UK tech employers should hire for judgment, not AI tool names

UK employers still plan to grow their technology teams. On 29 September the recruiter Robert Half published research saying that nearly half (47%) of UK employers plan to increase permanent technology headcount before the end of the year, and that a further 25% are expanding contract and project-based hiring. The skills those employers want most, according to the release, are cybersecurity (54%), agentic AI use (50%), generative AI solutions (48%), cloud (44%) and business intelligence (44%).

The research is Robert Half’s own, and Robert Half sells permanent placement and contract talent services. The survey was developed by the company and run by an independent research firm in June 2026, three months before publication. It covered 500 hiring managers and 1,000 professionals in the UK across six fields: finance and accounting, technology and IT, marketing and creative, legal, administrative and customer support and HR. The release does not say how many of those respondents work in technology.

The more telling numbers sit lower down the release. It says 45% of UK technology professionals say they are now expected to develop new AI-related skills, and nearly two in five (38%) spend more time overseeing and validating AI-generated outputs. At the same time, 53% say AI has reduced the time they spend on routine tasks and 37% say their roles have become more strategic.

Read together, those figures describe a job changing shape rather than disappearing. For many respondents, routine work takes less time and checking takes more. For UK tech employers, that makes judgment about whether machine output is right the skill to hire for, and job specs, assessments and pay should be built around it rather than around tool names.

‘Agentic AI use’ on a job spec

The release puts agentic AI use second on employers’ skills list, at 50%. As a line in a job advert, that phrase invites candidates to list the tools they have tried. The release itself points somewhere more useful: it says the findings point to a growing demand for professionals who can work alongside AI systems and apply the human judgment needed for accuracy, quality, security and compliance.

Matt Weston, senior managing director, UK & Ireland at Robert Half, said businesses are “not simply looking for more people”. He added: “They’re looking for professionals with the right blend of technical expertise, commercial understanding and the ability to work effectively alongside AI.”

For a hiring manager, the practical translation is three questions. What does this person have to check? How often? And what breaks when they miss something? A service desk analyst reviewing AI-drafted fixes, a developer approving generated code and a security analyst triaging machine-ranked alerts are doing different jobs, but each is signing off work they did not write. The job spec should say so.

The same problem in the hiring process

Employers are meeting the verification problem twice: in the work, and in judging who can do the work. An earlier Robert Half release, published on 14 September and drawn from the same June survey, found that two-thirds (67%) of hiring managers say the need to screen for AI-generated materials has increased the time it takes to hire. Almost half (48%) of UK employers say AI-generated CVs or application materials are creating notable challenges, and 23% say they need additional in-person interviews or screening to validate candidates’ skills and experience. Those figures cover hiring managers across all six fields, not technology alone.

Weston said that when large volumes of applications appear highly polished or follow near-identical formats, “it becomes harder to evaluate candidates, validate skills and experience and confidently identify the strongest fit for the role”.

A polished CV cannot show whether a candidate spots a plausible wrong answer. A short work sample can. Give shortlisted candidates a piece of AI-generated code, a firewall rule set or a data summary with errors planted in it, and ask them to find the errors and explain the fix. That tests the task nearly two in five tech professionals already say takes up more of their time, and it puts the candidate, not the CV, in front of the evidence.

Pay for the sign-off, not the volume

If routine work is shrinking, output measures such as tickets closed, pull requests merged or reports produced start to reward the part of the job AI is taking over. The risk is that the people doing the careful review look slower on those measures than colleagues who accept whatever the tool produces.

People leaders can correct for that in four ways:

• Write roles around the decisions the person signs off, and the checks that go with them.
• Assess for error-spotting and explanation, not familiarity with named tools.
• Set progression on the quality of review and the incidents avoided, not output volume alone.
• Budget working time for the new skills 45% of tech professionals say they are expected to develop. The release does not say who pays for that learning or in whose time.

The 25% of employers expanding contract and project-based hiring face a version of the same question. A contractor brought in to deliver a project at speed is still signing off AI-assisted work, and the statement of work should name who checks it and against what standard.

Weston’s own summary was that “AI is automating some tasks, but it isn’t replacing the need for skilled technology talent. In many cases, it’s making those skills even more valuable.” The skill it is making more valuable is knowing when the machine is wrong. Employers that hire, assess and pay for that will get more from their new headcount than those advertising for the latest tool name.