AI recruitment UK is currently a noisy market pretending to be a settled one. Job titles appear and vanish. CVs list tools the way they once listed Microsoft Office. Boards ask for “someone who can do AI” as if that were a trade like electrician. Meanwhile the useful hire is still a person who can sit inside a real business — operations, performance, a transformation programme — and make a change that survives contact with customers and constraints.
Viktorija Sherriff’s practice is not a volume shop for that noise. VS Recruitment is a boutique UK headhunter. She starts a named search for a £450 engagement fee and works her own network. She does not scrape job boards. She does not spray CVs because a profile mentioned a model. If you want theatre, there are agencies for that. If you want someone who can ship, that is a search.
The job title is not the brief
Digital transformation roles have always suffered from vague naming. AI has made it worse. “AI lead”, “automation manager”, “head of innovation” can mean a person who has implemented a workflow that saved a team genuine hours, or a person who has run a lunch-and-learn and a pilot that never left a slide. A specialist search starts by refusing to hire the title. What does the week look like. Who do they sit with. What would have to be true in six months for you to call this a success.
That conversation is closer to built sector recruitment than it first appears. In construction you do not hire a “buildings person”. You hire someone who can run a site, a contract, a facility. In AI hiring you should not hire a “tools person”. You hire someone who can change a process in a UK company that already has customers, a board, and a regulator somewhere in the background.
What she looks for, without fake credentials
There is no CSCS card for generative tools, and VS Recruitment will not invent one. Viktorija vets the way she vets digital marketing recruitment: by talking about the work. Has this person shipped something that other people now use. Can they talk to operations without condescension. Do they know the difference between a proof of concept and a process that has an owner. Are they honest about what a model cannot do. Prompts are not a department.
She also looks at where the role actually sits. An AI hire inside a performance team needs to be numerate and commercially literate — the same instincts that matter in PPC recruitment. An AI hire inside a facilities or property business needs to respect how buildings are actually run. A decade in the built sector and in digital/PPC is relevant here. The transformation that sticks is usually the one that understands the existing trade.
Why volume fails this market even faster
A CV spray is always a bit of a lie. In AI hiring it is a louder lie. Keyword matching now returns everyone who attended a webinar. You then interview fluency. The person who can talk about agents, copilots and “the stack” will outscore the person who quietly automated a reconciliation and documented it so a colleague could run it. If your business is not a lab, you want the second person. They will not apply to a generic advert. They will take a call from a headhunter who can describe the work without sounding like a conference.
Confidentiality matters more than people admit. Plenty of UK firms are exploring AI roles without wanting the market to think they are either behind or naive. A network-only approach is built for that. The role is not posted. The conversations are named. The shortlist is short on purpose.
Duties that did not disappear because the role is new
New job titles do not cancel old law. Employers still check right to work, using the process on GOV.UK. How you describe a role and how you select still sits under the Equality Act 2010; the statute is on legislation.gov.uk. If the work touches health and safety systems on a site or in a building, the HSE construction guidance is still the adult reading list — an AI workflow does not replace a dutyholder. VS Recruitment is a headhunter, not your lawyer or your safety adviser. A careful search just does not skip the boring parts.
Write the brief around a process
The most reliable way to hire well in AI is to describe a process that needs to change, rather than a technology you would like to use. “We spend too long turning site survey notes into quotes”, “our customer service team triages every email by hand”, “month-end reporting takes days of copying between systems” — each of those is a brief. “We need someone to do AI” is not.
A workable brief covers:
- The process, who runs it today and roughly how much time it takes.
- The data involved, where it lives and who owns it.
- Who the hire will work with in operations, IT, finance and data protection.
- Who signs off a change, and who maintains it once the hire moves on to the next problem.
- What success looks like in six months, in terms the operations team would recognise.
A brief like that attracts people who want to deliver and quietly puts off people who mainly want to experiment.
Interview questions that separate delivery from fluency
- Tell me about something you shipped that other people now use. Who uses it, and what changed for them?
- How did you decide whether the output was good enough? What did you measure?
