Anthropic and OpenAI are opening consultancies: what a UK buyer should ask, and where independents fit

What changed

On 23 October 2026 the Financial Times reported that Anthropic and Hg, the London-based private equity firm known for software buyouts, have agreed to back a new independent AI consultancy in Europe. According to the report, as carried by City AM and Dealroom, the firm is expected to launch by the end of the year with hubs in London and Berlin, will embed AI consultants and engineers inside European businesses, and will target B2B software companies and other white-collar organisations alongside Hg's own portfolio. Anthropic would fund it from its balance sheet; Hg would invest from the funds it manages, including the London-listed HgCapital Trust. Anthropic was approached for comment and Hg declined. Nothing has been published on a name, leadership, headcount, pricing or legal structure, so every detail above is reported rather than confirmed.

It is not a one-off. On 4 May 2026 Anthropic, Blackstone, Hellman & Friedman and Goldman Sachs announced a standalone enterprise AI services company with Anthropic engineers embedded in its team. The Wall Street Journal, cited by Fortune, put committed capital at about $1.5 billion, with General Atlantic, Apollo, GIC and Sequoia among the other backers. Later reporting named it Ode, said it launched with about 100 engineers after absorbing the services firm Fractional AI, and described a Claude-first approach that will use other models where a design requires it. Anthropic's chief financial officer, Krishna Rao, gave the rationale at the time: enterprise demand for Claude was significantly outpacing any single delivery model.

A week later OpenAI did the same thing at larger scale. The OpenAI Deployment Company, launched on 11 May 2026, is majority-owned and controlled by OpenAI, started with more than $4 billion of initial investment led by TPG with Advent, Bain Capital and Brookfield as co-leads, and counts McKinsey & Company, Bain & Company and Capgemini among its investors. OpenAI also agreed to acquire Tomoro, an applied AI consulting firm whose clients have included Tesco and Virgin Atlantic, bringing about 150 forward-deployed engineers into the new company, subject to closing. OpenAI's description of a typical engagement is a diagnostic, a small number of priority workflows chosen with the client's leadership, and then engineers working inside the organisation to build and deploy production systems on OpenAI models.

Behind all three is a labour-market fact. CNBC, citing Lightcast, reported on 15 October that US job postings for forward-deployed engineers rose from fewer than 200 in 2023 to more than 7,300 in the first nine months of 2026, with Deloitte and Accenture the largest advertisers. AWS has committed $1 billion to its own forward-deployed engineering group. Anthropic has committed $100 million to train 10,000 such engineers by the end of 2027, with its first cohorts drawn from Accenture, Bain, Capgemini, Deloitte and McKinsey. The model makers have concluded that the shortage of people who can put a model into a business is the constraint on their revenue, and they are buying and building that capacity themselves.

Why it matters for UK business

The first thing to be clear about is who these ventures are for. Every one of them is built on a private equity partner, and private equity partners bring portfolio companies: Hg already has around 150 AI specialists working across its portfolio, according to the FT, and the new firm is reported to work alongside them rather than replace them. OpenAI says its investment partners sponsor more than 2,000 businesses between them. Ode's chief executive has said it targets projects a client CEO personally owns. The engagement shape described in OpenAI's own post, a diagnostic followed by embedded engineers rebuilding priority workflows, is an enterprise programme. None of this is aimed at the UK SME deciding whether to roll Claude Team out to a finance function, and nothing in the reporting suggests it will be.

The second thing is what has changed for buyers who are in scope. Until this year the advice on which model to adopt came from a party that did not own the model: a systems integrator, a consultancy or an in-house team. Now the owner of the model is also, through a majority-owned or balance-sheet-funded affiliate, the firm offering to deploy it. Ode's stated Claude-first default is the honest version of this. A consultancy funded by one vendor will reach for that vendor's model unless a design forces otherwise, and that is a reasonable commercial position for it to take. It is not a reason to avoid such firms; it is a reason to contract with them as what they are, which is a channel, and to ask the channel questions.

Third, the move confirms something our explainer on the Claude Partner Network has argued since May: the partner tier a firm holds tells you less than who its named practitioners are and whose interests they serve. Anthropic's own consultancies will, presumably, sit alongside the firms it certifies and courts as implementation partners, including the large consultancies now investing in the rival OpenAI venture. City AM noted Accenture's shares fell more than 4 per cent on the day of the FT report. The interesting question for a UK buyer is not which brand wins that fight. It is whether the people advising you have a reason to prefer one answer before they have seen your data.

Fourth, there is a genuine upside. More deployment capacity in London is good for UK adoption, and the forward-deployed model, done well, leaves a client with working systems rather than slide decks. The failure mode is the one every embedded-engineering model has had since Palantir popularised it: the system works while the engineers are in the building and decays when they leave. That is a contracting problem, and it is solvable in the terms.

What to do, and what not to do

Do, if you are talking to a model-maker-backed consultancy:

• Ask for its model-neutrality policy in writing, and for two examples of engagements it delivered on another vendor's model. Ode has said it will use rivals where a design requires it; ask how often that has happened.
• Ask where client data is processed by the embedded engineers' own tooling, whether any of it can reach the parent company, and how that squares with your data-protection assessment. Our guide to running a DPIA for an AI assistant rollout covers the questions.
• Ask how a conflict between your interests and the parent's roadmap or pricing is handled, and who decides. An independent legal entity helps on paper; the governance documents show the reality.
• Write handover into the contract: documentation, code ownership, and a demonstration that the systems run without the consultancy's people on site.
• Agree how outcomes are measured before work starts, in your numbers rather than the vendor's case studies.

Do not:

• Treat the Hg venture as launched. It is a report, Anthropic has not confirmed it, and nothing is known about its leadership, pricing or structure.
• Assume a vendor-owned consultancy is the safer choice because it owns the model. Owning the model is the conflict, not the reassurance.
• Delay a decision that is ready to be made in the hope that the vendors' firms will serve SMEs. Nothing in the reporting says they will.

Where The AI Consultancy fits

We are the other kind of firm. The AI Consultancy is Certified in the Claude Partner Network, which gives us Anthropic's training and the Claude Certified Architect credential; it does not give us, and we do not take, any referral, reseller or deployment income from Anthropic, OpenAI or anyone else, as our independence page sets out. Our Claude consulting service exists for the UK SMEs and mid-market firms the new ventures are not built for, and our guide to choosing a Claude consultancy gives the checklist we would want applied to us. Where Claude is the wrong tool, we say so, and we have the ChatGPT and Gemini practice to prove it.

Verified on 24 October 2026 against the Financial Times report of 23 October 2026 as carried by City AM and Dealroom News the same day, Blackstone's press release of 4 May 2026 and Fortune's report of the same date, AI Chat Daily's report of 15 July 2026 on Ode, OpenAI's "OpenAI launches the OpenAI Deployment Company" (11 May 2026) and CNBC's report of 15 October 2026 citing Lightcast. The Anthropic and Hg venture is reported, not confirmed; we will update this briefing if either company publishes a statement. This briefing is general information, not procurement or legal advice.