What the CV pile actually costs a recruitment desk

Ask a recruiter how long they spend reading CVs and you will get a shrug and a guess. Ask them how many roles they lost last quarter because a competitor got a shortlist out first, and you will get a much more precise answer. The two questions have the same answer, and almost nobody connects them.

So let us do the arithmetic properly.

The number nobody writes down

A working desk handling four to six live roles takes somewhere between eighty and two hundred applications a week, depending on the sector and how much of it comes through job boards. Even skimming — and it is skimming, ten to fifteen seconds a CV to decide whether it is worth two minutes — that is roughly four hours a week gone before anybody has spoken to a candidate.

Four hours a week is about seventeen hours a month. On a fully-loaded recruiter cost of £45,000 to £55,000, that is somewhere around £500 a month of salary spent on the part of the job that produces no placements on its own.

That is per recruiter. A desk of six is £3,000 a month, or £36,000 a year, to read documents.

But the salary is not the expensive part

The cost that actually shows up in the P&L is not the hours. It is what the hours delay.

Contingency recruitment is a race with an obvious finish line: the first credible shortlist in front of the client sets the shape of the whole process. The agency that submits on day one is the agency whose candidates the client interviews first, whose candidates become the mental benchmark, and whose fee gets paid. Everybody else is submitting against an anchor somebody else set.

The gap between "CVs arrive" and "shortlist sent" is usually not skill. It is queueing. The CVs sat in an inbox on Tuesday because Tuesday was full of interviews, and got read on Thursday.

Take a desk placing thirty roles a year at an average fee of £8,000. If shortlisting a day earlier converts even three more of those — and on contingency roles it converts more than that — the delay is costing £24,000 a year on a desk of one. That dwarfs the salary arithmetic.

Where the time actually goes

It is worth being specific, because "reading CVs" is four different tasks and only one of them is judgement:

• Opening and parsing. Finding the dates, the titles, the tech, the registrations. Mechanical, and the same work every time.
• Checking against the brief. Does this person have the five things the client said were non-negotiable? Also mechanical, and the part most likely to drift by the fiftieth CV.
• Ranking. Which of the twelve plausible people are the best six. Half-mechanical: comparable evidence, compared consistently.
• Judging. Would this person survive that hiring manager? Is the career story credible? Is the gap in 2023 a problem or a sabbatical? That is the job, and it is not automatable.

The first two are most of the clock and none of the value. The last one is almost none of the clock and all of the value.

The drift nobody notices

There is a second cost, harder to see and worse in its consequences.

The criteria you apply to CV number four are not the criteria you apply to CV number seventy-four. Not because you got worse at your job, but because the brief lives in your head and heads move. By the middle of a big batch you have quietly relaxed the "must have" you started with, because you have seen forty people who do not have it. Then a strong one arrives, you tighten up again, and the person you rejected on Tuesday would have passed on Thursday.

Nobody catches this, because catching it would mean re-reading the first forty against the standard you ended up with.

That is not just inefficiency. In a regulated sector, or in front of a client who asks why a particular candidate was screened out, an inconsistent standard is a problem you cannot defend — because there is no record of what the standard was.

What actually fixes it

Not "AI reads CVs". That framing gets the boundary wrong and is why so many agencies tried a tool once and quietly stopped.

What fixes it is fixing the criteria before the reading starts. Turn the job description into an explicit, weighted list of requirements — written once, agreed once — and then judge every CV against that same list, with the line from the CV that justifies each verdict recorded next to it.

That does three things. The mechanical hours collapse, because parsing and checking are exactly what a machine is good at. The drift disappears, because the standard is written down and cannot move between candidate four and candidate seventy-four. And the judgement stays with the recruiter, who now opens a ranked list with the evidence attached instead of a folder of eighty PDFs.

The recruiter still decides. They just start from a different place.

The honest caveat

A vague job description produces a vague list of requirements, and a vague list produces a shortlist you should not trust. "Strong communicator, team player, fast-paced environment" cannot be checked against a CV by a person or a machine. Rubbish in, confidently-scored rubbish out.

The agencies that get the most out of this are the ones whose briefs were already specific — and the ones who discover, on the first role, that theirs were not, and fix it. That second group tends to get more out of it than the first.

---

If you want to see the arithmetic on your own desk rather than ours, the fastest version is to time yourself on the next batch. Fifteen seconds a CV, multiplied by the number that arrived last week. Most people are surprised, and almost nobody is surprised downwards.