You have probably seen the headlines recently. OpenAI launched a $4 billion consulting subsidiary last week, Anthropic announced a $1.5 billion services venture with Blackstone and Goldman Sachs the week before, and Google Cloud has been quietly partnering with McKinsey, Deloitte, and BCG to embed engineers inside enterprise accounts.

Three companies that compete fiercely on model capability all made the same move in the same month, so that must be worth paying attention to.

What is actually going on

The approaches differ, but the direction is the same.

OpenAI has created a majority-owned subsidiary called the Deployment Company, backed by $4 billion of initial investment and already valued at $10 billion. They acquired Tomoro, a UK-based AI consulting firm, to have experienced consultants on the books from day one.

Anthropic has partnered with Blackstone, Hellman & Friedman, and Goldman Sachs to launch an AI services company with $1.5 billion in committed capital. The model is their own engineers embedded inside client organisations, building solutions around how the business actually operates.

Google Cloud has gone the channel route, partnering with McKinsey, Deloitte, BCG, Accenture, and Bain, giving them early access to upcoming Gemini models and placing specialist engineers into client accounts through those partnerships.

All three are following a similar approach to Palantir, the US data analytics firm that built its business by putting engineers inside client organisations rather than shipping software and leaving implementation to the buyer.

Why are they doing this

What makes this significant is that these companies have the most capable AI technology on the planet, and yet they have all arrived at the same conclusion: selling the technology is not enough.

They can see from their own customer data that businesses are buying AI and struggling to make it work. The demand is clearly there. But there is a gap between purchasing the technology and getting value from it, and that gap is large enough that the labs think it is worth billions to address.

It is no surprise that the demo always looks good when someone spends an afternoon with clean data and a well-crafted prompt, and it is impressive. The problems start in the real world where data sits across systems that do not talk to each other, processes have exceptions that nobody documented, and the team doing the work was never consulted on how the tool should fit into their day.

The labs have watched this pattern repeat across thousands of enterprise accounts and decided the answer is to send their own people in. If the organisations that built the technology believe you need specialist help to implement it, that is a signal worth taking seriously.

What this tells you

The demand signal is clear. OpenAI’s enterprise revenue already makes up more than 40% of their total income and the market for implementation support is real and growing fast.

More importantly for anyone running a business: the companies that created GPT, Claude, and Gemini are telling you, with billions of pounds behind it, that getting AI implementation right requires proper expertise. The choice of who helps you deploy AI matters at least as much as the choice of which AI to deploy.

Where this leaves most businesses

The catch is that these new services operations are largely built around enterprise economics. OpenAI’s subsidiary is structured for major accounts. Google Cloud’s partners are the large strategy and advisory firms, where programmes can quickly run into six or seven figures. Even Anthropic’s venture, which references mid-size businesses, is backed by the kind of capital that suggests a very different definition of “mid-size” to most UK operators.

For many businesses, particularly those with 50 to 500 people, real operational complexity, and tight margins, an embedded OpenAI engineer or a traditional consultancy-led transformation programme is simply not realistic.

But even larger organisations are increasingly questioning whether AI implementation needs to come with the overhead, slow pace, and fee structures that large consultancies typically bring.

The underlying point still applies either way. If the labs themselves believe businesses need specialist implementation support to get value from AI, they probably do. The challenge is finding a partner that understands operations well enough to identify where AI creates real value and can deploy it at a pace, level of pragmatism, and commercial model that fits the business.