AI Hiring: What's Working, What's Broken, and Where Humans Still Matter
Some candidates go through three full rounds of interviews before they ever speak to a person.
Welcome to the era of AI hiring.
A chatbot screens the resume. A one-way video records answers for nobody in particular. An AI voice agent asks polite follow-up questions. And somewhere along the way, a tool is scoring how the candidate showed up, their eye contact, their pacing, their energy.
Now picture the parent who was up at 3am with a sick kid and logged on for that video interview. A human on the other side of the screen would notice, cut them some slack, and focus on what they actually said. An AI interviewer wouldn’t.
To be clear, I'm not anti-AI. I use it every day. But after nearly two decades in recruiting, I've never seen a tool this useful get used this carelessly.
So let's talk about where AI belongs in the hiring process, where it doesn't, and what founders should understand before handing the first three rounds to software.
"After nearly two decades in recruiting, I've never seen a tool this useful get used this carelessly."
Why AI hiring is everywhere now
None of this happened because recruiters got lazy. It happened because of math.
Greenhouse, which tracks hiring across more than 6,000 companies, found in its Hire Standard report that applications per job more than doubled between 2022 and 2025, from 116 to 244. Over the same stretch, the number of recruiters per organization fell by 56%. More applications, fewer people to read them.
And a lot of those applications aren't really written by the people sending them. In Greenhouse's 2025 Workforce & Hiring research, 67% of US job seekers said they use AI in their search, and 22% said they'd used bots to submit applications for them. There are tools that will fire off hundreds of applications a day on a candidate's behalf, many to jobs the candidate has never actually read.
Greenhouse's CEO calls this the AI doom loop. Candidates use AI to apply to more jobs. Employers use AI to filter the flood. Candidates use more AI to get past the filter. Everyone is solving their own problem in a way that makes the whole system worse.
So companies reached for AI hiring tools, and fast. Nearly two thirds of US job seekers (63%) have now been interviewed by an AI, up 13 points in six months, according to Greenhouse's 2026 Candidate AI Interview Report.
I get it. If I was staring at 250 applications for one role with a smaller team than I had three years ago, I'd want help too. The question isn't whether to use AI. It's how.
What's going wrong with AI in the recruitment process
Candidates often don't know it's AI
In Greenhouse's 2026 Candidate AI Interview Report, a survey of 2,950 job seekers, among US candidates who'd been through an AI interview, 70% said it wasn't clearly disclosed beforehand. One in five only figured it out once the interview had started. Imagine walking into a meeting and realizing halfway through that nobody is actually there. That's the first impression a lot of companies are making.
Soft signals are being scored by systems that can't read them
Some AI interview tools assess mannerisms and presentation like eye contact, vocal cadence, facial expression, and how "engaged" someone seems. These get treated as neutral data points. They aren't.
Disability advocates have warned that tools measuring eye contact and voice can disadvantage autistic and blind candidates. Researchers at the University of Washington found that AI resume screening ranked candidates lower when their resumes mentioned disability-related awards or memberships.
Then there's context. The parent running on four hours of sleep. The candidate whose first language isn't English and who pauses before answering. The person who's nervous because they really, really want this job. A human interviewer can account for all of that. An AI can't, so it penalizes people for things that have nothing to do with whether they can do the work.
Bias at scale is a different problem
Human interviewers are biased too. But human bias is inconsistent, which means some great candidates get through anyway. A biased algorithm applies the same flawed standard to every applicant, every time. It turns an uneven barrier into a uniform one.
When a separate University of Washington team tested language models on more than 550 real resumes, the models favoured white-associated names 85% of the time and female-associated names just 11% of the time.
Candidates stop being themselves
Researchers publishing in PNAS ran 12 studies with more than 13,000 participants and found that when people believe AI is assessing them, they play up their analytical side and downplay empathy, intuition and emotion.
In other words, AI doesn't just miss human signals. It teaches candidates to hide them. And those are often exactly the qualities you're hoping to hire for.
Nobody closes the loop
38% of US candidates who completed an AI interview never heard anything back. No rejection, no update, nothing. More people were left hanging than were moved forward.
I've said this before: ghosting isn't a time problem, it's a systems problem. If you have a tool smart enough to interview someone, you have a tool smart enough to tell them where they stand.
