How 500+ recruiters screened 20,000+ resumes in under 95 minutes of AI time

The problem: too many CVs, not enough hours

Recruitment agencies live and die by throughput. A single popular role can attract hundreds of applicants, and an agency running dozens of roles at once quickly drowns in resumes. The traditional options are all bad: skim the first fifty and hope, lean on brittle keyword filters that reject strong people for using the "wrong" words, or burn recruiter hours reading CVs one by one.

When we opened up free AI screening, one of our partner agencies rolled it out across their desk. Over the measured period, 500+ recruiters ran screenings covering more than 20,000 resumes.

The result: 20,000+ resumes, under 95 minutes of AI time

The headline number is the one that surprises people: the combined AI processing time to score those 20,000+ resumes against each role's criteria was under 95 minutes. That is the total compute the model spent reading and evaluating CVs. It is not wall-clock time spread across a day, and not counting the human review that follows. Averaged out, that's roughly a quarter of a second of AI reasoning per CV.

What used to be days of collective skimming became something recruiters could kick off between calls and come back to with a ranked shortlist waiting. The recruiters still made the decisions. The AI just did the reading, fast, consistently, and against the exact criteria each recruiter set.

What that saves in hours and dollars

Put the throughput in human terms. A recruiter doing a considered first-pass on a CV takes around four minutes. Screening 20,000 resumes by hand is therefore roughly 1,300+ recruiter-hours (about 167 full working days, or the better part of a year for one person doing nothing else).

The AI did the reading in under 95 minutes of processing time. Even after accounting for the human review that still follows a screen, the manual reading effort (the part that used to consume those 1,300+ hours) is essentially eliminated.

This agency runs desks across the US, Australia/New Zealand, Singapore and the UK. Using a conservative blended loaded cost of about US$45 per recruiter-hour across those markets (base salary plus benefits and overhead), the reading effort on that single batch of 20,000 CVs represents roughly:

• ~1,300+ recruiter-hours saved (≈ 167 working days)
• ~US$60,000 in recruiter time on one batch of resumes

Scale that across a year of reqs and the numbers compound quickly, which is exactly why the agency moved screening onto AI and kept its recruiters focused on the conversations only humans can have.

Savings are estimates for illustration. Assumptions: ~4 minutes of manual screening per CV; a conservative blended loaded recruiter cost of ~US$45/hour across the agency's US, ANZ, Singapore and UK offices. Your mileage will vary with role complexity and review depth.

Why it worked

Three things made the difference for this agency:

1. Criteria they controlled. Every screening started from a job description that generated weighted criteria the recruiter could edit. No opaque relevance score. The recruiter decided what mattered and how much.

2. Reasoning, not just ranking. Each candidate came back with a per-criterion breakdown explaining the score. Recruiters could trust the shortlist because they could see why someone ranked where they did, and override it when their judgement differed.

3. Contextual reading over keywords. The model understood that "led a team of twelve" satisfies a leadership requirement even without the word "leadership." Strong candidates stopped falling through keyword cracks.

Beyond CVs: Voice AI screening is scaling too

Resume screening was the entry point, but the same agency is now rolling out Voice AI screening, and adoption has been fast. Across the rollout, Voice AI has already assisted 130+ requisitions, conducting 18,713 combined voice minutes of candidate screening conversations.

That's over 310 hours of first-round screening calls handled by AI (consistent questions, every candidate given the same fair shot, and structured summaries back to the recruiter), freeing the team to spend their time on the shortlisted people who matter most. The pattern mirrors what happened with CV screening: once recruiters saw the time it gave back without sacrificing quality or control, adoption climbed on its own.

What it felt like on one recruiter's desk

The aggregate numbers are the headline, but numbers this big flatten out the human story underneath them. So let me zoom in on one recruiter, a composite we'll call Priya, to show what a Tuesday inside that 20,000-CV wave actually looked like. She is illustrative of the recruiters at the agency, not a single named person, but her day is stitched together from what several of them told us.

