Satya Nadella just gave CEOs a blueprint for the next phase of AI. Here's why it matters.

When the CEO of one of the world's largest technology companies publishes a treatise about the future of the firm, people pay attention.

This week, Microsoft's Satya Nadella did exactly that.

His essay, A frontier without an ecosystem is not stable, quickly attracted millions of views because it did something surprisingly rare in the AI conversation: it gave business leaders a clearer way to think about what comes next.

Not better models. Not bigger models. Not even AI itself.

But how businesses learn, retain knowledge and create value in a world where AI becomes part of the fabric of work.

And the whole Decidr office found themselves nodding along.

Nadella articulated many of the ideas we've been thinking about for years: knowledge sovereignty, institutional memory, organisational intelligence, and the need for an orchestration layer that sits above individual AI models.

Here are the ideas we think every CEO and business leader should pay attention to.

Human capital and token capital: a new way to think about company value

One of the ideas getting attention from Nadella's essay is his distinction between human capital and token capital.

"Every company is going to have to build what I think of as human capital and token capital. Human capital comprises the knowledge, judgment, relationships, ingenuity, and pattern recognition of its people, while token capital is the firm's AI capability it builds and owns."

At first glance, this sounds like another AI buzzword. But it's actually a useful way to think about where value sits inside a business.

We’ve referred to this as AI sovereignty in the past.

Human capital is familiar. It's the expertise of your people. The judgement of your leaders. The relationships your team has built. The pattern recognition that comes from experience.

It's why a great salesperson outperforms an average one, why an experienced project manager spots risks earlier and why some organisations consistently make better decisions than others.

Token capital is newer. It's the AI capability a business builds and owns. Not just models, but workflows, agentic apps, knowledge systems and the accumulated intelligence encoded into those systems.

The interesting part is that Nadella doesn't see these as competing forms of capital.

For years, the AI conversation has often implied a tradeoff: more AI means less need for people.

Nadella argues the opposite.

"Human capital does not become less valuable as token capital grows. It only becomes more valuable!" he writes.

The better your people are, the more valuable the systems they train and guide become. And the better those systems become, the more widely that expertise can be applied across the organisation.

In other words, AI doesn't reduce the value of human knowledge. It creates new ways to capture, distribute and scale it.

This idea sits at the heart of Decidr's view of the agentic organisation. Most businesses already possess enormous amounts of human capital. The challenge isn't creating it, but making it available beyond individual people, teams and systems.

How do you turn expertise into something your business can actually use, learn from and build upon?

That's ultimately a knowledge and structure problem, not simply an AI problem.

The learning loop becomes the new IP

The second major idea in the essay is that organisational knowledge can, and should, compound.

Most businesses already possess valuable knowledge.

Nadella argues that organisations need to start turning that knowledge into systems that improve over time:

"Companies need to turn their workflows, domain knowledge, and accumulated judgment into AI systems that improve with each use."

Imagine a business where every customer interaction, operational decision and workflow creates a small piece of feedback.

The system learns from what worked. It learns from what didn't. It learns from exceptions, edge cases and outcomes. Then the next decision starts from a stronger position than the last.

This is what Nadella calls a learning loop. The result is an organisational asset that becomes more valuable through use.

Decidr calls this the schema layer: the encoded logic of how your organisation actually makes decisions, not just what it decides.

When that layer is owned by you and runs inside your environment, every workflow execution makes it more precise to your business. The model underneath becomes interchangeable. The institutional intelligence doesn't.

Nadella puts it neatly: "The companies that build this early will have an advantage that is hard to replicate, regardless of any new individual model capability."

Why ownership matters (and the model doesn't)

The final section of the essay is perhaps the most important.

Nadella warns about a future where businesses create the knowledge but don't capture the value. We’ve referred to this as the great knowledge heist.

"The last thing any of us want is a world where every company across every sector is ceding value to a few models that eat everything they see," Nadella says.

He draws a parallel with the first phase of globalisation. While globalisation created enormous value, it also hollowed out industries and shifted economic power away from the communities and businesses that originally created that value.

Nadella is warning that AI could create a similar pattern.

Industries spend decades building expertise. Businesses develop specialised knowledge. Employees accumulate judgement and experience. Then the value created by that knowledge flows to a small number of platforms.

Every time a task gets offloaded without capturing the judgement behind it, that knowledge migrates into someone else's infrastructure and doesn't come back.

And the industries that ignore this won't see the damage in their near-term numbers, just as globalisation looked fine in the GDP print while the underlying value was already moving. By the time the extraction is visible, it's done.

His recommendation is not to slow down AI development. It is to build a stronger ecosystem around it, so that businesses own their value and use AI to compound it.

They need a way to capture institutional knowledge, preserve it, improve it and continue benefiting from it.

This is the argument Decidr has been making all along

Nadella's essay expresses something we've watched play out in businesses for years.

The businesses that come to us aren't short of AI tools. They're short of a structure that makes those tools work as a coherent whole, one that learns, retains and improves rather than starting from scratch every session.

That structure is what DecidrOS is built to provide. The schema layer encodes how your organisation actually reasons. The task graph maps every goal, workflow and dependency into a live, connected picture of how work gets done.

And every decision, every outcome, every evaluated action gets stored permanently so the organisation doesn't just use AI: it builds an intelligence asset it owns.

This is exactly the distinction Nadella draws between an organisation that compounds and one that creates dependency.

AI sovereignty isn't a niche concern for large enterprises. It's the question every founder and operator should be asking right now, while the patterns are still forming and the cost of getting it wrong hasn't fully landed.

The model is the engine. But your organisation's expertise is the fuel. The question is whether you're building the infrastructure to hold onto it.