A hundred CVs nobody has time to read
Picture posting a single job opening and finding a hundred CVs in your inbox within a week. Your recruiter — or you, the owner, if there is no recruiter — can't read them all properly. So the first twenty get a careful read, the rest a quick glance, and somewhere in that pile the right person may be sitting unseen.
This is where AI in recruiting genuinely saves time. A model can read every CV to the end, pull out years of experience, specific skills, languages and commute distance, and rank candidates by how well they fit your role. There's no magic to it — it's just systematic reading that never gets tired on the eightieth document. The output is not a decision. It's a sorted pile, with the profiles worth looking at first sitting on top.
Job ad, pre-screening and calendar — three concrete jobs
In practice, AI in recruiting earns its keep on three things. The first is drafting the job ad: you describe who you're after and the model writes a readable text you can edit, instead of a blank page. The second is pre-screening — automatic ranking plus a one or two sentence summary of each candidate, so instead of a hundred PDFs you read a clear list.
The third, often underrated, is scheduling. Half of recruiting time disappears into emails like "would Tuesday at ten work for you?" An assistant connected to your calendar can offer candidates open slots and confirm a time without you lifting a finger. None of these jobs requires the AI to judge anyone as a human being — it simply handles the admin that would otherwise eat hours.
Where it breaks down: bias and fairness
Now the uncomfortable part. A model learns from data, and data about people carries old patterns baked in. If the people who historically succeeded in a role mostly shared one profile, the model easily memorises that as "a good candidate looks like this" — and quietly starts penalising women, older applicants, graduates of a different school, or anyone with a gap on their CV due to parental leave.
This is not a theoretical risk. There are documented cases where a company's recruiting tool penalised CVs containing the word "women's" (as in a women's sports club) entirely on its own. The danger is that it's invisible: the tool never tells you "I'm discriminating," it just hands you a ranking, and you can't see what it ranked by. That's why we never use AI in recruiting as a sieve that throws candidates out on its own.
A human always has the final word
The rule we give every client is simple: AI prepares, the human decides. The model may rank, summarise and suggest a reading order. It may not automatically reject anyone, nor decide by itself who gets invited to an interview.
The distinction matters. When AI merely sorts the pile and you review at least the top thirty profiles with your own eyes, you get both speed and control. When AI decides and you just rubber-stamp its picks, you take on responsibility for decisions you never saw — including any hidden bias. In Europe this isn't just an ethics question either: automated decision-making about people is regulated, and a candidate has the right to a human review. We build the system so you can actually honour that right.
What your company gets out of it in practice
A realistic payoff for a smaller company looks like this: the time to do a first sort of CVs drops from hours to minutes, job ads take a few minutes instead of half a day, and scheduling stops being an annoying email ping-pong. The hiring process doesn't get fully automated — and it shouldn't. The interview, judging the fit with your team, and the final yes stay with you.
To sum up: AI in recruiting is an excellent assistant for routine and a weak — even dangerous — judge of character. We deploy it exactly where it clears paperwork, and keep it well away from where human futures are decided. If someone promises you fully automatic hiring "with no human in the loop," be wary. What you save in time you may pay dearly for in fairness and reputation.