Three tech cycles, one pattern: what Web 1.0, mobile and SaaS taught me before AI

I built my first web project in 2002. Back then, people earnestly explained to me that the internet was a fad, that the bubble had just burst, and that "serious" people wouldn't bet their careers on it. Twenty-four years later, I hear the exact same sentences, word for word, except now we're talking about AI. And it's precisely that repetition that interests me.

Because after living through three major technology cycles from the inside, Web 1.0, mobile and SaaS, generative AI, I'm convinced of one thing: the pattern is almost identical every time, and there's a lot to learn from it for anyone trying to navigate the current wave.

The pattern that repeats

Every major technology cycle I've lived through follows the same four-stage curve:

1. The skeptics phase. A new technology appears. A minority sees the opportunity. The majority shrugs — or worse, openly mocks.
2. The evangelists phase. A few spectacular cases emerge. The press goes wild. Gurus rise up. Everyone "does" the technology. Budgets explode.
3. The disillusionment phase. The promises don't hold. Projects fail. Many conclude it was overhyped, just a fad, that we should have been more cautious.
4. The operators phase. While others check out, a minority keeps building, quietly. Three to five years later, those operators have built a gap that's impossible to close.

That's exactly what happened with the web between 2000 and 2008. Exactly what happened with SaaS and mobile between 2010 and 2018. And it's exactly what's playing out before our eyes with AI since 2022.

The real winners are almost never first

Here's a counterintuitive observation that 24 years have hammered into me: the entrepreneurs who dominate a new cycle are almost never the spectacular early adopters.

The first to launch on the web in 1999-2000? Most went bankrupt. The ones who won? People who arrived in 2003-2005, who benefited from mature infrastructure and the lessons learned by others.

The first to bet big on mobile apps in 2008-2009? Many burned through their cash. The long-term winners arrived around 2012-2014, with better business models.

With AI, we're already seeing the same thing take shape. The "GPT-wrapper" startups launched in 2022-2023 with absurd valuations are starting to collapse. The real winners will be those arriving now, with a more mature understanding of the problem to solve.

There's a big difference between being early and being on time. The first position is romantic. The second is profitable.

What early adopters get wrong

I've often had this conversation with clients who said: "But François, if we're not there first, we lose the first mover advantage." That's a largely false belief, and it costs a lot.

The first mover advantage is real in only two cases: when there's a powerful network effect (Facebook, eBay), or when massive capital infrastructure is required (railroads, telecoms). For 95% of SMBs, that doesn't apply. What matters is the best mover advantage — arriving with superior execution in a window where the market is ripe.

The classic mistake: confusing the tool with the transformation

Every cycle, I see the same mistake repeat. Entrepreneurs buy the tool and believe they've done the transformation.

In 2003, SMBs ordered a website from me and believed they had become "digital." In 2014, companies subscribed to a CRM and thought they had a sales department. In 2025, leaders pay for ChatGPT Enterprise and imagine they have an AI strategy.

The tool is never the transformation. The transformation is what happens around the tool: the processes you accept to revisit, the people you train, the decisions you make differently. Without that work, the tool just becomes another expense. I developed this idea in more depth in what 24 years in digital taught me about AI.

What's really different this time

Now, I want to push back on my own analogy. The pattern repeats, but AI has something the previous waves didn't: the speed at which the cycle compresses.

The web took eight to ten years to become unavoidable. SaaS, six to eight years. Generative AI became critical for entire sectors in less than two years. The window between "this is new" and "you're behind" shrinks every cycle.

For someone who's lived through previous cycles, that changes the strategic calculation. You can no longer afford to wait three years to see what sticks. But you also shouldn't rush toward the first shiny tool. The discipline of arriving on time, neither first nor late, has never been more demanding.

Three principles that survive every cycle

If I had to condense those 24 years into three rules transferable from one wave to the next:

1. Start with the problem, never the technology

The worst question: "How can I use this new technology?" The right one: "What's my biggest bottleneck, and can this technology solve it?" That inversion changes everything, and it's also the first step I recommend for SMBs that want to start concretely with AI.

2. Build systems, not tricks

A brilliant ChatGPT hack is a trick. A documented, measured, integrated workflow is a system. Tricks vanish when the employee leaves. Systems stay.

3. Measure before and after

If you don't know where you were before, you can't prove the transformation worked. And without proof, the project dies at the first budget cut.

These three principles held in 2002, in 2014, and they keep holding in 2026. That's what makes me say: technology changes. The disciplines, they hold.

Frequently asked questions