Find "digital transformation" and replace with "AI"?

Here we go again

A few months ago I took part in a panel discussion on business transformation and artificial intelligence. The title was promising: "From technological promise to real impact." Reality was more modest. For most of the session, the conversation revolved around the usual suspects: use cases, tools, productivity, agents. AI as the solution. AI as the future. AI as inevitability.

When my turn came, I said something that shifted the mood in the room a little: that the agentic transformation we are being sold is not very different from that digital transformation which, in 80% of cases, ended up as mere digitization. That companies are using AI to automate exactly the same processes they already had, without asking whether those processes make sense in a world where AI exists. That we are repeating the mistake. There was a somewhat awkward silence, the kind that falls when someone goes off script. I suppose nobody expected an answer like that from someone who, coming from a technology consultancy, was supposed to be selling blind faith in the future.

The same old mistake, seen from the telcos

The thing is, I know this pattern well because I lived it from the inside. When telecommunications moved from circuit switching to packet switching, the change was not just technical. It was economic, strategic and existential. In a world of circuits, each call occupied a dedicated channel for the entire conversation, and was billed by the minute. In a world of packets, the marginal cost of carrying one more data packet was essentially zero. That didn't just change the infrastructure. It changed the cost structure, the billing model and, ultimately, the entire business model.

What did the telcos do? They digitized their networks. They moved from analog infrastructure to digital infrastructure. And they declared victory. Some went further and adapted operational processes, such as service provisioning, network management or customer service. But almost none of them sat down to think about what it meant for their business model that carrying data was essentially free. That on top of those zero-marginal-cost pipes, services they couldn't even imagine could be built. That Google, Netflix and Amazon were going to build empires on an infrastructure the telcos had built and paid for, without capturing a single cent of the value flowing through it.

It wasn't an external threat that defeated them. It was their own inability to read the change at the right layer.

The three layers

Every technological disruption hits at three levels, and their order matters. The first is strategy, where the key question is whether the disruption has the potential to change what my business can or should be. The second is processes, where the central point is whether the things I do, and how I do them, still make sense. And the last level is operational reality, the moment when it is worth asking what tools and infrastructure I use to execute it all.

And the typical mistake is to tackle it the other way around. In many cases as the product of a fit of nerves or the pressure not to be left behind. You start with the tool (the new technology, or the new infrastructure) and work your way up. You digitize the network without rethinking the process. You optimize the process without questioning the strategy. And when someone builds a completely new business on the same technology that you only used to do the same old thing a little faster, you wonder what happened.

The telcos did exactly that. They digitized the infrastructure layer. Some touched the process layer. Almost none reached the strategy layer. And those of us who tried (Terra among them, with all our mistakes) arrived too late or with too many ties to the traditional business.

The same mistake, 2026 edition

The same thing is happening with AI. Companies are rolling out generative AI tools (Copilot for Office, customer service chatbots, coding assistants...) without rethinking anything else. Which is nothing more than digitization under another name. The report nobody read gets automated. AI generates the twenty-page document that the recipient's AI will then have to summarize. Everyone gets thrown into "vibe coding" to churn out applications with no technical foundations, riddled with security holes, solving problems nobody had. People are hired with job ads asking for five years of experience in a technology that is two years old.

Once again, all of this is operating at the infrastructure layer and, at most, the process layer. The strategy layer, where the real challenges lie, such as what can I do now that was impossible before? which part of my business model is becoming obsolete? or where is the opportunity that exists only because this technology exists? remains untouched.

But this time three things make it worse

If the mistake were simply a repeat of what already happened with the Internet, with the cloud or with ERP, it would be enough to repeat the diagnosis and wait for history to correct itself. But there are at least three ruptures that make this disruption qualitatively different from the previous ones. And all three are being ignored with chilling nonchalance.

The first is cognitive substitution. Previous technologies automated processes, reinvented channels or took on manual tasks. AI is the first with the potential to replace cognitive work on a broad scale. Repeat slowly after me: cog-ni-tive. That changes the rules of the game in an unprecedented way. What happens when a middle manager oversees thirty AI agents and five people? What metrics do you use to evaluate their performance? How do you manage the working relationship with someone whose team consists mostly of machines? Nobody is thinking about it, because everyone is obsessed with deploying tools as quickly as possible.

The second is the talent pipeline problem. If AI eliminates the routine tasks that were historically handed to junior professionals (the first draft of the contract, the first check of the figures, the first line of code that doesn't work), you are eliminating the path that turned juniors into seniors. Professional experience is not a line on a résumé: it is the sediment left by a chain of mistakes corrected by someone with judgment. If you remove the pain caused by mistakes because the machine prevents them, how do you produce the professionals who, ten years from now, will have to make the decisions the machine cannot make?

The third is the speed at which competitive advantage becomes obsolete. In previous disruptions, implementation was the barrier: years of integration, trained teams, phased rollouts, obsessive KPI tracking. With AI, implementation can be a matter of weeks. That means the advantage doesn't come from having the technology everyone will have almost simultaneously, but from how fast you redesign your organization around it. And if the advantage from implementation is fleeting, the only sustainable thing is the capacity for continuous redesign. Do you think a license rollout solves that? What solves it is an organizational culture that most companies don't have, and don't know how to build.

The question nobody asked

At the panel, when I finished, an attendee asked me what his company should do. I answered with another question: if AI disappeared tomorrow, what would change in your business? If the answer is "almost nothing," then what you have is not an AI strategy. You have a subscription.

Real transformation begins when someone with strategic vision looks at the tool and doesn't ask "how do I do the same thing faster?" but "what can I do now that I couldn't do before?" That question isn't answered by a massive rollout of Copilot licenses across the whole org chart. It is answered by a small team with the capacity to reinvent, time to think and access to the right technology.

And there is a second question nobody is asking that should keep any HR director up at night: who is going to train the people who replace us, if we are eliminating the path along which today's decision-makers were trained?

The telcos held the infrastructure of the future in their hands and used it to keep doing the same old thing. Today, in AI, companies have the most powerful tool for reinvention that has existed since the Internet. The question is whether those who apply it will use it to reinvent or to digitize.

I know what the answer was last time. Care to bet it happens again?

This article is part of a series that began with "Less Wood, It's War!", and continues with "The end of the token open bar, or something more serious?", "What if the Genesis Mission were a huge mistake?", "Where's the pea?", "From packet to token, and back to square one", "Investing in typewriters", "Will the "killer app" arrive in time?", "Are they making room for us?", "Spot the 8 differences", "Will AI end up in a backpack?", "Will Nvidia be the new Kodak?", "Is CUDA Nvidia's magic potion?", "Neither conscience nor consciousness", "Neither Hawking nor Einstein. The true origin of intelligence", "Yogurts and tokens: welcome to the AI confusopoly", "The mathematician who left mathematics. What will we leave behind?". To be continued...