Every Company is an AI Company

In 2011, Marc Andreessen wrote that software was eating the world. It was a declaration that every business (banks, retailers, hospitals, law firms) would eventually become software businesses whether they wanted to or not. The insight was radical at the time and trivially obvious within a decade. In 2019, I wrote a follow-up arguing that the next layer was design: that software alone had become a commodity, and that companies capable of thinking clearly about problems, users, and systems would be the ones that compounded. Companies where design was a strategic function, not a decorative one. I still believe that. But something has happened since then that demands another update to the thesis.

We are living through a third paradigm shift. And like the ones before it, it will not replace what came before. It will subsume it.

How paradigm shifts actually work

In my 2019 piece I borrowed the Kuhn Cycle to explain these transitions: the idea that technological paradigms don’t evolve linearly, they rupture. There is a period of normal science (a shared intellectual framework, stable methods, agreed-upon problems) followed by accumulating anomalies, a crisis, and then a revolution that redraws the map entirely. After the revolution, a new period of normal science begins, until the next crisis.

The software paradigm followed this arc. The App Store didn’t just create a distribution channel. It created a new ontology for companies. Suddenly, the question was not “do we use software?” but “are we software?” Incumbents that failed to make that conceptual leap didn’t just fall behind; they became irrelevant. Kodak understood digital photography technically. They didn’t understand that they had become a software company whether they liked it or not.

The design paradigm followed the same arc. As software became ubiquitous (as standing up a backend, shipping a mobile app, or building a SaaS product became progressively easier and cheaper) software alone stopped being a competitive differentiator. The anomalies accumulated: why did some products with inferior technology win? Why did technically superior products die in the market? The crisis was quiet but real. Engineering-first companies started losing to design-first companies, and the survivors learned to fuse both. Apple had been proving the thesis for thirty years. The rest of the industry finally caught up.

Now the same arc is playing out again, faster and more violently than either of the previous two.

The rupture we are inside of

The anomalies are already obvious. A two-person company can now build and operate what previously required a team of twenty. A founder can ship product, analyze data, generate content, handle customer support, draft legal language, and synthesize competitive intelligence. Not by working harder, but by working with systems that reason. The constraint is no longer access to capital, talent, or technology. The constraint is clarity of thought about what to build and why.

That is the rupture. Not AI as a feature (a chatbot added to a product, a recommendation engine bolted onto a feed). That is the software era’s way of thinking about AI: as a module, a component, a service you call. The paradigm shift is something deeper. AI is becoming the operating system of the company itself.

Intelligence is no longer something you buy from a vendor or hire from a university. It is something you design into the structure of how your business works.

The companies that will define the next decade are not adding AI to their workflows. They are rebuilding their workflows around AI. Their product strategy is informed by systems that synthesize market signals continuously. Their customer interactions are mediated by agents that learn from every exchange. Their internal operations run on documented reasoning, not institutional memory. Their teams are smaller, faster, and more levered because the cognitive work that used to require human hours can now be delegated to systems that don’t sleep and don’t forget.

This is not automation in the old sense. Automation replaced repetitive physical and clerical work. What we are talking about now is something qualitatively different: the delegation of judgment. Not just “do this task” but “figure out what to do, and do it.” That is a different category of capability, and it changes what a company fundamentally is.

What “every company” actually means

When Andreessen said every company is a software company, he didn’t mean every company writes code. He meant that every company’s competitive position would be determined by its relationship with software (whether it understood it, invested in it, organized around it, thought in its terms). The companies that got this became durable. The ones that treated software as a vendor relationship, as IT infrastructure, as someone else’s problem, got eaten.

The same logic applies now, at a different level of abstraction. Every company is an AI company does not mean every company builds models or hires machine learning engineers. It means every company’s competitive position will be determined by its relationship with AI: whether it understands it, builds with it, organizes around it, thinks in its terms.

The question is not whether your industry will be affected. Every industry will be affected. The question is whether you are building a company with the muscle to navigate that, or whether you are still debating whether AI is real.

In 2019, I wrote that the Designer’s role becomes strategic during a paradigm shift. The same is true now, but the sentence needs updating. The most important role in the company is whoever thinks most clearly about how intelligence (human and artificial) should be structured. Not a CTO role, not a data science role, not an “AI lead” role. A thinking role. A systems role. A design role, in the deepest sense of that word.

The layers stack, they don’t replace

Paradigm shifts don’t erase their predecessors. I was explicit about this in 2019, and it bears repeating. Software didn’t replace manufacturing or distribution. Design didn’t replace engineering. AI will not replace software or design.

What happens is that the previous layer gets abstracted. Once you have AI, you can produce software faster, more cheaply, with fewer people. The constraint moves up. The bottleneck is no longer whether you can build the thing. It is whether you know what to build, whether you have the judgment to evaluate what the system produces, whether you understand the problem well enough to give the right direction. Design, in the Kuhnian sense, becomes more important, not less.

The companies that will be destroyed by AI are not the ones that hire too many software engineers. They are the ones that have organized themselves around processes that require no thinking: that have replaced judgment with procedure, insight with compliance, strategy with template. These companies are already brittle. AI simply makes their brittleness visible faster.

The companies that will compound through this transition are the ones that have preserved (or are rebuilding) their capacity to think. To ask the right questions. To know what a good output looks like. To understand their customers well enough to direct an AI system and evaluate its results. That is not a technical skill. It is a design skill. And it sits on top of software the same way software sat on top of distribution.

The new question every company must answer

In the software era, the diagnostic question was: do we think of ourselves as a software company? Companies that answered no eventually found themselves disrupted by companies that answered yes.

In the design era, the question was: is design a strategic function or a service function in our organization? The difference between Apple and its competitors in the 2000s was not technology. It was that design sat at the table where product decisions were made.

The question for the AI era is this: is intelligence (the capacity to reason, synthesize, decide, and act) something we are building into the architecture of our company, or something we are treating as a feature?

This is the question that will sort the durable companies from the ones currently borrowing time. Not a question about tools or vendors or models. A question about organizational philosophy. About how you think about what your company actually does.

What comes next

I’m a founder. I build products. I’m working at the intersection of AI, product, and design. I’m not writing this as an outside observer. I’m writing it as someone who is, right now, in the middle of rebuilding a practice around the recognition that we have entered a new paradigm.

From inside that process: the transition is harder than it looks. Not technically. The tools are extraordinary and getting better at a rate that is genuinely disorienting. The hard part is conceptual. It requires letting go of the way you have organized your thinking about what a company is, what roles are for, what strategy means, what a product team does. These are deep assumptions. Built in the software era, some of them refined in the design era. They need to be re-examined again.

The cycle continues. Our knowledge is never complete. That is not a reason for anxiety. It is the most interesting thing about being alive and building things right now.

Every company is an AI company. Not because AI is the answer to every problem, but because refusing to engage with it seriously is now a strategic choice with consequences. And those consequences compound.

The companies that understand this earliest will not just survive the transition. They will define what comes after it. And the question of what comes after AI (what paradigm will eventually subsume this one) is, for now, delightfully open.