Focus
Nothing Has Changed in the Town of Bedrock
In the space of seven days in May 2026, the two most consequential AI companies on the planet both announced they were getting into consulting. The model alone is not getting the job done.
In the space of seven days in May 2026, the two most consequential AI companies on the planet both announced they were getting into consulting. Anthropic launched an enterprise AI services joint venture backed by Blackstone and Goldman Sachs. OpenAI raised $4 billion, acquired a London applied AI consultancy called Tomoro, and began embedding engineers directly inside client organisations. McKinsey and Bain & Company joined as investors in the same entity they will now partly compete with.
Model capability alone is not producing enterprise value at scale.
The Flintstones lived in a world where stone-age principles ran underneath modern-looking surfaces. Fred still had to run under the car to make it go. The town of Bedrock looked different. The fundamentals had not changed.
Enterprise technology has always worked the same way. ERP arrived in the nineties and promised to unify operations. The organisations that got value from it invested heavily in process redesign, change leadership, and sustained training. The ones that did not, got expensive shelfware and a decade of workarounds. Capability delivered. Value contingent on what you built around it.
McKinsey's 2023 research on digital transformation found that fewer than one-third of programmes delivered their intended value. The barrier was rarely the technology. It was the road around it: who owned which process, whether the measurement existed, whether leadership stayed long enough to see it through.
Anthropic and OpenAI have built extraordinary engines. What they are now discovering, at scale, is what every enterprise technology wave discovered before them. The impressive dashboard is on top. Fred is still running underneath.
The gap between AI capability and AI value is real. A composite picture from NHS deployments illustrates the pattern: an AI documentation tool goes live for clinical teams, adoption sits at 11% six months in, not because the tool fails but because change is hard. The capability exists. The delivery infrastructure does not.
Key to all of this is having a workflow owner, a training plan, and someone accountable for the answer.
Most AI investment conversations focus on return on investment. A better focus would be: does this make your people genuinely more capable?
Before buying another AI licence, ask whether a workflow owner, an adoption metric, a training plan, and a value review cadence exist.
The town of Bedrock has not changed. Neither have the rules for getting value from the technologies that arrive there.