The Last Wave Is Adoption

The technology is moving at a speed with no precedent. The companies meant to use it are moving at a crawl. That gap is the whole opportunity.

There is a strange asymmetry running through the current technology cycle. The technology moves at a pace with no real precedent, and the organisations meant to use it move at the pace they have always moved, which is to say barely.

Every few months a model arrives that makes the previous one look quaint. What was a research demonstration two years ago is now a commodity call priced by the token. And yet walk into most large companies and you find the same committees, the same procurement cycles, the same pilot that has been running since last spring and will keep running until it is quietly retired without anyone admitting it failed. The capability is sitting on the shelf. Almost nobody is using it properly.

This gap is the entire story. It is also, for anyone paying attention, the opportunity.

The comfortable assumption inside most boardrooms is that AI is a thing you will get to, a project for next year's budget, a workstream to be owned by someone in IT. That assumption is wrong in a way that will prove expensive. The firms that win the next decade will not be the ones that bought a chatbot licence and declared victory. They will be the ones that rebuilt their processes around the technology, embedded it into how work actually gets done, and treated it as infrastructure rather than a feature. The rest will spend a few years insisting they are fine, and then they will not be.

None of this is a prediction about the technology. It is a prediction about behaviour, and behaviour is slow.

Where the value goes next

It helps to look at where the money has already gone, because it tells you where it is going.

The first wave was silicon. The chips. One company in particular turned the training of large models into the most valuable franchise of the decade, and the market rewarded it accordingly. If you needed to compute, you paid the toll.

The second wave was the physical layer wrapped around the silicon. Data centres, and the energy to run them. The unglamorous realisation that intelligence at scale is, in the end, a power problem. Grid capacity, cooling, land, and utilities. The picks and shovels behind the picks and shovels.

The third wave, the one we have lived through most recently, was the models themselves. The large language models that turned raw compute into something that could read, write, reason, and occasionally embarrass itself in public.

Each of these waves created enormous value. Each was also, in its own way, a race to build the engine.

The next wave is different, and we think it is the durable one. It is the application layer. It is what you actually do with the engine now that it exists. Agentic systems that carry out real work, software that does not merely answer a question but completes a task, tools that fold into a business and quietly remove cost from it. This is where efficiency gets manufactured. This is where a company of fifty does the work that used to need three hundred.

The engine was the expensive part to build. The lasting profit is in the use. Adoption and utilisation, the boring words, are where the next decade of value creation will sit, precisely because they are the part almost everyone is currently too slow to take seriously.

The firms that understand this are not waiting for permission. Everyone else is running another pilot.