Your best candidate is in the CVs nobody read

A published role attracts somewhere between 80 and 400 applications. Some CVs are read carefully. Some are scanned in seconds. Some are filtered out on a single keyword. And a large share are never opened at all.

So ask the real question: where is your best candidate? There's a good chance they're sitting in the pile nobody had time to read.

The real problem is volume

Everyone talks about how hard it is to recruit. Almost nobody talks about the stage that actually decides the outcome: the first screening. It's the hidden — and hardest — part of the process.

When a recruiter has 200 CVs to work through and three other open roles, they don't read — they decide fast. And a fast decision always favors the same signals:

• recognizable job titles;
• familiar companies;
• the exact keywords from the ad;
• clean, well-formatted CVs.

The problem is that none of these signals measure competence. They measure readability. And readability quietly screens out people who deserved an interview:

• candidates with transferable skills a job title doesn't capture;
• candidates from smaller companies, who "look less impressive";
• career changers, rejected too early;
• people who described their experience using different wording than the ad.

The Harvard Business School and Accenture "Hidden Workers" report documented this bluntly: rigid filters and applicant tracking systems exclude perfectly qualified people, because they focus on what a candidate lacks rather than what they can do.

Speed damages quality

A recruiter is measured on time metrics: time-to-shortlist, time-to-hire. Those are useful numbers — but they pull the process in a specific direction.

Under time pressure, you keep the candidates who are easy to understand quickly. Not the best ones. The strongest fit isn't always the fastest to decode, and that's exactly where the mistake slips in.

It needs saying plainly: a CV is not a person. A job title is not a skill set. A keyword is not proof of competence. When screening runs on those shortcuts, it isn't talent doing the deciding — it's formatting.

Where AI-assisted screening genuinely helps

The point is not to replace the recruiter. The point is to give them back something they've lost: the time to look. An AI screening agent can:

• read every CV, with the same attention, without fatigue;
• compare each profile against the actual role requirements;
• summarize the relevant experience;
• flag missing information;
• highlight transferable skills;
• group applications by level of fit;
• suggest screening questions;
• prepare a structured shortlist for human review.

The word that matters here is assist. The AI is the first reader — not the hiring manager. It reads everything so that a human can then decide on the full picture, rather than on the first twenty CVs in the pile.

The goal isn't more automation — it's better attention

This is the distinction that separates bad automation from good.

Bad automation removes people too early: it reproduces the rigid filter, just faster. Good automation does the opposite — it helps the human look more carefully.

Weak AI says: "reject this candidate." Useful AI says: "does not match the exact job title, but has three years of relevant project experience — worth reviewing."

The first replaces a judgment. The second prepares one. Only one of them actually moves recruitment forward.

What stays human

The line has to be drawn without ambiguity. The AI is not there to decide who you hire. These stay firmly in the recruiter's hands:

• the final shortlist;
• interview decisions;
• communication with candidates;
• assessing cultural fit;
• the salary question;
• the fairness review of the process.

And one non-negotiable requirement: the AI must explain why it highlights a candidate. A recommendation without a reason helps no one — it just moves the black box somewhere else.

That's exactly Nova's role. She doesn't choose your hires. She simply makes sure the strong candidate isn't lost because nobody had time to read the CV.