Hundreds of applications for one role: how to shortlist without reading every résumé

When one job post brings in hundreds of applications, the instinct is to screen faster: keywords, filters, an AI résumé ranker. That speeds up a step that has stopped working. The better move is to change the order. Ask every applicant for a short piece of real work first, and read résumés only for the people whose work was good.

Why has résumé screening stopped working?

A résumé was useful when writing a good one took effort and the effort told you something. Now a candidate can produce a tailored résumé and cover letter for your exact job post in under a minute, and many do. The result is a pile where everyone has the right keywords and nobody stands out.

Recruiters are saying so. In Robert Half's survey of more than 2,000 US hiring managers, published in March 2026, 84% of HR professionals reported heavier workloads from reviewing and validating applications, and 67% of HR leaders said AI-generated applications have slowed their hiring. Greenhouse's 2025 report found 34% of recruiters spend up to half their week filtering out spam and junk applications.

Adding an AI screener on your side turns this into an arms race: candidates' tools write for your tool, and your tool rewards whoever's tool wrote best. Neither side is measuring whether the person can do the job.

What do most teams do, and what does it cost?

Faced with three hundred applications, most teams fall back on what is quick to check:

• College and marks. Easy to filter, and it throws out capable people from colleges you do not know.
• Years of experience. Among the weakest predictors of performance that hiring research has measured.
• First come, first read. The recruiter reads until they have enough, and applicant number 180 is never opened.

Each of these is fast and each gives up on finding the best person in the list. The cost is hidden, because you never meet the people you screened out.

What works instead?

Move the work to the front. In place of "apply, then we screen, then a few of you get a task", make it "apply, do a short task, and we read the results".

It sounds like more effort. In practice it is less, for three reasons.

1. The list shrinks by itself. People who applied to two hundred jobs with one click do not do a thirty-minute task. The ones who do it want this job.
2. You compare like with like. Every applicant did the same task and is marked the same way, so ranking them is quick and defensible.
3. You read fewer résumés, with better attention. You look at the résumés of the top fifteen, not the first fifty.

How do you set it up?

1. Pick one task from the real job that takes thirty to sixty minutes. Our guide to work sample tests has examples for six roles.
2. Give each applicant their own version of the material, so answers cannot be shared and AI tools do not give everyone the same result.
3. Write the marking before you post the job.
4. Put the task link in the job post itself, as the way to apply. Asking for a name, an email and the task is enough at this stage.
5. Give a clear deadline, such as three days from when they start, and send one reminder.
6. Read the ranked results, then interview from the top, asking each person about their own work.

Allow AI tools and say so. Read more on what to test now that AI can finish a take-home.

Won't good candidates refuse to do a task?

Some will, and it is worth being honest about that. Three things keep the number small.

• Keep it short. Thirty minutes is a fair ask. Three hours is not.
• Make it clearly about the job. Candidates accept a task that shows them what the work is like. They resent a puzzle.
• Tell them what happens next. Say that everyone who hands in is looked at and that the interview will be about their own work. A candidate with a plain résumé and real ability hears that as good news.

That last group is the reason to do this. The people a résumé screen is hardest on (the first in their family to go to college, the career changer, the graduate of a college nobody on your team has heard of) are the ones a task gives a fair chance.

What should you measure?

Three numbers tell you whether it is working:

• Completion rate: of those who opened the task, how many handed in. If it is very low, the task is too long or the brief is unclear.
• Hours of your team's time per hire, before and after.
• How your hires do at ninety days. This is the one that matters, and the only one that tells you whether your shortlist was the right shortlist.

Wooble is built for this. You pick the role, share one link, and every applicant does a real piece of the job on their own material. Applicants need no account. You get a ranked list with the reason for every place, and your first hand-ins are free. See how it works.