On paper, this should be an easy time to hire. The latest KPMG and REC Report on Jobs found UK candidate availability kept rising sharply in September, while vacancies fell again. KPMG's Group Chief Executive said some sectors could be seen as an employer's market. More people are chasing fewer roles, and employers have more AI than ever to help sort through them.
Businesses have invested in AI search, automated shortlisting and outreach at scale. 72% of American recruiters said AI had made finding candidates faster; in the UK, only 58% did. More than one in five British recruiters (22%) said AI had slowed them down, against 13% in the US.
Of the UK recruiters AI did make faster, 45% are no more confident they've found the right person, against a third in the US.
“Recruiters find people faster, then spend the time they saved working out whether those people can do the job.”
— Wouter Durville, CEO and co-founder, TestGorilla
The time didn't disappear. It moved
That's what I mean by the productivity paradox. Recruiters find people faster, then spend the time they saved working out whether those people can do the job.
Three in four UK recruiters spend an hour or more of their own time on a single candidate before they trust that person can do the job. 30% spend two hours or more. At the same time, 69% review 26 or more profiles to find one person worth contacting.
The work AI was meant to remove has shifted to later in the process, where mistakes cost more.
The reason is the material recruiters are working with. 75% of UK recruiters believe half or more of the CVs and profiles they see are written by AI. Checking whether candidates really have the skills on their CVs is the top sourcing challenge in the UK, named by 49%. When a sourced candidate’s profile claimed AI skills, 45% said those skills often or almost always turned out weaker at interview.
I co-founded TestGorilla believing the CV should be one signal but never the only one. That belief got a real test recently: we had more than 7,000 applications for a single marketing role, and a skills assessment showed us who could actually do the work before anyone spent time on interviews.
Across the employers hiring on our platform, the people who got the job scored around 16 points above the average applicant on skills tests, but only about 5 points above on their CV score. A strong CV helps, but employers’ decisions follow skills far more closely.
More candidates should be good news. Without a reliable way to tell them apart, it means more claims to check.
The cost of moving fast on the wrong signal
An overstated CV rarely fails at sourcing. It passes screening, gets an interview and sometimes gets an offer. A survey of UK and US hiring leaders found 59% had made a bad AI hire in the previous 12 months: someone who sounded fluent in the interview and couldn't apply it on the job.
By the usual measures, those hires looked like successes: the roles were filled. That's why the paradox is so hard to spot from inside a recruitment dashboard.
What to measure instead
One of the biggest challenges for employers is how to judge whether a change to their process has worked. Most recruitment reporting can't answer that because it counts activity: applications received, profiles reviewed, time-to-hire, cost-per-hire. AI improves all of them, which is why so many businesses believe it's working. None of them tell you whether you hired the right person.
To judge whether AI is improving outcomes, I'd track four numbers alongside the ones you already report.
Shortlist pass rate. Of the people you shortlist, how many hold up when their skills are checked, whether through a skills test, a work sample or a structured interview? If that number is low, your sourcing is guessing.
Interviews per hire. Hiring managers' time is the most expensive in the process. A better shortlist should mean fewer interviews before an offer.
Mis-hire rate. How many hires don't work out in their first six months? If that number rises while time-to-hire falls, the speed is costing you.
Quality of hire. For the people who stay, how do their managers rate them at six months, and how well do they perform on the role's core skills? Mis-hire rate tells you who didn't work out. This tells you who's actually good.
Three things to change this quarter
First, agree what good looks like before anyone searches. Every role needs written criteria, and the need is greatest in newer roles: a third of UK recruiters say they have no clear picture of what good looks like in an AI role. Write the criteria down, give them to whoever sources, and score every candidate against them.
Second, move evidence earlier. 92% of UK recruiters say they would use an objective check before interviewing a sourced candidate. A short job simulation or skills test tells you more than another read of a polished profile.
Third, make every name on a shortlist come with a reason. Most sourcing platforms confirm that someone exists, is reachable and worked where they say they did. Far fewer explain why a candidate was ranked where they were. Ensure your tools can show which criteria each person matched and on what evidence. If a recruiter can't tell a hiring manager why someone is on the list, the shortlist isn't finished.
Then put the shortlist pass rate and mis-hire rate in the same report as time-to-hire. When they move in different directions, you're looking at the paradox in your own numbers.
Finding people stopped being the hard part of recruitment some time ago. Knowing who can do the job is still hard, and most businesses aren't measuring it yet.
Author bio
Wouter Durville is CEO and co-founder of TestGorilla. TestGorilla is a proof-first AI recruiting platform. Recruiters source from 900 million profiles, and then prove who can do the job with TestGorilla's 350+ skills tests and 100+ AI interviews. TestGorilla's mission is to place 1 billion people in their dream jobs. www.testgorilla.com
