If you guessed "most of them," you'd be wrong. An independent audit of major AI screening tools found one widely-used model's scoring swung by a coefficient of variation of 29.5% when the same résumé ran through it multiple times.
In practical terms, a candidate averaging a score of 75 could land anywhere between 53 and 97, depending purely on which run happened to process them. Same CV. Same job. Same software. The same candidate scored like three different people.
That's not a rounding error. Whether you've got a talent function of fifty or you're the one personally reading CVs between calls, that difference is happening inside your hiring process right now.
Point solutions get purchased fast and integrated slowly. A team adopts a screening tool to solve one visible problem, too many resumes, not enough hours, without rebuilding the workflow around it. The tool runs, a shortlist comes out, and nobody checks whether that shortlist is a fair or repeatable read of the candidate pool.
That’s the trap: adoption looks like progress on a dashboard. Consistency is the thing that actually determines whether you’re finding the right person, or just the person the algorithm liked that day.
Organisations in the 44% aren't running AI instead of judgment — they're running it with judgment still in the loop. A resume gets parsed and ranked by AI, but a person reviews the edge cases, the near-misses, the ones the algorithm scored oddly. The technology does the repetitive first pass; a human still makes the call that carries risk.
That's the model worth building toward, whether you've got a dedicated recruitment team or you're doing this alongside running finance, product, or the whole company: AI for volume, a human for judgment, and a process that treats "the tool ran" as the start of quality control, not the end of it.
With Myn, that looks like 99%+ repeatable matching across repeat runs on the same brief, run the same brief again and you get the same shortlist, not a different draw of the dice.
Every shortlist also gets a human review before it reaches you, which catches edge-case candidates that the AI scored inconsistently on its own.
Organisations working with Myn report saving 20 hours a week per team member on sourcing and screening, time a talent team reinvests in candidate experience and stakeholder work, and time a founder or exec reinvests in, well, running the business.
If you're leading talent or people at scale, you don't just need a tool that works, you need one you can defend when someone asks how you know it's fair and repeatable.
If you're a founder, CEO, CFO, or CTO doing your own hiring, you don't have a dedicated ops team to catch a screening tool quietly cutting your shortlist in half. The tools you pick, and how you use them, are the whole system, there's no safety net behind you.
This is exactly why Myn builds AI sourcing and screening with a human always in the loop, rather than handing anyone a black-box tool and leaving them to catch its blind spots alone. The AI does the first pass at scale; a person makes the calls that actually decide who gets placed, whether you're running a talent function of fifty or hiring your first ten people.
The gap between using AI and having it work isn't a technology problem. It's a process problem, and it's one every organisation needs an answer for, not just the enterprise firms with the budget to build one.