What You Don't Build Today Will Be Difficult to Catch Tomorrow

The last 30 years I believed I had lived through the most important technological revolutions. I was wrong.

Of course the arrival of personal computers changed how I work, the internet changed how I communicate, smartphones changed how I live. Yet in each case, the underlying structure of society remained largely intact. Businesses still operated as businesses. Governments still governed. Institutions adapted without fundamentally changing their nature. I have build companies and products for each of these waves that I thought were revolutions.

When I started to build with AI in 2011, there were 2 movements, the symbolic one with models following rules and the cognitive one with models finding rules. We managed to build pretty good predictive models, the first "okay" chatbots and the first impressive driverless cars (thanks to massive progress in computer vision) but we were far from where we are today.

Generative AI feels and is different. Not because it is another powerful technology, but because it is beginning to transform something that has always been scarce: intelligence itself.

This idea has been articulated in different ways by some of the most influential thinkers and builders of our time. While their backgrounds range from venture capital and enterprise software to scientific research and AI infrastructure, they are increasingly converging on the same conclusion. We are not entering another software cycle. We are entering a world where intelligence becomes abundant.

I would describe this shift as a transformation of the foundations of society itself.

We are not simply dealing with a tool for productivity or content generation. It is becoming infrastructure. Like electricity before it, AI will disappear into the background while simultaneously becoming indispensable to nearly every aspect of modern life.

Andrej Karpathy, one of the most influential AI engineers of the past decade, arrives at a similar conclusion from a different direction. His concept of "Software 3.0" suggests that software itself is being redefined. We moved from software written by humans to neural networks trained on data. Now we are entering an era where natural language becomes the programming interface and intelligence becomes programmable. In Karpathy's framing, AI is not being added to software. Software itself is being rewritten.

Marc Andreessen's vision of technological abundance reaches a similar destination. In his Techno-Optimist Manifesto, he argues that the history of human progress is the story of making scarce resources abundant. Industrialization scaled production. The internet scaled information. Artificial intelligence scales expertise. What was once available only to highly trained professionals may soon be accessible to anyone, anywhere, at marginal cost.

Sam Altman often frames the future in even simpler terms. He has repeatedly suggested that intelligence will eventually become as ubiquitous as electricity or internet connectivity. It will not be something people buy as a standalone product. It will become part of the environment. Intelligence on demand. Intelligence embedded into every workflow, every application, every decision.

Jensen Huang frequently speaks about AI agents as digital employees. Not software. Not assistants. Workers. In his vision, every company will eventually manage both human employees and AI employees. Organizational charts will evolve to include agents alongside people.

Satya Nadella extends this idea further. He has suggested that the business applications we rely on today may become far less important than the agents operating across them. In a world where intelligence can access systems directly, the software interface itself becomes secondary. The critical question is no longer which application someone uses. It becomes which agent is performing the work.

Even long before the current AI boom, MIT scientist Max Tegmark was arguing in Life 3.0 that humanity was approaching a moment when intelligence itself becomes programmable. His work focused less on products and more on civilization, but the core insight remains remarkably consistent with today's leading technologists.

What is striking is not that these individuals agree on every detail. They do not. What is striking is that they come from entirely different disciplines and continue to arrive at the same destination. The old world was built around scarce intelligence. The new world is being built around abundant intelligence.

I believe that distinction changes everything.

For centuries, organizations scaled by adding people. More work required more employees. More expertise required more specialists. More growth required more management layers. The economics of organizations were constrained by the availability of human intelligence. Those constraints are beginning to weaken. I see that daily in my conversations with business leaders.

When intelligence becomes abundant, expertise is no longer the bottleneck. The bottleneck shifts to execution, judgment, trust, coordination, and access to proprietary data. Competitive advantage no longer comes from simply knowing more. It comes from acting faster, designing better systems, and making better decisions.

This is why the debate about whether AI will replace jobs often misses the bigger picture. The deeper question is what happens when a company of twenty people can accomplish what previously required two hundred. What happens when entire operational workflows execute autonomously. What happens when intelligence itself becomes a utility.

The implications extend far beyond business.

Healthcare will change. Education will change. Government will change. Scientific discovery will accelerate. Entire industries that evolved around the scarcity of expertise will need to reinvent themselves. This is not because AI is another application layer. It is because intelligence is becoming infrastructure.

For the first time in human history, we are learning how to scale intelligence itself.

I don't believe we are witnessing another technology cycle and when I am asked my message is a simple one, do not wait: not because AI is fashionable or because your competitors are talking about it but because it is way more profound than what we think.

And the transition has already started. The organizations that are learning how to work with AI today are not simply adopting a new tool. They are learning how to operate in a world where intelligence is abundant, where software can reason, where agents can execute workflows, and where a team of twenty can accomplish what once required two hundred.

If you are still waiting for certainty or for the technology to mature or for regulations to stabilize or waiting for someone else to prove it first, you might be waiting too long.

You do not need to transform your organization overnight and automate everything. But you do need to start. Don't think about automation in a simple linear way, starting with agentic AI today means compounding knowledge, data, processes, and capabilities that become increasingly difficult to replicate tomorrow.

Just start, start small but start.

Book a Demo

See how AI Agents can transform your operations