Where the time actually goes
Break a typical thirty-day fill into its parts and the shape is consistent. Two or three days pass before anyone opens the applications. Another week goes to reading them, usually in gaps between other work. Phone screens take a few more days to schedule around candidate availability. Only then does the client see anyone.
The interview loop itself, the part everyone tries to optimise, is often the fastest stage. It has a scheduled slot and a named owner. The screening queue has neither.
Why screening resists scheduling
Reading two hundred resumes is not one task. It is two hundred small judgement calls, each cheap on its own and expensive in aggregate. Work like that expands to fill whatever attention is left over, which on a busy desk is very little.
It also degrades. The hundredth resume gets less care than the tenth, and the candidates at the bottom of the pile are systematically disadvantaged by nothing more than upload order.
What AI compresses, and what it does not
Automated screening removes the queue rather than the judgement. Every resume is read against the job description at the same level of attention, and each one comes back with a score, the reasoning behind it, and the gaps that pulled it down. What took a week now takes minutes.
It does not remove the recruiter. A model can tell you a candidate has never shipped anything at the scale this role requires. It cannot tell you the hiring manager will forgive that for someone who interviews well, or that this particular client has been burned before by exactly this profile. That knowledge lives with the person running the desk.
- Compressible: reading, parsing, first-pass matching, gap identification, scheduling a first conversation
- Not compressible: client relationships, judgement calls on borderline candidates, negotiating an offer
The number worth measuring
Time-to-hire is a client-side metric and depends on their calendar as much as your work. Time-to-shortlist is yours: how long between receiving a job description and putting credible candidates in front of the client.
It is also the number worth selling. An agency that reliably delivers a ranked, explained shortlist within days of a brief has a concrete answer to why it should keep the account.
Keeping the bar where it was
Faster screening is only an improvement if quality holds. Two habits matter. First, insist on explainability: a score with no reasoning cannot be checked, and cannot be defended when a client questions a ranking. Second, keep a person in the loop on rejections. Automated rejection is where speed turns into a fairness problem nobody notices, because the people affected never find out.