Start a search
We recruit for what's next.
We will contact you within 24 hours. Alternatively you can book a convenient time now.
Book a time
We recruit for what's next. What's next arrives with no title, no talent pool and nobody to benchmark it against. We find the people who can already do it.
What we do
Most firms in this market stop at research and engineering, because that is the part they can read. The finance lead, the general counsel and the first commercial hire are just as hard, and considerably less well served.
Research and engineering
Models, infrastructure, evaluation, applied research, security
Product and design
Product management, design, forward deployed engineering
Go to market
First commercial hire through VP, marketing, partnerships, customer
Corporate
Finance, legal, people, operations, data protection
Leadership
Founding team, functional heads, executive and board search
From the first hire in a discipline to replacing the person who built it.
Why this market is different
Whole disciplines have appeared to build, audit and pay for AI, and job architecture has not caught up. There is no salary band to work from, no comparable hire to point at, and a keyword search returns nothing, because the people already doing the work do not use the title yet. These are simply the newest examples, not the extent of what we recruit.
Build it
Prove it
Run it
Designs multi-agent ecosystems where autonomous systems carry complex, multi-step business logic without a person in the loop, and makes sure that logic does not drift or fall into recursive loops.
Why it is hard to fill
Almost nobody has run agents in production for longer than a year. Filter on years of experience and you remove every candidate who can actually do it.
Structures multi-turn context windows, persona frameworks and tool-calling parameters so output stays consistent across complex enterprise workflows.
Why it is hard to fill
The title is contaminated. Advertise it and you get thousands of applicants with no engineering background, which makes inbound the worst signal in the market.
Generates, cleans and tags high fidelity artificial datasets to train advanced models, now that clean real-world training data is depleted or restricted by privacy law.
Why it is hard to fill
Hardly anyone uses the title on their CV. The people doing the work call themselves data scientists or research engineers, so a keyword search returns nothing.
Runs technical evaluations on deployed models to catch hallucination, security exposure and systemic demographic bias, before the system meets a regulator.
Why it is hard to fill
Has to be credible to a regulator and to an engineering team at once. Audit backgrounds cannot read the model. Machine learning backgrounds cannot write to a regulator.
Keeps enterprise deployments inside evolving global frameworks such as the EU AI Act, maintaining model risk documentation and liability protocols.
Why it is hard to fill
The EU AI Act pushed the whole market at the same small pool simultaneously. Most available candidates have policy depth and no deployment experience.
Verifies, grounds and fact-checks what AI scribes and copilots produce in high-stakes fields such as medicine, law and corporate compliance.
Why it is hard to fill
Needs a real professional qualification plus a willingness to work next to a model. That combination almost never answers an advert.
Manages token budgets, model right-sizing and GPU compute cost, so an enterprise AI operation does not burn capital faster than it generates revenue.
Why it is hard to fill
It sits between finance and infrastructure, so one function advertises it and the other understands it. Cloud FinOps people miss the model economics. Machine learning people miss the profit and loss.
Oversees the physical and behavioural interaction between human staff and collaborative robots across manufacturing, logistics and clinical environments.
Why it is hard to fill
It belongs to operations, not technology, so it never reaches a technical recruiter. The people who can do it come from safety, industrial engineering or clinical operations.
How we work
Any agency can buy an index of eight hundred million profiles, and most now have. It tells you who exists. It does not tell you which company is about to need them, or which person has just become reachable. A vacancy is the last event in a long sequence, so we work from the earlier ones.
What arrives
No account, no login, nothing to learn. Names and contact details are withheld until you say someone is worth meeting. One click releases them and we make the introduction.
Your name is never in the outbound approach. Candidates see an anonymised description of the business and the work, and nothing goes out that names you.
AI Governance Lead
For candidates
Start a search
Eight questions, about two minutes. We come back with whether this is a search we can run well and how we would run it. If it is not one for us, we will say so.