You've got 100 CVs. You've also got a business to run.

The job went live on Monday. By Wednesday you had 100 applications. By Friday it was 180. Somewhere in that pile is the person you want to hire — but finding them means reading every single one, and that job doesn't stop while you do it.

This is the reality for most hiring managers at small companies. There's no talent acquisition team to hand this off to. It's you, a spreadsheet, and a very full inbox. Here's how to get through it without losing a week of your life — or missing the best candidate in the batch.

Why manual screening breaks at scale

The numbers are brutal. Employers receive an average of 250 applications per job posting, with entry-level roles regularly hitting 400 or more. Of those, research consistently shows that 75–88% are unqualified for the role they applied for. You're spending most of your time reading CVs that were never going anywhere.

The average recruiter spends 23 hours screening CVs per hire. That's nearly three full working days, and that figure doesn't include the interviews that follow. Meanwhile, 60% of companies reported their time-to-hire increased in 2024 — partly because the volume of applications is growing, and partly because the rise of AI-generated CVs means 64% of recruiters are now seeing more look-alike applications that are harder to differentiate quickly.

At small companies, this time cost is particularly painful. You're not a recruiter — you're a founder, a department head, or an office manager who got handed this job. Three days buried in CVs means three days of actual work not getting done.

The biggest time sinks in manual CV review

Most of the time wasted in manual screening comes from a few predictable places:

  • Reading every CV in full before filtering. Most reviewers start at the top of the document and work down. This is the wrong order. You end up spending equal time on a CV that fails your must-haves in the first 30 seconds and one that genuinely warrants careful reading.
  • No defined criteria before you start. Without a written list of must-haves and nice-to-haves agreed in advance, every CV becomes a fresh debate with yourself. Does five years' experience matter? Does a degree? If you haven't decided upfront, you'll re-decide it 100 times.
  • Inconsistency across the batch. Human inter-rater reliability for CV review is only 60–70% — meaning if you reviewed the same pile twice, in a different order, on a different day, you'd reach different conclusions on a significant chunk. The contrast effect is real: a mediocre CV after several weak ones looks better than it is.
  • Reconsidering maybes. A "maybe" pile that you revisit three times costs more time than the original read. Every CV you don't decide on stays on your desk.
  • Unconscious bias slowing decisions. Research shows that names, university prestige, and employer brand cause reviewers to make snap judgements that aren't related to the job. The result: strong candidates get missed and weak candidates get through, making later stages harder.

A better process: what to check and in what order

Before you open a single CV, do this:

  1. Write your must-haves. Three to five criteria that are non-negotiable. Right to work, specific qualification, minimum years in a relevant role, location. If a CV fails any one of these, it's a no — stop reading.
  2. Write your nice-to-haves. Another three to five things that distinguish a good fit from a great one. These score points; they don't disqualify.
  3. Decide your shortlist size upfront. You want eight people to interview? Commit to that. It stops you expanding the shortlist indefinitely.

Then, when you're in the pile:

  • Scan for must-haves first — check only those specific fields, not the whole document. This eliminates a large proportion of your stack in seconds.
  • Score the survivors on nice-to-haves — assign a point per criterion met. Total the score. Rank. Don't re-read at this stage.
  • Review only the top-ranked CVs in full — read the details, check for red flags (unexplained gaps, contradictory dates, job-hopping without context), and form your shortlist.
  • Do one pass, not three. Decide yes, no, or reserve on each CV. Don't leave maybes unresolved.

This structure alone can cut your review time roughly in half. But it still requires sitting down and doing it manually, which at 100+ applications is a significant block of time.

How AI screening changes the maths

When you apply the structured approach above with AI assistance, the numbers shift dramatically. AI systems can process 100 CVs in 15–20 minutes versus 10–13 hours manually — a 30–40x speed difference. More importantly, AI doesn't get tired, doesn't succumb to contrast bias, and applies the same criteria to CV 100 as it did to CV 1.

In one documented case, a mid-sized firm received 400 applications for a software role. Two recruiters spent 40 combined hours reviewing CVs over five days manually. With AI screening, the same pool was ranked in 3 minutes, and recruiters spent only 2 hours reviewing the top 10%.

What matters for a small company isn't the enterprise ROI story — it's simpler than that: you get a ranked list with each candidate's fit score, key strengths, identified gaps, and suggested interview questions, rather than staring at 100 PDFs. The AI has done the first-pass scoring against your job description. You're making final calls on a shortlist, not wading through the whole pile.

A blind panel study found that AI-shortlisted candidates were rated "genuinely qualified" by expert reviewers 93% of the time, compared to 73% for manually screened candidates. You're not sacrificing quality — you're getting more of it, faster.

Getting through 100 CVs in under an hour: the actual workflow

Here's a concrete step-by-step that works whether you're doing this manually, with AI assistance, or a combination of both.

Before you start (10 minutes)

  • Write your must-have criteria. Be specific — "2+ years in a client-facing sales role" not "sales experience."
  • Write your nice-to-haves, scored 1 point each.
  • Decide: you want to interview how many people? Fix the number.
  • If using a tool like cvtally, paste your job description and any specific criteria. Let it run while you do something else. The AI scores every candidate against your JD, flags gaps, and generates suggested interview questions per candidate.

First pass — eliminate (15 minutes for 100 CVs, or already done if using AI)

  • Check only must-haves. No, no, no, yes — move fast.
  • Target: cut the pile by at least 50–60%. For most roles, 75–88% of applicants won't meet your must-haves.
  • Don't read cover letters at this stage. Don't read summaries. Check the specific fields.

Second pass — score survivors (15 minutes)

  • Score each surviving CV on nice-to-haves. One point per criterion. Record the score.
  • If AI has done this step, review the scores and flag any that feel off. Override as needed.
  • Rank by score. Your shortlist candidates will rise to the top.

Third pass — read the top candidates properly (20 minutes)

  • Read the top-ranked CVs in full — you're now looking at 10–15 people, not 100.
  • Check for red flags: unexplained gaps longer than 3 months, titles that don't match the described responsibilities, qualifications that can't be verified.
  • If AI flagged gaps or inconsistencies, check those specifically.
  • Pick your shortlist. Use the AI-suggested interview questions as your starting point for the calls.

Total time: under an hour.

Manual-only version of this process takes 2–3 hours for 100 CVs when you're disciplined. With AI handling the scoring and flagging, you're reviewing a ranked shortlist in 30–45 minutes total.

The one thing that wastes the most time

It's not reading CVs. It's reading them without a system. The hiring managers who spend a full week on 100 CVs are the ones who start with CV number one and work through the pile treating every application as equally deserving of their full attention. They're not wrong to be thorough — but thoroughness applied at the wrong stage costs them hours they don't have.

Set your criteria. Filter fast. Go deep on the people who survive the filter. That's the whole system.