In eighteen months, three of the largest structural bets in the AI industry were placed on the same thing. Not on better models. On the work of getting existing models into companies that need them.
If you want to understand where enterprise AI actually is in 2026, that pattern tells you more than any benchmark.
Three announcements
February 2026: OpenAI’s Frontier Alliances. OpenAI announced multi-year partnerships with McKinsey, Boston Consulting Group, Accenture and Capgemini. The division of labour is explicit. McKinsey and BCG handle strategy and operating-model work. Accenture and Capgemini handle implementation, wiring the technology into existing enterprise systems. OpenAI’s own forward deployed engineers work alongside them, and the partner firms build certified practice groups with early access to models before general availability.
May 2026: The Deployment Company. OpenAI raised over $4bn from 19 outside investors for a joint venture valued at $10bn before the money, with TPG leading alongside Brookfield, Advent and Bain Capital. OpenAI keeps majority ownership and control through super-voting shares. Its purpose is to embed engineers inside enterprise customers and run complex multi-team deployments. To staff it quickly, OpenAI acquired Tomoro, a London consultancy founded in 2023 whose client list includes Tesco and Virgin Atlantic.
July 2026: Ode. Anthropic, Blackstone and Hellman & Friedman launched Ode on 15 July, joined by Goldman Sachs, General Atlantic, Leonard Green, Apollo, GIC and Sequoia. It is led by Chris Taylor as chief executive and Eddie Siegel as chief technology officer, who co-founded Fractional AI, the applied-AI services firm acquired earlier in the year to form Ode’s operational core. The stated model is teams that partner closely with chief executives to define and execute priority AI initiatives, aimed at mid-size companies moving from experiment to operation.
What the pattern means
Read those three together and the argument is hard to miss. The organisations with the best possible view of what frontier models can do have concluded that the constraint is no longer the model.
That is a genuine shift. For three years the assumption was that capability was scarce and deployment was routine. If capability were still the bottleneck, this capital would be going into research. Instead it is going into engineers who sit inside client organisations, learn how a specific business works, and build in that context.
It also explains something that confuses a lot of business owners. You can subscribe to the same models as a FTSE 100 company, for a few pounds a month, and get almost nothing from them while they report real gains. The difference was never access. It is that somebody spent months connecting the model to their actual data and their actual processes, and nobody did that for you.
The part nobody says out loud
Look at who these ventures are for.
Ode’s own language is mid-size companies. Frontier Alliances pairs OpenAI with four firms whose smallest typical engagement would exhaust most SMEs’ annual technology budget. The Deployment Company is built for complex multi-team deployments, which is a phrase that describes organisations with multiple teams.
Every tier of this new industry is aimed at companies larger than the overwhelming majority of British businesses. There is no version of Ode coming for a 30-person firm in Swindon. That is not a criticism of Ode, which is doing something sensible for the market it chose. It is an observation about which market got chosen.
What that actually means for a smaller business
The useful conclusion is not that you are being ignored, though you are. It is that the method being sold at that scale is not exotic, and most of it works at a smaller one.
Strip away the capital structure and the delivery model is consistent across all three. Someone technical goes into the business. They work out where the value actually sits, which is almost never where the demo suggested. They build against the company’s real data and real edge cases. They stay long enough to fix what breaks in contact with reality. Then they measure whether the thing worked, and they are accountable for the answer.
None of those steps requires a billion dollars. They require someone senior enough to make the judgement calls and technical enough to build the result, which is a combination that has been hard to buy at SME scale for a long time. That is the actual gap, and it existed before any of these ventures launched.
The objection
You may reasonably be thinking that this is a convenient argument for a firm that sells exactly that. It is, and pretending otherwise would be silly.
So here is the version that does not depend on trusting us. The three announcements above are public. The language about mid-size companies is in Ode’s own press release. The consulting partners in the Frontier Alliances are named by OpenAI. You can read all of it and reach your own view about whether anything in that market is aimed at a business your size. We think the answer is obvious, but the evidence is not ours.
What we would push back on is the conclusion some owners draw from it, which is that AI is an enterprise story and they should wait. Waiting was reasonable when the models were changing monthly. The models are now the stable part. The unstable part is whether your business has anyone able to connect them to the work, and waiting does not fix that.
What to do about it
Stop evaluating tools and start mapping tasks. The pattern in every one of these ventures is that value is found by understanding a specific business, not by selecting a product. Our guide to finding your first real AI use case is the small-company version of the same exercise.
Ask who is accountable for the outcome. The reason forward deployment works is that the engineer owns whether the problem got solved, rather than whether the specification was met. If you are commissioning AI work, that distinction is the one worth negotiating over. We wrote about how to tell whether an AI system actually works for the buyer’s side of that conversation.
Check whether your systems can be reached at all. The most common blocker we find is not the AI. It is that the data lives in four places and none of them have an interface. That is a plumbing problem, and it is solvable, but it is the prerequisite rather than the project.
Frequently asked questions
What is a forward deployed engineer?
A software engineer embedded inside a client’s organisation who builds working systems in the client’s own environment and owns the outcome end to end, rather than delivering recommendations or building at arm’s length. We cover the role in detail in our plain-English guide.
Why are AI companies investing in services rather than research?
Because the constraint has moved. Getting value from AI inside a specific business depends on that business’s data, processes and edge cases, which cannot be shipped in a product. The capital is following the scarce input, and in 2026 the scarce input is people who can do that integration work.
Does any of this apply to a small business?
The ventures do not, since all three target mid-size companies and above. The method does. Mapping tasks, building against real data, measuring the outcome and staying to fix what breaks are not scale-dependent practices.
Is AI worth it for an SME yet?
It depends entirely on whether you have identified a specific, repetitive, measurable task rather than an ambition. Businesses that start from a task tend to see returns. Businesses that start from a tool tend not to.
Flux Dynamics is a fractional CTO who builds, working with UK businesses that are too small for the firms above and too ambitious for a template. Tell us what you are trying to fix and we will tell you honestly whether AI is the answer.