The news
Anthropic and the private-equity firm Hg have announced a European consultancy, with presences planned in London and Berlin and operations live from the end of 2026. When a frontier lab decides delivery is worth building itself, that is the loudest confirmation possible of what practitioners have said for two years: the bottleneck in enterprise AI is not model quality. It is production delivery.
Buyers should read this as good news. The category now has a lab-grade participant, which means the market for actually shipping AI in production is large enough to fight over.
What a lab-owned consultancy offers
Three things nobody else can match, and they are real. The deepest access to the vendor’s own stack: models, agent infrastructure and deployment patterns exactly as the lab intends them to be used. Roadmap proximity: new platform capability tends to show up in the lab’s own delivery work first. And no translation loss between the people who build the platform and the people who deploy it for you.
What it is structurally hard for a lab-owned arm to promise
None of this is a criticism. It is geometry. A consultancy owned by a model vendor cannot make certain promises its ownership contradicts.
Model neutrality. A lab-owned arm cannot, as a founding promise, tell you when a competitor’s model is the better choice for one of your workflows. An independent partner can. Nugawi, an applied-AI consultancy in London, is an approved partner of both Anthropic and OpenAI: we deploy the right model for each workflow, not one model for everything, and we say so in writing before you sign anything.
Freedom to switch. Ask what happens to your system the day you change primary model, or want to run two vendors side by side. The honest answer from any lab-owned arm is that you are buying into one vendor’s trajectory. That can be exactly right if you have chosen that vendor deliberately and want its full stack. Buy with your eyes open.
Incentive alignment on outcome. Labs are product companies. A consultancy’s only product is your outcome. When the engagement gets hard, in month five, you want the party whose entire business depends on the system working.
Five questions to ask any AI consultancy
Lab-owned or independent, these five questions separate a production partner from a body shop:
- Who owns the code, the data and the prompts, and from which commit? Yours from the first commit, in writing, is the only acceptable answer.
- What happens when we change models? A production system designed for model change is a few weeks of work. One welded to a single vendor is a rebuild.
- Where does our data live, and whose terms apply? UK or EU residency and a clear controller and processor position, not a shrug.
- What is fixed, and what is open-ended? Fixed fees and fixed scope exist: a 90-day production build with a defined end state, or a Discovery Sprint that ends in a decision. If everything is a retainer, the incentive is to stay.
- Who is accountable for the outcome, and how is it measured? Published results, named deployments, numbers your CFO recognises.
Where Nugawi stands
We are an independent, approved partner of both Anthropic and OpenAI, based in London, and our practice is built on the promises a lab-owned arm structurally cannot make: model neutrality, fixed fees and fixed scope, and IP from the first commit. Our Claude consulting practice sets out how we deploy, govern and hand over. If you are weighing the options, a Discovery Sprint is the fastest way to make the comparison concrete: your workflows, your data, a written recommendation, and a firm price.
The consultancy you choose should be the one whose incentives survive your next model change.