The Application Funnel May Be Breaking

For years, recruiting has been built around a fairly simple idea:

Post the job. Collect applications. Screen the applicants. Interview the best. Hire one.

But I’m starting to wonder whether the application funnel itself is becoming the problem.

Greenhouse’s March 2026 Hire Standard report analyzed more than 640 million applications across over 6,000 companies between 2022 and 2025. In that dataset, applications per job increased 111%, while recruiters per organization fell 56%. Its findings also give recruiters credit: monthly hires per recruiter increased 122%.

That is not evidence that recruiters suddenly forgot how to recruit. It is a sign of people doing more work inside a system under pressure. Those figures describe Greenhouse’s dataset, not every employer, and they do not establish that AI caused the changes.

Still, they make the question harder to ignore: are we improving the hiring process, or simply getting better at surviving it?

Two automation systems walk into an ATS

Candidates can use AI to find more jobs, tailor résumés, write cover letters and apply at greater scale. Employers can respond with automated screening, matching, assessments and rejection.

The cycle practically writes itself.

Candidates hear less, so they apply more. Employers receive more, so they automate more. A better-looking application becomes easier to produce, while deciding what it actually demonstrates becomes more important.

We’ve built two increasingly sophisticated automation systems that can end up fighting each other across an ATS.

Somewhere in the middle is a person who wants a job and a manager who needs somebody capable of doing the work.

They would probably appreciate being introduced.

I’m not arguing against automation. Scheduling, organizing evidence and communicating promptly are useful things to automate. My concern is what we are optimizing.

Processing a larger pile faster is valuable. Understanding why the pile became necessary is a different question.

Why did 5,000 people need to apply?

I don’t think the future of recruiting is simply figuring out how to process 5,000 applications faster.

I think the better question is: Why did 5,000 people need to apply in the first place?

That number is an illustration, not a benchmark. The point is the design.

An application asks a person to enter a process before either side may have enough information to know whether it makes sense. A job description can obscure what matters behind a long wish list. A résumé can obscure capability behind titles and familiar company names.

Then we ask software to reconcile the two documents.

If the inputs are vague, faster comparison does not necessarily produce better understanding.

Talent Intelligence should help us get upstream of that exchange: understand the work, the available capability, the constraints and the person’s interest before treating another application as a successful outcome.

Capability first. Proxies second.

Start with a different question:

What must this person actually be capable of doing?

Before searching, the hiring manager and recruiter should separate the essentials from the preferences. Which capabilities must be present on day one? Which can transfer from a different environment? Which gaps can reasonably be learned? Which credentials or contextual requirements are genuinely necessary?

That creates a clearer basis for discovery and evaluation.

Titles, companies and industries still matter. They provide context and clues. They should not automatically become the definition of capability.

A familiar title can hide a very different job. An unfamiliar title can hide highly relevant work.

The practical challenge is to connect a requirement to evidence: what did the person do, under what conditions, and how does that relate to what this role needs?

A list of extracted skills is not enough. Otherwise we’ve simply built another keyword matcher wearing a nicer suit.

Industries don’t own skills. Workers do.

This is where skills adjacency becomes interesting.

Traditional sourcing often asks: “Where do people doing this job work?”

A broader question is: “Where are people learning how to do this work?”

Those questions can lead to different places.

Consider a hypothetical technical role requiring experimental design, failure analysis and complex equipment troubleshooting. Some relevant experience may exist outside the employer’s immediate competitors.

That does not mean every related skill transfers cleanly. The operating environment, depth of expertise and time required to become effective still matter.

It means adjacency deserves investigation rather than automatic rejection.

I would want the evidence separated clearly: demonstrated capability, credible adjacent experience, something that still needs verification, and an actual gap.

“Not documented” and “cannot do it” are not the same conclusion.

Nor should a polished résumé be mistaken for verified proficiency. A structured conversation or relevant work sample may still be needed.

Better matching cannot become a closed door

There is an obvious danger in saying fewer people should need to enter the funnel.

We could build a system that makes the same narrow assumptions earlier—and calls the result efficiency.

That would miss the point.

Getting upstream should make opportunity easier to understand and relevant people easier to discover. It should not make an opaque algorithm the only route to consideration.

People still need an accessible way to express interest, provide missing evidence and be considered beyond a system’s initial assumptions. Recruiters need room to question the matching logic.

Candidate intent matters, too. Someone who looks suitable on paper may not want the role, its location, its compensation or its working conditions. Matching should help both sides make an informed choice.

Fewer unnecessary applications would be useful. Fewer opportunities for capable people would not.

Give human attention a better job

I would measure progress beyond application volume and screening speed.

Are hiring teams getting clearer about what success requires? Are they finding relevant capability outside the usual titles and companies? Are candidates receiving useful information sooner? Are fewer people spending hours on processes that could have been ruled out with an honest conversation?

And when a person reaches an interview, does the team have a better question to ask than “Walk me through your résumé”?

That is where I want automation to earn its place: helping people arrive at a more informed conversation, with the uncertainty visible rather than hidden behind a score.

The technology can organize evidence. People still need to test assumptions, exercise judgment and own the decision.

Brett’s Perspective

The goal shouldn’t be more applications, more screening and more automation.

It should be better matches, less noise and more human attention where it actually matters.

Talent Intelligence offers a way to ask better questions before the funnel fills up: what capability is needed, where it exists, what could transfer, and whether there is a real reason for both sides to talk.

Maybe the next breakthrough in recruiting isn’t a better application funnel.

Maybe it’s needing the funnel less.