Investment in AI has accelerated exponentially. Capital has poured into chips, data centers, and new models. New models are released at a steady pace, and spending continues to rise.
Yet across sectors, many organizations are reaching the same conclusion. Adoption is proving harder than invention, especially for general-purpose technologies, as it once was for electricity. As a result, AI adoption remains on the sidelines. Systems exist, but changing how work actually gets done is far more difficult. The constraint is no longer technological capability. It is whether institutions and organizations are prepared to absorb AI. Institutions and organizations play distinct roles. Institutions are the rules, incentives, standards, and accountability structures that reduce uncertainty and make new behavior safe and trusted. Organizations operate within those rules and change workflows accordingly.
Germany pioneered the chemical industry, but the United States diffused it by embedding chemistry into manufacturing and everyday commerce. Productivity followed only after institutions evolved and organizations redesigned workflows. The United States also created the discipline of chemical engineering and applied it across sectors such as food and automobiles, drawing on its long-standing ability to attract the best and brightest talent globally and turn invention into industrial scale.
This is not a new pattern. Research on technological change shows that economic advantage rarely comes from being first to invent. It comes from the ability to diffuse new technologies broadly and productively. As Jeffrey Ding argues in Technology and the Rise of Great Powers, leadership is shaped less by breakthrough innovation and more by the capacity to absorb and deploy technology at scale.
As Nobel Prize winner Douglass North observed, institutions are the rules and incentives that shape behavior, while organizations are the actors that operate within them. That distinction explains why diffusion depends on institutional change as well as organizational capability.
Two years ago, Nandan Nilekani, co-founder of Infosys and the founding Chairman of the Unique Identification Authority of India, said that India will be the use-case capital for AI. As the architect of Aadhaar, the world’s largest biometric identity system, he said AI will change India, and India will change AI. Drawing on his experience building Aadhaar, Nilekani has argued that AI needs India as much as India needs AI because India’s scale, institutions, and use cases will shape how AI is deployed in the real economy.
Aadhaar shows what diffusion at scale actually looks like. India has more than 1.4 billion Aadhaar holders. Identities have been authenticated digitally over 164 billion times, leading to an estimated half-trillion dollars in savings by reducing leakage, duplication, and friction across the economy. UPI, built on top of these digital rails, is now the world’s largest real-time payment system, processing roughly 20 billion transactions per month.
Biometric technology was not new when Aadhaar was introduced. What changed was absorption at two levels. Institutionally, the Unique Identification Authority of India set standards, verification rules, and accountability, absorbing trust and risk at the system level.
Organizationally, banks, government agencies, and private firms absorbed biometric identity into workflows and everyday operations. Individuals did not have to judge reliability on their own. Diffusion theory shows that leading sectors emerge when general-purpose technologies are adopted across organizations and supported by institutions. Like earlier general-purpose technologies, AI will create value only when institutions absorb uncertainty and risk, and organizations can turn use cases into repeatable workflows at scale.
CEOs increasingly understand that the challenge is no longer access to AI capability. When executives ask for AI use cases, they are not asking for demonstrations of technical performance. They are asking whether AI can be trusted inside real systems. They want to know whether it can be used consistently, without shifting risk onto individuals.
Many AI initiatives stall because organizations treat adoption as a tooling problem without sufficient institutional adaptation. AI is added to existing workflows rather than integrated into how decisions are made. Accountability is unclear. Users are left to judge outputs, manage errors, and decide when systems are safe to rely on.
Feedback loops are often missing or informal. The broader majority does not tolerate uncertainty.
Diffusion helps explain why. Technologies spread when institutions set clear standards, incentives, and accountability that make new behavior safe and routine, and when organizations absorb technology through learning and use. They stall when uncertainty and liability are pushed onto individuals or isolated teams. Until responsibility is clearly owned at the institutional level and organizations build the capability to integrate AI into workflows, use cases remain pilots rather than sources of lasting value.
Diffusion is often misunderstood as a question of speed. In reality, it is a question of learning, capability, and the application of technology across sectors. Organizations learn through use.
Capability develops as workflows change and skills mature. Application across sectors is what ultimately produces productivity gains.
Organizations that diffuse technology effectively redesign workflows and clarify ownership within stable institutional frameworks. Responsibility becomes clear. Workflows change before tools scale. Productivity gains come from redesigning processes, not from adding software. Trust is anchored in institutions and expertise, not models alone. Systems improve through real-world experience rather than isolated pilots. These capabilities develop over time. They cannot be added at the end of deployment.
The next phase of AI competition will not be decided by who builds the most powerful models. It will be shaped by which societies build institutions that absorb uncertainty and risk, and organizations that absorb technology into daily work. The race to build artificial general intelligence (AGI) has a destination and a winner. The race that matters for economic impact is diffusion. Organizations carry diffusion forward, but institutions shape the incentives, rules, and trust that determine whether it succeeds.
That change does not come from organizational adoption alone. It comes from institutional change that makes new ways of working safe, repeatable, and trusted.
The most valuable AI systems will not look dramatic. They will look ordinary. They will fade into routine decisions and familiar processes. They will be embedded.
History suggests that leadership in general-purpose technologies belongs to those who diffuse them effectively. AI will follow the same path.
One race is about power. The other is about productivity. The narrative decides which race we think we are running.
Waiting for perfection is not a strategy. The question is whether the United States can afford to engage seriously in the diffusion race only after the geopolitical race is decided.
The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.
