There Are Two GPTs. Your Company Only Knows One.

Does your company already have a Chief AI Officer? An “AI Center of Excellence”? An artificial intelligence roadmap with its own budget and dedicated team?

If the answer is yes, I want to tell you something that already happened before. Almost to the letter.

In 1997, IBM launched one of the most expensive advertising campaigns in its history: roughly $500 million invested in establishing a concept the company itself had coined a year earlier — “e-business.” The agency Ogilvy & Mather produced dozens of black-and-white commercials showing confused executives facing this new thing called the internet, and the underlying promise was simple: doing business online isn’t just another channel, it’s a transformation that needs its own strategy, its own team, its own budget. It worked. By 1999, more than 10,000 clients were paying IBM to help them build their “e-business strategy.”

Sound familiar? Replace “e-business” with “AI” and you have the corporate narrative of 2026.

This isn’t a forced metaphor. It’s literally the same movie, with the same script.

And here’s the core problem — the one almost nobody names: everyone is obsessed with the wrong GPT. Which model is better, which chatbot to integrate first, which launch is disruptive enough to justify the headline. That’s the GPT everyone knows: Generative Pre-trained Transformer. But there’s another GPT, far less sexy, that actually explains everything happening right now: General Purpose Technology. While companies fight over the first, they completely ignore the second. And it’s the second one that predicted, decades ago, exactly this moment.

A general purpose technology isn’t just another tool in the toolbox. It’s a layer that ends up permeating everything an organization does: how it sells, how it serves customers, how it makes decisions, how it operates. Electricity was one. The internet was another. And artificial intelligence, according to virtually all recent economic research on the subject, is following exactly the same path. That means the relevant question was never which tool is most disruptive, but something far less glamorous: how does that technology disappear inside every process until it goes unnoticed.

What happened to IBM’s “e-business”? The same thing that happened to every “internet team,” “digital strategy,” and “online department” that companies built between 1997 and the early 2000s. They didn’t fail. They dissolved. They dissolved because the internet became too fundamental to keep living in isolation on a separate team. Today no bank has a “Head of Internet.” Nobody in a boardroom asks what the company’s “internet strategy” is, because the internet stopped being a strategy: it’s the operating system for how the entire business runs.

That is exactly what is going to happen to your AI office.

Not because artificial intelligence is going to fail. Quite the opposite. Precisely because it’s going to work, and because it’s going to work everywhere at once, it will stop making sense to keep it isolated in a team with its own name. The fate of every “AI Center of Excellence” being built today in Latin America is the same as every “e-business team” in 1998: to cease existing as a separate entity — not because it failed, but because it ended up dissolved inside every department that once looked at it from the outside.

The problem is that, while that’s happening, many companies keep chasing the next sexy tool, the disruptive launch, the model that will change everything — instead of understanding the structural pattern that already determines how this story ends. They’re making the same miscalculation that traditional companies made nearly thirty years ago: treating technology as a project with a delivery date, instead of treating it as a transformation with no endpoint.

I have nothing against naming an AI leader, or building a team that accelerates internal learning. It’s probably necessary at this stage. The mistake isn’t starting there. The mistake is staying there — always chasing the next model, instead of understanding that this layer, sooner or later, has to disappear inside every decision the company makes.

This is the first column in a series where I’ll dismantle, week by week, the parallel between the internet wave of the ’90s and the AI wave of today: why chasing the most disruptive tool is the wrong question, why results take so long to show up, why small companies have an advantage they don’t know they have yet, and what replaces, in practice, the AI office your company has probably already built.

For now I’ll leave you with the question that opens the whole series: if internet departments disappeared, what’s in store for yours?