AI will probably not replace humans. But the belief that it can will destroy thousands of companies.
It will not be the first time.
The AI era has not democratised entrepreneurship. It has democratised the illusion — that is the reading I have held since the start of this cycle, and what the files I see passing through confirm: SMEs with €5M to €50M in revenue, executives with an "AI-powered" deck and no documented processes, founders convinced a tool will compensate for twenty years of structural debt.
Tools cost less. The real competence gap — the kind that holds when the founder is gone, the kind that survives a crisis cycle — is widening.
For fifty years, every technological revolution has told the same story. Computing was supposed to kill paper. The internet was supposed to kill stores. The cloud was supposed to kill servers. MOOCs were supposed to replace schools. Social media was supposed to replace media. Today, it is artificial intelligence.
The pattern is identical: take an accelerator, promote it as a replacement. A few years later, discover it only eliminated the projections we placed on it. Paper did not disappear. Neither did stores. Servers still run — elsewhere, differently, less visibly.
AI will probably follow the same trajectory. That is precisely why it will profoundly reshape the economy.
THE PRESENT: The replacement fantasy
Every week, an announcement promises to eliminate entire professions. Valuations reflect that belief. In the field, the dynamic is more prosaic: executives who want to cut payroll before stabilising their processes, founders who confuse execution speed with structural solidity, teams using ChatGPT without governance and producing outputs nobody reviews.
The problem is not AI. It is the narrative around AI.
Coffee, or how to step outside the AI narrative
To understand why, take a deliberately absurd example.
Coffee.
Nobody wakes up in the morning saying: "I no longer need coffee. I have ChatGPT." Nobody. Because the idea is ridiculous.
AI can compare coffees. It can recommend a roaster. It can optimise a plantation. It can find the best suppliers. It can write a marketing campaign. It can sell coffee.
But it does not produce the coffee. It does not grow the beans. It does not harvest the plant. It does not heat the water. It does not fill a cup. It does not replace the fundamental need we have to drink a coffee.
It accelerates everything around coffee. It never replaces coffee.
And yet, we find it perfectly credible that it could replace a human being. A human being is infinitely more complex than a cup of coffee.
The law of fundamental needs
AI replaces no fundamental human need.
Coffee. Water. Food. Sleep. Transport. Housing. Clothing. Health. Education. Relationships. Love. Trust. None.
A technology can accelerate the satisfaction of a fundamental need. It almost never eliminates the need itself.
Coffee is only the first example on that list. It is not an amusing metaphor — it is a systemic law. And that is where what I call the Substitution Illusion begins: we confuse an accelerator with a replacement.
What I have seen recently
An executive — services SME, roughly €12M in revenue — had terminated three vendors and frozen two hires. "AI does the work, we no longer need all these people." Six months later, he calls back. Not because the tools failed. Because he had removed the human layer that held the system together: deliverable review, client relationships, file memory, implicit quality control that nobody had documented because "it just happened."
That is one case. The pattern is general: we replace what we do not yet understand we still need to keep.
The Substitution Illusion
Definition
The Substitution Illusion: the belief that a technology replaces the element it accelerates, when in fact it only increases its capabilities.
This is not a conference metaphor. It is the name I give to a pattern found across five technological cycles — and that I see reproducing at accelerated speed with AI. Formalising it serves one purpose: recognising it before making a restructuring, layoff, or vendor termination decision.
See the full definition in the System Lexicon →
The real risk
The danger is not AI. It is the decision made under the effect of fantasy. Some companies will lay off too early, cut consultants, experts, subcontracting SMEs, artisans — convinced a SaaS subscription and a language model are enough. Others will refuse any integration and fall structurally behind. Both extremes are costly.
Technology never destroys a company directly. Human choices made in the belief that it can do alone what it only accelerates — that is what destroys companies.
For the operational dimension — the 85% of AI projects that never reach production, the four dominant failure modes — see: Why thousands of companies will die from AI automation.
THE PAST: The Technology Cycle Law
All major technologies go through the same cycle. I call it the Technology Cycle Law — not because it is a physical law, but because the pattern is stable enough to serve as a reading grid:
← AI today sits between phases 2 and 3
Technology becomes indispensable after phase 7 — not because it replaced anything, but because it merged into the system.
Five cycles, one pattern
| Revolution | Promise | What happened |
|---|---|---|
| Computing (1980s) | Paperless office | More paper than before — processes ten times faster |
| Internet (2000s) | End of stores | Retail reorganised, not eliminated |
| Cloud (2010s) | End of servers | Costs spiked, failures, then de facto standard |
| Smartphone (2010s) | Pocket computer replaces everything | New ecosystem, not substitution |
| AI (2020s) | End of human work | Phases 2-3. Phase 4 is approaching. |
The cloud is the most recent and instructive example for French executives. In 2012, "migrating to the cloud" was sold as a revolution. Costs exploded. Projects failed. IT directors lost confidence. Then tools matured, prices stabilised, skills professionalised. Today, the cloud is infrastructure — nobody sells it as a revolution, everyone uses it.
AI is at the "cloud migration 2014" stage. That is worth knowing before betting the organisation on it.
THE FUTURE: Where value is recreated
The disillusion phase will arrive. Companies will discover that AI is expensive, requires governance, produces errors, and needs humans to validate what it generates. Investors will tighten criteria. Narratives will become measured again.
Then maturity will come — and that is when AI will become genuinely useful. Not by replacing humans, but by becoming an invisible layer, like the cloud or the internet today. Companies will no longer say "we use AI." The company will function, and AI will be inside it.
The sorting will be brutal and predictable: total refusal → lag; total replacement → frequent collapse; leverage + human → reinforcement.
Companies that emerge stronger will keep what creates their value — relationships, trust, field execution, capacity to decide under uncertainty — while accelerating what can be accelerated.
Displacement, not suppression
Technological revolutions almost never replace existing systems. They shift the centre of gravity of value creation.
The internet did not kill physical commerce: it shifted value toward logistics, in-store experience, client data, brand. The cloud did not eliminate infrastructure: it shifted it toward integration, security, architecture, consulting.
With AI, the shift is already visible: data governance, output quality control, process design, augmented client relationships, regulatory arbitrage. These are not jobs "replaced by AI." They are jobs created by the displacement.
Those who seek to replace the human look at technology. Those who seek to understand where value is recreated look at the system. The patterns are readable — if you know where to look.
The day someone actually replaces coffee with artificial intelligence, I might start believing it can replace a human being.
Until then, it does what all great technologies have always done.
It augments the world.
It does not substitute for it.
In summary
AI does not erase reality. It accelerates our capacity to act on it. The nuance changes everything.
Technological revolutions shift the centre of gravity of value creation. They rarely replace the fundamental needs that structure our economy.
Understand the systems
The System Economy decodes how economies and companies actually work. Explore more analysis on AI, SMEs, and technology cycles.
Explore Insights →
