Anthropic and Hg, the London-based private-equity firm, have agreed to fund an independent AI consulting company together, with London and Berlin as the first two hubs and a launch planned before the end of the year, according to the Financial Times, which reported the agreement on Friday.
The vehicle will be capitalized from both sides. Anthropic is putting in its own money, and Hg will commit through one of its funds, people familiar with the arrangement said. The company is being set up as an independent business rather than a division of either owner. Its staff will be engineers and advisers placed inside client organizations, with the job of moving AI systems from demonstration to production.
For Anthropic, the project is a first in Europe, the people said. For Hg, whose portfolio is concentrated in software and business services, it is a way to sell AI work to companies the firm already controls and to the customers those companies serve.
The commercial reasoning is not complicated. Access to a frontier model has become close to a commodity at the low end of the market. What companies still struggle to buy is the middle of the work: data pipelines, permissions, evaluation sets, the rewiring of a claims process or a procurement workflow so that a model can be trusted to touch it.
Systems integrators have moved into that gap. So have the large accounting and strategy firms, which built AI practices and staffed them with thousands of consultants over the past three years. Their pitch to a board is that the model is one input and the process change is the product.
A model developer chasing enterprise revenue has two routes. It can license the model and let partners carry the implementation risk, or it can take a stake in the implementation itself. Anthropic has done both. It has agreements with consultancies and cloud providers that resell its models, and it sells Claude directly to large customers through its own sales force. The Hg venture adds a third channel, part-owned by the model maker and part-owned by an investor with a portfolio to upgrade.
The geography follows the money. Hg is among Europe’s most active software investors, and its holdings skew toward mid-market business-software companies in Britain, Germany and the Nordic countries, many of them old enough to carry technical debt and manual processes that predate cloud computing. Berlin also puts the venture in the market where domestic model developers, Aleph Alpha among them, have argued that European companies should keep sensitive data with European vendors.
Customer demand is real but uneven. Banks, insurers and manufacturers have run pilots for two years; far fewer have put systems into production with a budget line attached. Analysts said the gap between pilots and production is where consulting revenue now sits, and where model vendors lose deals to firms that can promise delivery rather than a demonstration.
The regulatory clock adds urgency. The European Union’s AI Act puts documentation, testing and human-oversight duties on systems used in hiring, credit scoring, medical triage and other high-risk settings, and compliance work is exactly the kind of unglamorous project a consulting firm can bill for. A venture with a model vendor’s engineers on staff can read a model’s behavior in ways a generalist adviser cannot.
The independence question is the one that will decide how the venture is received. A firm that is partly owned by a model developer has an obvious reason to recommend that developer’s model, and corporate buyers have grown sensitive to that conflict after three years of vendor-led transformations that ended with a prototype and an invoice.
Hg’s partners have an answer ready: the new company will be independent, will work with other models where a client’s data or budget demands it, and will be paid for outcomes rather than seats. Whether clients believe that is another matter. Procurement teams at large European companies have started asking suppliers to disclose ownership ties, according to two people who advise on such contracts.
There is also the talent problem. The people who can build these systems are scarce and expensive, and they are the same people the accounting firms, the cloud providers and the AI labs are all trying to hire. A new firm with two offices will not have scale on day one. It will have something else: direct access to Anthropic’s engineers and to the product roadmap, which is a recruiting argument in a market where the technology changes every quarter.
For Hg, the calculation is closer to portfolio math. A buyout firm that owns dozens of software companies needs its portfolio to adopt AI without breaking, and one capability shared across those companies, with a preferred vendor attached, is cheaper than thirty separate engagements.
For Anthropic, the venture is a hedge. Enterprise software has rarely been won by the company with the best engine; it has been won by the company with the best distribution. The lab has spent two years building a direct sales force and partnerships. Signing a private-equity firm with hundreds of portfolio companies is a distribution move dressed as a services deal.
Neither company disclosed the size of the commitment, and neither would say how the venture will be governed or how profits will be split. Both companies declined to comment on terms.
The first test will be visible by spring. If the London and Berlin offices fill with engineers who can point to production systems at named clients, the idea of a vendor-backed consultancy will look early rather than late. If the offices fill with salespeople, European buyers will treat the venture as a channel and price it accordingly.
