Original author: Satya Nadella, Microsoft CEO
Compiled by: Peggy
I've been thinking a lot lately about what the future of businesses will look like in an AI-driven economy.
This transformation is unlike any previous platform migration. In the past, we used digital systems to enhance human capital; this time, for the first time, we are able to establish a true cognitive loop between people and digital systems. This is a very disruptive concept because it will change the way we understand "work" itself within an organization.
The real key issue is not how a particular digital tool or system is used, but how organizations can continue to learn, accumulate intellectual property, differentiate themselves, and thrive in a world where AI models can continuously absorb and commodify human and organizational expertise.
Every company must build what I call human capital and token capital. Human capital includes employees' knowledge, judgment, networks, creativity, and pattern recognition abilities; while token capital is the AI capabilities that the company itself builds and owns.
Importantly, as token capital grows, human capital will not become less important. On the contrary, it will only become more important. I believe that human agency will become the core driver of token capital growth. Humans will set ambitious goals, connect clues across disciplines, build relationships, and identify truly important patterns. Without the guidance of human direction, computing power will simply go in circles.
This means the real opportunity lies not in choosing the best model, but in building a learning loop on top of that model, allowing human capital and token capital to compound and grow together. You can outsource a task, or even a job, but you can never outsource your learning. The future of a business lies in its ability to enable this learning to continuously compound between humans and AI.
This requires a new architectural approach: every company should be able to build intelligent agent systems that continuously improve over time while retaining control over its intellectual property. A company should be able to replace a "generalist" model without losing the "veteran employee" level of expertise accumulated in its learning system. This will be a key test of a company's control and sovereignty in the future.
Enterprises need to transform their own workflows, domain knowledge, and long-accumulated judgment into AI systems that can continuously improve with each use. Private benchmarks should measure whether the model truly improves on the business outcomes the enterprise cares about, rather than just looking at external benchmarks. Private reinforcement learning environments should allow models to become stronger based on real-world trajectories within the organization. Enterprise knowledge bases will make institutional memories searchable and improve the efficiency of token usage.
This closed loop will become a new form of intellectual property for the company. I see it as a "climbing machine." Moreover, unlike most assets, it grows exponentially. Each workflow improvement generates better training signals, thereby accelerating the accumulation of the company's unique tacit knowledge. Companies that establish this system earlier will gain an advantage that is difficult to replicate, regardless of how much the capabilities of individual models improve in the future.
What we least want to see is a world where every company in every industry hands over value to a few models that devour everything they see. If all value is ultimately captured by a few models, the political and economic structure simply will not tolerate such an outcome. An AI that hollows out an entire industry in the future will not be permitted at the societal level.
Consider what happened in the first phase of globalization: the entire industrial economy was hollowed out by outsourcing. On the surface, GDP figures may have looked good, but the real industrial shifts and job losses were real, and their consequences are still being felt today. We cannot bring this dynamic into the AI era—allowing a few AI systems to capture all the economic returns while the knowledge of entire industries is commodified and hollowed out at their feet.
In my view, our priority must be building a cutting-edge ecosystem, not just a cutting-edge model. Only in this way can value flow widely to every company, every industry, and every country. In such an ecosystem, each organization can have its own learning loop, encoding its own institutional knowledge and enabling human capital and token capital to grow together through compound interest.
This is also the platform spirit I have always believed in: the value created on the platform should be greater than the value captured by the platform itself; every company should be able to continuously innovate and create its own value.
When this is achieved, businesses will create value for themselves and for the economic environment in which they operate. Employees' professional skills will be amplified, their judgment will become part of the system, and will become replicable and scalable, with these benefits flowing back to the company and its surrounding community.
This is how businesses create value for themselves and the broader economy. It is also the stable equilibrium we should work together to build.