- What went wrong, and what did you do about it?
- How did you handle personal or commercially sensitive data?
- How did you document the work so somebody else could run it?
- What would you not automate in our business, and why?
People who have delivered describe users, data, testing and the decision to stop or continue. They talk about cost and risk in plain language. People who mainly know the vocabulary tend to describe tools and pilots with no named users and no measurement. The last question is often the most revealing: someone who has done the work knows where automation creates more problems than it solves.
Data protection belongs in the brief
Many AI projects in UK businesses touch personal data: customer emails, employee records, call notes, CVs. That brings UK GDPR and the Data Protection Act 2018 into scope, and the Information Commissioner’s Office publishes guidance on AI and data protection. You do not need to hire a lawyer, but you do need someone who knows when to involve one. Ask candidates whether they have worked with a privacy lead or contributed to a data protection impact assessment, and what changed in the design as a result.
Red flags in AI hiring
- A portfolio of demos and proofs of concept with no named users and no live process behind them.
- Claims of large efficiency gains with no baseline, no time period and no explanation of how they were measured.
- Impatience with data protection, security or the people who currently run the process.
- A preferred tool for every problem before they have heard what the problem is.
The right hire is usually calmer than the market. They can explain what a system cannot do as clearly as what it can, and they are comfortable saying that a spreadsheet and a better process would solve half the brief. The AI recruitment UK page sets out what gets checked in a search.
Where the role should sit
An AI hire who reports into IT will be pulled towards infrastructure and security. One who reports into operations will be pulled towards day-to-day fixes. One who reports to the managing director may find that every department wants something at once. None of those is wrong, but the reporting line shapes the work far more than the job title does. Decide it before the search, and say it in the brief.
The setting matters too. An AI hire inside a performance marketing team needs to be numerate and commercially literate, with the same instincts that matter in PPC. An AI hire inside a facilities or property business needs to respect how buildings are actually run, including the statutory work that cannot be automated away. Candidates who have worked near the trade they are joining tend to settle faster, because they understand why a process looks the way it does before they try to change it.
What a good AI shortlist looks like
Two or three people who can each describe a change they delivered, the people who used it and what happened next. Each should be able to talk to your operations lead without translation. They should have a clear reason for moving and a realistic view of what can be done in the first six months. If the shortlist is full of impressive titles but thin on users and outcomes, it is the wrong shortlist, however good it looks on paper.
How to start
Write the work, not the fashion. Sector: AI, or AI sitting on top of built or digital. Role: what you need done. Constraints: in-house or contract, seniority, whether this is confidential. If it is a fit, the £450 fee opens the search and Viktorija works the network. If it is not — if you want a hundred “AI interested” CVs — she will decline. That is the service.
AI recruitment UK will calm down eventually. The businesses that hire well in the meantime will be the ones that treat it as a specialist search, the same way they should already treat construction recruitment and PPC recruitment. If that is the way you want to hire, start a search. A short note is enough. Precision, not volume.
Frequently asked questions
What does an AI hire in a UK SME actually do?
Usually they take one or two business processes, work out where automation or AI tools can help, build or configure the change with the people who run the process, measure whether it works and document it so it can be maintained.
Do I need a data scientist or an AI generalist?
Most operating businesses need someone who can deliver change with existing tools and data rather than build models from scratch. A data scientist makes sense when you have substantial data and a problem that genuinely needs modelling. The brief should make that difference clear.
How do I check that AI experience is real?
Ask about users, measurement and failure. Real delivery comes with named users, a before-and-after and at least one thing that did not work. Fluency without those is a warning sign.
Does hiring for AI change our legal duties?
No. Right to work checks, the Equality Act 2010 and data protection law still apply. If the role touches personal data, UK GDPR and ICO guidance apply to what the person builds as well as to how you recruit them.
What does the £450 fee cover?
It opens a named, network-only search. It is not a listing charge, and if the brief is not a fit — for example, a request for a hundred “AI interested” CVs — Viktorija will decline before anyone pays.