is a comfortable assumption buried in most national conversations about AI, and the evidence dismantles it. The assumption is that access is the problem, and that once a country’s people can reach the technology, benefits follow. They do not. Young American workers in the most exposed occupations have seen measurable relative employment declines. Danish workers, by the best registry evidence, have seen almost no effect on employment, hours, or wages. Same technology, same timing, same access, opposite outcomes. In India, AI screening for diabetic retinopathy has helped screen hundreds of thousands of people and saved the sight of thousands, not because the algorithm was singular but because it was embedded in clinics, referral pathways, and follow-up care.
The technology is portable. The benefit is not.
What separates the two cases is the enabling environment, which is built by decisions about education, data, institutions, procurement, regulation, and trust. It is built, in other words, by leadership. The UN’s trade and development body counts 118 countries absent from the major AI governance discussions, and fewer than a third of developing countries with a national AI strategy of any kind. The gap that matters is not a compute gap. It is a fluency gap, and fluency is learnable.
What the book gives a leadership team
National fluency is the deliberate, learnable capability of a country’s leadership and institutions to adopt, adapt, govern, and shape artificial intelligence on the country’s own terms. It does not require building frontier models, which is out of reach for almost everyone and unnecessary for almost everyone. It requires institutions and people who understand the technology well enough to use it, question it, contract for it, regulate it, and negotiate over it without flinching.
The National AI Fluency Ladder
A five-level scale on which a leadership team can locate its country today and see what the next level asks of it. Position is uneven by design, so a country can stand at Level 4 in health and Level 1 in justice.
|
Level 1 |
No strategy, no owner, no assessment. The name describes the government, not the country: adoption proceeds unguided and invisible to policy. |
|---|---|
|
Level 2 |
The subject has reached the cabinet, and there is a strategy document, often a handsome one. More than seventy countries have published one, and most currently live here. |
|
Level 3 |
The technology is deliberately at work on national problems, at a scale citizens encounter, with someone responsible for the outcomes. |
|
Level 4 |
The country stops accepting the technology as shipped: it localizes models for its own languages, evaluates against its own benchmarks, and holds its data and compute arrangements on terms it understands. |
|
Level 5 |
Influencing the terms on which AI develops and enters societies. Its current occupants are mostly small and middle powers, which is proof that it does not require superpower scale. |
The seven pillars
The work itself, organized so that a cabinet can divide it. None of the seven requires wealth on the scale of the superpowers. All require sustained leadership attention, which is the scarcest resource in any government, and because the pillars multiply rather than add, the return on any one is capped by the weakest of the rest.
- Start with problems, not technology. A strategy is a decision framework, not a document.
- Foundations. Compute, data, energy, and connectivity, and the honest buy, build, or partner choice behind each.
- A nation that can use AI. Literacy, workforce, talent, a fluent public service, and leaders who practice.
- Teach the machines your language. Corpora, consent, open data, and benchmarks in the languages your citizens actually speak.
- The state as first adopter. Government as the proving ground, with procurement used as the power it is.
- Guardrails that build trust. Governance as road conditions rather than a brake, sized to real regulatory capacity.
- The economy of the fluent nation. Complementary investment, good jobs by design, and the fiscal base that pays for schools and clinics.
The book closes with four working instruments: a national fluency diagnostic, a hundred-day strategy sprint, a first hundred days for a newly appointed minister, and a practical map of the global governance machinery.
Who this book is written for
Ministers of ICT, finance, education, health, and planning. Permanent secretaries and senior civil servants. National AI strategy teams and the offices of presidents and prime ministers. Central bank governors and heads of regulatory agencies. And the institutions that work alongside them: development bank knowledge teams, UN agencies, bilateral donors, and the diplomats carrying their countries’ positions into the Global Dialogue.
The book assumes no technical background.
Read the executive summary
A short brief covering the argument, the ladder, the seven pillars in brief, a first self-diagnostic your team can run in a working session, and the honest headwinds. Written for the person who has to decide. It arrives by email, free.