The Next Phase of AI: From Thinking to Doing

Back in 1995, Carl Sagan warned about the long-term consequences of losing industrial capability. Today, we’re at a different kind of turning point—one driven not just by globalization, but by technology itself.

AI has already reshaped the digital world. It can write, analyze, generate, and communicate. But for all its capabilities, it doesn’t really move.

That’s starting to change.

The next phase of AI isn’t just about thinking—it’s about acting in the physical world. The line between software and robotics is blurring, and we’re seeing the early stages of what some describe as a breakout moment for embodied AI.

You can already see it in how companies are operating. Tesla is training its systems on real-world data from millions of vehicles, with projections that this could scale to tens of millions—potentially 100 million vehicles by 2040. Every mile driven feeds into a feedback loop that improves performance in real time.

Amazon offers another signal. Its warehouses have steadily increased automation, moving from roughly five humans per robot in 2017 to closer to two per robot by 2024, with estimates suggesting automation could drive up to $10 billion in annual savings.

These aren’t isolated examples—they point to a much larger shift.

So far, most of AI’s impact has been in the knowledge economy. But the physical economy is where the scale really sits. Globally, there are over 4 billion workers, representing roughly $40 trillion in annual labor value. Even a 1% shift toward automation translates into hundreds of billions of dollars in impact.

The opportunity is amplified by the sheer scale of physical systems already in motion. There are approximately 1.2 billion vehicles worldwide, traveling around 12 trillion miles annually, with humans spending an estimated 720 billion hours per year inside cars. That’s not just transportation—it’s a massive platform for autonomy, data capture, and new forms of productivity.

And data is the key driver.

Real-world, sensory data is what allows AI to move from simulation to reality. Companies that can capture it—at scale—gain a compounding advantage. It’s the difference between systems that can generate answers and systems that can operate in the world.

There’s also a broader competitive layer to this. Advances in AI, robotics, and automation are increasingly tied to geopolitical positioning, with echoes of past “Sputnik moments” driving urgency in areas like defense, manufacturing, and infrastructure.

At the same time, this isn’t just a large enterprise story. Small businesses—representing over 90% of global economic activity—are beginning to access these capabilities through cloud platforms and robotics-as-a-service, lowering the barrier to entry.

We’re still early. Much of this is experimental, and the technology will continue to evolve.

But the direction is clear.

AI is moving from understanding the world to operating in it.

And once that transition fully takes hold, the impact won’t just be incremental—it will reshape how work, industries, and economies are structured.