AI made applying easier, and hiring harder.

A few years ago, a recruiter filling a role might have reviewed around a hundred applications. Today, that number is closer to 300 per hire, and it's taking companies nearly a quarter longer to fill a role than it did before the pandemic. Nobody singlehandedly willed this into being; it’s just what happens when both sides of hiring pick up the same tool at the same time.

Both sides armed themselves, and nobody won

Here's the mechanism: candidates started using generative tools to tailor applications at scale, sending out dozens of polished, customized resumes in the time it used to take to write one. Companies responded by deploying their own screening tools to manage the flood. Each side's tool was a reasonable answer to the other side's tool, and the combined result industry commentators have started calling a doom loop: applications that increasingly read alike, sorted by systems built to handle volume rather than to notice what's distinctive about any one of them.

Volume was supposed to be the problem AI solved. Instead, it's the problem AI helped create, and then had to be deployed again to manage. And the cycle goes on and on, ad infinitum.

The confusion companies are sitting in

If you're leading a hiring team right now, the honest question isn't "should we use AI?" Most already do, in some form or another. The better, more important question is where would it do good versus where is it adding a filter on top of others that aren’t themselves working.

Two things are true at once: nearly every recruiter and hiring manager surveyed in recent industry research reports having spotted or suspected candidate deception in the past year, whether that's an AI-polished resume overselling real experience or, in rarer cases, fabricated credentials outright. At the same time, more filtering doesn't fix that. It just processes the same homogenized pool faster, and a resume that reads well is not the same thing as a person who'll do the job well.

That's the bind. More automation on the intake side doesn't solve a problem that's happening because of automation on the intake side. You can read that again.

What's left to differentiate on?

When resumes converge and the first-pass screen is a commodity every company can buy, the signal that's left to tell candidates apart isn't more of the same kind of information, processed faster. It's a different kind of information.

A person describing themselves on paper is answering the question "how do I present well." A person talking through an actual problem is answering a different question entirely, one that's considerably harder to optimize for at scale in an afternoon.

The takeaway

The arms race isn't going to slow down by adding another filter to the intake pipeline. Every company running that play is already running it, and the applications keep converging anyway. The more useful question is what kind of signal a hiring process is actually built to read, not how fast it can process the signal it already has.