The AI Trap: why technology alone is not transformation.

I have had a version of this conversation more than once this year. "We rolled out Copilot to a hundred people six months ago. Adoption is at seventy-eight percent. Our workflows haven't changed. What did we miss?" The people asking are not naïve. The budgets work, the teams use the tools, the dashboards glow green. And yet the business they were promised — the one where AI redesigns how value gets created — has not arrived.

That gap is the whole subject of this piece. The last one ended on a hard edge: a mirror will show you what your business actually is, but seeing is not the same as acting. This is the acting. And the reason the acting so rarely happens is almost always the same.

They bought the bike. They never built the legs.

Bought is not adopted. Adopted is not capable.

A short, honest aside on why I reach for that image, because it is not decoration. I have finished the Ironman distance three times — 3.8km swum, 180km on the bike, 42.2km run, in that order, in one day. That distance is supremely uninterested in your equipment: the lightest frame ever made will not carry you to the line if you have not put in the hours. So I read business performance partly through that lens, and I would rather say so plainly than pretend it is a framework I found in a book. Resilience is conditioning. There are no shortcuts — only the work. From here I will mostly speak the language the boardroom speaks — capability, operating model, decision rights — but the idea underneath stays the same.

The trap is quiet, and almost everyone walks into it: mistaking the purchase for the outcome. Acquiring a technology and being changed by it were never the same thing. We learned this in the digital wave. We are about to learn it again — faster, harder, at higher cost.

It helps to separate three states, rather than collapsing them into one. Equipped: the licences are provisioned, the platform is live, the vendor is paid. Active: people use it daily, and the adoption metric looks healthy. Capable: the way the business decides, operates and creates value has been rebuilt around what the tool makes newly possible — so it can now do things it genuinely could not before. Most companies reach Active, log the number, and declare victory. But high usage of a tool is not a transformed organisation. Two firms can hold the identical licence and sit on opposite sides of that line: one writes its emails faster; the other has become a different, more capable business. Tech-enabled is not tech-savvy — and the difference is not a feature you can buy.

But notice what that ladder quietly assumes: that there is a top rung. There isn't. Capable is defined against what the tools make possible today — and the tools, and the possible, keep moving. The same thing happens in sport: a state of the art holds right up until someone climbs past it, and then it is merely the new baseline. The sub-eight-hour Ironman was once the outer edge of the imaginable; the two-hour marathon barrier — the Schallmauer that stood for a generation — fell this year. The summit does not stay put. It recedes the moment you reach it.

Which raises the obvious question — what is next? — and here is where it gets interesting, because that question is far too valuable to leave lying around. The incumbents claim it fast. Microsoft has already named the firm on the far side of this work: the Frontier Firm, an organisation "structured around on-demand intelligence" and "hybrid teams of humans and agents," which — they assure us — every company will be on its way to becoming within two to five years. It is a good phrase. I will use it. I will also note, in the spirit of the Jester whose whole job is to say the quiet part, that it arrives in the very same announcement as the agents, the Agent Store and the Copilot upgrades you will need to buy to get there. Naming the next summit and selling the climbing equipment is a time-honoured move, and a profitable one. Whether the Frontier Firm is the real horizon or a well-drawn sales map is a debate worth having out loud — and you should have it before you sign. The one thing not up for debate is the direction of travel: the bar keeps rising, and it is rising faster. So whatever you call the next stage, the response that works is the same one that always worked. Up your training regimen.

The digital wave — an opportunity mostly missed.

What I think of as the digital decade — roughly 2010 to 2020 — was the largest enterprise technology adoption cycle since personal computing. Companies bought content platforms, marketing automation, CRM systems, the lot. Most implementations were technically successful. Most transformations were not.

Nicholas Carr had already explained why. In The Big Switch he traced how computing was becoming a utility — drawn from the wall like electricity, available to everyone on identical terms. And a utility, by definition, confers no advantage simply by being switched on. The strategic value was never in having the technology. It was in what you reorganised around it. The equipment commoditises. The capability does not.

I watched this play out at Allianz. The data was already there; the platforms worked. What was missing was the reframe of digital from a marketing cost line into a P&L position — a change not of tooling but of how the organisation was structured to create and capture value. Sixteen new revenue streams in twelve months came out of that reframe, not out of any new system. The technology was the occasion. The redesign was the work.

And the pattern is not anecdotal. Three years into the AI cycle, the broad surveys of corporate adoption are still struggling to find the productivity it was meant to unlock — the same flat returns the digital decade produced, and for the same reason. The capability did not arrive because nobody built it. It does not ship with the licence.

The reason transformations fail isn't the technology. It's the change.

This is where the work actually lives — and where David Teece's whole career points. Durable advantage comes from an organisation's dynamic capabilities: its ability to sense a shift, seize it, and reconfigure itself around it. The technology is available to everyone on the same terms; the capacity to reorganise around it is not. That is where the advantage hides.

John Kotter spent decades cataloguing why change efforts collapse, and almost none of his reasons are technical. They fail because urgency is manufactured rather than felt; because the new way of working is never anchored into how the organisation operates; because leadership declares victory the moment the tool is live and stops doing the unglamorous work of rewiring everything around it. Read those failure patterns next to a typical AI programme and the resemblance is uncomfortable: the platform goes in, the dashboard turns green, the steering committee disbands — and the operating model it was meant to change is exactly where it started.

This is the part procurement cannot help you with. There is no vendor for decision rights, no demo for governance, no comparison matrix for the way a business has organised itself for a decade. The redesign lives in the places software never touches — who decides, who is accountable, how value moves — and those are the places that decide whether the technology amounts to anything at all.

What it looks like when it actually happens.

DarkTools came to me with what looked like a website problem. They were on WordPress and WooCommerce, the database was buckling under a complex B2B product catalogue, updates were failing. The brief was a redesign — a better website. New equipment.

We never got to talk about the website. Or rather, I refused to, for about two weeks, because the website was the wrong subject. The real question was not what platform do we need; it was what business are we actually in? And the honest answer reframed everything: their edge was not a prettier storefront — it was the prospect of becoming best-in-class at giving a demanding B2B audience the supreme product information they need to specify their projects. Get that right and you do not have a better website; you have a better offer in your category. That single shift — from tool-first to model-first — dictated everything that followed:

Business case and data logic first. Platforms and tools follow. That is the entire argument of this piece, spoken by a client's situation rather than by me.

And here is the part that ties back to where the last piece left off. Notice that none of the three moves above is, strictly, an AI move. That is not an accident — it is the sequence. You do the homework first: you know what business you are in, you get your data honest, you build the centre of truth. Then — and only then — applying AI at scale becomes the thing you are finally ready for, rather than the thing you reach for to avoid the homework. That is the mirror's lesson made operational: see yourself clearly, build the business back around what you see, and the wand has something worth amplifying. At DarkTools that next chapter is loading now. More on it soon.

Transformation is not a sprint. It's the longest race there is.

So you have understood the trap. You know the bike alone is not the point, and you have started — slowly, expensively, properly — to build the legs. There is one thing left that nobody tells you, and it is the thing that actually decides if you finish the race.

Pace.

Here is the orthodoxy of our age, and I want you to hear how seductive it is: move fast. Fail fast. Rapid-prototype your way to the future. Ship, learn, repeat. It sounds like courage. It sounds like exactly what a bold company should do with a powerful new tool. And on the AI bike leg, it is the single most reliable way to blow yourself up before you ever see the finish line.

Because here is what "move fast" forgets: there is a marathon after the bike.