Positive signs are emerging that the U.S. employment picture is improving. Private sector hiring rose in December after shedding jobs the previous month. Annual pay was up. And layoffs hit their lowest pace since the summer of 2024.

Of course, that’s little consolation to the knowledge workers who have thrown their hands up in frustration looking for a job. They may want to blame artificial intelligence for kicking out their polished resumes, but here’s the deal: AI is unfairly getting the blame for what are uniquely human decisions.

The real issue here is automation amplifies bad filters at scale. So, if an HR team has bad practices to begin with, its tech stack is likewise going to reject more candidates.

As someone with a long career in HR technology, I have collaborated with a myriad of employers large and small. It is time to pull back the curtain on what is largely a misunderstood process once an applicant clicks the “submit” button on an application.

‘AI rejected me two hours after applying’

One of the worst gut punches for a candidate is the email that starts with “Thank you for applying to such-and-such a role. We have decided to move forward with other candidates whose skills better align with the role.”

The job seeker is confused because the resume checked every box in the job description. Did a human even look at the CV? Probably not if there were 500 other applicants. That’s what AI is for.

More likely, the system is screening candidates on filter criteria set by the talent acquisition team (i.e. “must have X years of experience” or “master’s degree required”). If that criteria is too narrow or broad, AI doesn’t instinctively disposition candidates—it does so with relentless perfection because that’s what humans told it to do.

There goes the notion that a sinister little black box is making corporate hiring decisions. As if it were that simple.

There’s an oft-stated statistic that says 75% of resumes are rejected by HR technology before a human even sees them. The stat has been repeated so often that it has become an accepted fact, but it has been challenged as not grounded in credible research. Seventy-five percent? Please. That’s an organization that clearly does not want to hire anyone.

A more credible fact is that there are some 27 million “hidden” U.S. workers—people who can work but are screened out by common practices and automated screening/ranking, according to a joint Harvard Business School/Accenture study. The report was released in 2021, which just goes to show how pervasive the problem was and still is.

The point isn’t that the robot is evil. But rather, once a bad proxy becomes a rule, software turns it into a brick wall.

Read the full article here.