Priya had a warehouse-operations role that had pulled 620 applications in nine days. Under the old routine, she would have blocked out most of two days to read them, and she would have quietly resented the role by the end. Instead she pasted the job description into free screening, edited the weighted criteria until they matched what the hiring manager actually cared about, uploaded the folder, and went to lunch. By the time she was back, she had a ranked list with per-criterion reasoning for every one of the 620.

The part that stuck with her was not the speed. It was candidate number 411, a strong applicant whose resume never used the word "supervisor" but described "coordinating a shift crew of fifteen." A keyword filter would have dropped him. Priya, reading from the top on a normal day, might never have reached him. The screen surfaced him at number six, with a plain-language note explaining why. She placed him three weeks later.

The moment adoption tips

Agencies do not adopt tools by memo. They adopt them when one recruiter shows another something that saved their afternoon. That is exactly how this rolled out. A handful of recruiters tried free screening on their worst high-volume reqs, walked away with hours back, and mentioned it in the team channel. Within the measured period, 500+ recruiters were running screenings of their own.

Nobody mandated it. The tool spread because the people using it stopped dreading their inboxes, and that is contagious in a way a training session never is. One team lead described the shift to us, and we share it as an illustrative paraphrase rather than a quoted client: "I did not have to sell it. I just had to let the recruiters who tried it talk at standup."

What the recruiters did with the time

It is tempting to frame this as a cost story and stop at the ~US$60,000 figure. But ask the recruiters where the reclaimed hours went and the answer is more interesting than a spreadsheet. They did not go home earlier, mostly. They spent the time on the phone.

Screening was never the job recruiters loved. The job they loved was the conversation where a candidate's real story comes out, the one that never fits on a resume. When the reading collapsed from 1,300+ hours to under 95 minutes of AI time, that conversation is where the hours moved. More candidates got a proper call. More hiring managers got a considered shortlist instead of a rushed one. The quality of the human parts of recruiting went up precisely because the mechanical part stopped eating the day.

There is an honest caveat here, and we would rather say it than hide it. The screen does not make the hiring decision, and it should not. A ranked list with reasoning is a strong starting point, not a verdict. The agency's recruiters still reviewed shortlists, still overrode the ranking when their judgement differed, and still owned every placement. The AI did the reading. The people did the deciding. That division is the whole point.

The number that made the finance team pay attention

Recruiters adopted screening because it gave them their afternoons back. The finance side of the agency cared for a different reason, and it is worth spelling out because it is the argument that gets a rollout funded past the pilot.

Run the arithmetic on that single batch. Twenty thousand resumes at roughly four minutes of considered manual reading each is more than 1,300 recruiter-hours. At a conservative blended loaded cost of about US$45 an hour across the agency's US, ANZ, Singapore and UK desks, that reading effort is worth around US$60,000, on one batch of CVs. The agency does not run one batch a year. It runs high-volume reqs continuously, so the reading cost that screening removes recurs every month, on every popular role.

The finance lead's reaction, which we paraphrase for illustration rather than quote, was not excitement about the technology. It was relief that a recurring six-figure line of pure reading effort had a lever attached to it for the first time. Recruiters felt the change as a better week. Finance felt it as a cost that stopped scaling linearly with volume.

Both are true at once, and that is why the tool did not stall after the pilot. The recruiters wanted it because their days improved. The people who sign off on budgets wanted it because the maths held up. A tool needs both constituencies to survive contact with a real agency, and this one had them.

Try the same tool, free

The screening those recruiters used is the same one you can try right now, no signup required to see your results. Paste a job description, upload a folder of CVs, and get a ranked shortlist with reasoning in minutes.

Screen your resumes free →

Figures reflect aggregate usage observed at a GoVerse partner recruitment agency. "Under 95 minutes" refers to total AI processing time across the screenings, not human review time.