A few weeks ago, my executive assistant sent the same email fourteen times to job applicants: "We need to cancel your interview." Every one of those people had cleared our initial AI screen, blocked time on their calendars, and started preparing to talk with me about a director-level role on my team. We posted the job, drew more than 100 qualified applicants, and used AI to evaluate them against the job description. I asked her to book the top candidates who cleared it. Then 150 more applications arrived, and before I looked up she had over 30 interviews on my calendar.

That was the moment I realized the process had outrun the thinking behind it. I stepped in, re-evaluated every candidate myself, and made the call to cut the interview list roughly in half.

Some of the candidates wrote back. They were angry, and they were right. We had led them on, and we had looked disorganized doing it. The decision was mine, and so was the mistake. I moved fast and forgot that speed always has a cost, and that the cost rarely lands on the person who chose to move fast. It lands on the person who rearranged their week.

What I did on a small scale is happening across the job market at industrial scale. In an April 2026 Enhancv survey of 1,066 U.S. job seekers , 50.5% said they had been rejected at least once in the past year without a single word from a human, and 68.5% said no employer ever told them AI was involved. Half of candidates are being filtered by a system they cannot see, cannot question, and were never told existed. AI did not create that problem. It industrialized it.

So I did something either brave or reckless, depending on your perspective: I wrote about the fourteen cancellation emails on LinkedIn. The post set off a firestorm. One reader told me I was "casually playing with people in a time where job hunting is quite literally hunting." Another argued that turning a hiring mistake into a post was itself the problem. Plenty defended me and told me to give myself grace. The disagreement was the point. Hiring right now is raw enough that a single cancelled interview can become a referendum on whether a company respects people at all.

Is AI Actually Making Hiring Better, Or Just Faster?

The speed gains are real. In a ResumeBuilder survey of 948 business leaders , 82% of companies said they use AI to review resumes, and 51% already run it somewhere in hiring. Posting a role, ingesting hundreds of applications, and ranking them against criteria now takes hours instead of weeks. For any leader running a lean team, the appeal is obvious.

But efficiency in volume processing is not the same as efficiency in hiring. The companies with the fastest time-to-offer are often the same ones with the highest early attrition, because speed optimized for the employer's convenience and speed optimized for finding the right person are not the same objective. The problem was never that AI screens candidates. The problem is that most companies never defined what they were screening for. AI is fast at sorting. It is only ever as good as the criteria it is handed.

What Does A Hiring Process Built For Humans Look Like?

The framework that reset my thinking comes from Who: The A Method for Hiring , by Geoff Smart and Randy Street. Built from more than 1,300 hours of interviews with over 20 billionaires and 300 CEOs, the method reports a 90% hiring success rate , against the roughly 50% most managers achieve. The central move is almost embarrassingly simple: decide what success in the role looks like before you talk to a single person.

Smart and Street call this the Scorecard. It names the mission of the role, the specific outcomes the hire must deliver in year one, and the competencies that separate someone who can do the job from someone who merely interviews well. The Scorecard comes before the job description, before sourcing, and long before anyone books an interview. If I had written one first, I would have known exactly who I was looking for before any applications arrived. I would have closed applications before opening a calendar. Thirty slots would never have been booked, and fourteen people would never have gotten that email.

The cost of skipping that step does not stay with the candidate. Smart and Street put the price of a single senior hiring mistake at $1.5 million once you count direct costs, lost productivity, and team disruption. The sharpest thing anyone said in that LinkedIn thread reframed the whole problem in one line.

If the interviewers' time isn't reserved, the organization isn't ready to interview. And if it isn't ready to interview, it isn't ready to post Polly Rowland, CPG career coach

How Do You Use AI Without Treating People As Throughput?

The leaders getting this right treat AI as a way to execute a Scorecard, not a substitute for having one. The pattern is consistent.

Write the Scorecard before you post the role. What does this person need to accomplish in 90 days, and in a year? What does an exceptional hire look like next to a merely acceptable one? If you cannot answer that in writing before you launch the search, you are not ready to hire.

Close applications before you schedule a single interview. Running open applications and open calendars at once is where the chaos starts, because one cannot close fast enough while the other fills up in real time. Some founders go further and add deliberate friction to the application itself. One, Spencer Carlson, described how moderate friction filtered out mass applicants and sharply raised the share of strong first interviews.

Screen against the Scorecard, not against keywords. Keyword matching rewards people who have learned to game resume filters. Scorecard screening asks whether a candidate’s actual history of outcomes maps to what the role demands. The inputs you feed the AI decide the quality of everything it hands back.

Communicate at every stage, and do not hoard the process. A candidate who gets a clear rejection in three days feels respected; one who waits three weeks for a form email does not. Delegation helps here too. As Erin Lockley noted in the thread, she coaches her team on exactly what they are looking for and lets them lead many of the interviews rather than routing every conversation through herself.

What Does It Cost To Get This Wrong?

The trust data should worry any founder who assumes hiring is still a buyer’s market. Only 26% of candidates trust AI to evaluate them fairly , according to a Gartner survey of 2,918 job seekers, even as most never learn that AI touched their application at all. Skepticism is highest at the exact moment companies are most likely to hand the decision to a machine and walk away.

That gap is expensive in ways no time-to-hire dashboard captures. A candidate who gets a fast, honest answer is far more likely to reapply, refer a friend, or become a customer than one who vanishes into a black hole. In a market this brutal, someone saying yes to an interview is already handing you their hope. As one founder in the thread, Steve Benbow, put it, a broken hiring process can damage trust with good people faster than most brand work can rebuild it.

The next person you hire will either be someone your team celebrates in six months or someone you quietly regret, and the difference is usually not the candidate but whether you knew who you were looking for before you started looking. Write the Scorecard, close the applications before you open the calendar, and let the machine execute a standard you set rather than one it invents. Hiring is the rare place where slowing down at the start is the fastest way to get it right.