You see anyone pointing out that AI "transformation" will take 10+ years, not the 2-3 of a limited project? Seems like an elephant in the corporate room. 🐘

AI transformation gets treated as a limited initiative, explicitly or, more often, implicitly, in roadmaps and in the idea of being "done".

AI isn't your normal tech "transformation"

But it's not like implementing an ERP. It'll be a problem when it takes longer and Boards finally demand real return on investment. How long can AI drag earnings down before it's a five-alarm fire? A quarter? A year? By 5% or 10% of earnings?

Whatever it is, it'll take loads of money and time:

Consider the evidence from prior tech revolutions: "[For] every $1 of investment in computer hardware, companies made $5-10 of intangible investments [e.g., rework and processes] … So far, however, there is little evidence" of that, finds The Economist. The historical multiplier may be off because AI isn't hardware. But still. I don't see a 5-10x investment happening in processes. Some in data. Not at that level. We're talking several years of investment.

The transformation will slow down

Why? Some of it is "just" execution of what’s well underway. But progress will slow down after the current wave.

Much of what we call "work" was never a formal "process." That capture will take gobs more time than well-defined processes.

Worse, John Cutler just called out that even many defined processes are merely "pointers to context". You gotta capture that context too to automate work.

And there’s a structural limit on the possible acceleration too: Fares Fourati et al recently found that a ton of work cannot be fully automated, mainly because work objectives are emergent. (I.e., no pre-set objective for AI to pursue on its own. They emerge through human work.) The resulting "human-AI co-construction" will have a maximum pace. Plus, we'll need to learn such ways of working to begin with.

And that's just stage 1. Once most companies "have" AI, the same thing will happen as when they first "had the internet:" Cost-cutters and efficiency fans will come in, and Optimization will be upon us. Once that’s stable, lots more becomes possible: Voilà, Expansion work, use cases built on more basic ones. And all of it will face sudden interruptions when new AI capabilities force mid-stream step changes. So another few years.

In the end, it's just the latest tech

Basically, I expect it'll take just as long to evolve companies for AI as it did for the internet and desktop revolutions. The tech speeds up in each case, leaving everything else as the bottleneck. We're talking years, with permanent (process) teams instead of project teams. And the costs will hit Board patience long before we reach some kind of end.

Am I missing something?

T.I.S.C.

Sources:

Economist research on intangible investments (requires paid account): https://www.economist.com/finance-and-economics/2026/07/28/ai-revenues-are-growing-fast-but-not-fast-enough

John Cutler’s piece “AI Needs Context, Not Tickets”: https://www.linkedin.com/pulse/ai-needs-context-tickets-john-cutler-y4zwc/ (Plug for him: Also subscribe to his excellent newsletter, The Beautiful Mess)

Fares Fourati, Hinrich Schütze, Eyke Hüllermeier, and Iryna Gurevych on “The Boundaries of Automation: A Theory of Persistent Human Participation”: https://arxiv.org/abs/2607.21547