Fraud is getting harder to spot
Companies are also dealing with candidates who aren't who they say they are. Gartner predicts that by 2028, one in four candidate profiles worldwide will be fake, and 6% of the candidates it surveyed admitted to interview fraud. A process with no human contact until the offer stage is the easiest place for that to slip through.
The legal ground is shifting
If you hire in Ontario and have 25 or more employees, you've been required since January 1, 2026 to disclose in public job postings whether you use AI to screen, assess or select applicants. In the US, the ongoing Mobley v. Workday case is testing whether AI hiring tools, and the companies using them, can be held responsible for discriminatory outcomes. "The vendor said it was fine" is getting harder to stand behind.
(I'm not a lawyer, so talk to one about your specific situation.)
What's lost without the human touch
Here's what all of this builds toward. Judgment, empathy and context aren't nice-to-haves in hiring. They're part of what makes a hiring decision fair. Strip them out in the name of efficiency and you pay for it somewhere.
You lose the two-way conversation. An interview is supposed to be the candidate evaluating you as much as you're evaluating them. AI can't really answer questions about the team, the culture, or what the first 90 days look like. The best candidates notice, and some of them conclude that if leadership can't make time for a first conversation, they won't make time later either.
You lose candidates. Greenhouse found that 38% of candidates have walked away from a hiring process specifically because it included an AI interview, and another 12% say they would. These are people who decided not to work for you before you ever got the chance to decide on them.
You lose customers. Someone close to me went through several rounds with a major consumer electronics brand, spent hours building a final presentation, and never heard back. Our household will never buy their products. That's one candidate. Multiply it by everyone an automated process has ghosted and the employer brand damage stops being theoretical.
And you raise the odds of a bad hire. When the signals you're scoring aren't the signals that predict success, you end up hiring the people who perform well for software. A bad hire costs more than a good one saves, and I've written about what a bad hire actually costs a startup.
What AI hiring gets right
Now for the part that might sound like it contradicts everything above.
When AI is used in the right place, it works. In a field experiment out of Chicago Booth, researchers randomly assigned 70,000 applicants to be interviewed by either human recruiters or an AI voice agent. Human recruiters made every hiring decision either way. The AI-interviewed group was 12% more likely to get an offer, 18% more likely to start the job, and 17% more likely to still be there after 30 days.
Why? The AI interviews were more structured and consistent, so recruiters had better information to decide with. Given the choice, 78% of applicants actually picked the AI interviewer.
There are caveats. These were entry-level customer service roles, not a VP of Sales. But the lesson is the one I keep coming back to: AI is great at collecting consistent information. People are still better at deciding what that information means.
That matches my own work. AI helps us source faster, summarize calls, tighten scorecards and take admin off our plates. It doesn't sit in the debrief and tell a founder, "On paper this person's great, but here's what I'm worried about." That part is still our job.
8 rules for using AI hiring tools without losing the human
Tell people when AI is involved. Put it in the posting and the interview invite, and explain what the tool actually measures. In Ontario it's the law. Everywhere else, it's just respect.
Get a human in by round two. If a candidate's third conversation with your company is still with software, you don't have a process. You have a maze.
Don't let AI judge body language. Eye contact, pauses, fidgeting, nerves, "energy." Leave the little human stuff to humans. If a tool scores how someone shows up rather than what they say, turn that feature off.
Leave room for context. The sleepless parent, the nervous candidate, the person who needs an extra second to think. Build accommodations into the process instead of making people ask for them.
Let AI sort, and let people decide. No candidate should be rejected without a human looking at the decision.
Don't ghost people. Every candidate gets an answer. If AI can screen them, it can send the update. Ontario now requires employers to tell interviewed applicants the outcome within 45 days anyway.
Let candidates ask for a person. 46% of candidates want the option to request a human interview instead. It's one checkbox.
Test the human stuff with a human. Candidates hide empathy and intuition when they know AI is assessing them, so don't try to measure those traits in an AI round. Assess them in a live conversation with someone on your team.
Frequently asked questions about AI hiring
Is it legal to use AI in hiring in Ontario?
Yes, with disclosure. Since January 1, 2026, Ontario employers with 25 or more employees must state in publicly advertised job postings if they use AI to screen, assess or select applicants. Other jurisdictions have their own rules (New York City requires bias audits, and the EU AI Act treats hiring AI as high-risk), so check with employment counsel for your situation.
Can AI replace recruiters?
It can replace a lot of recruiting tasks. If I started a recruitment team today compared to five years ago, that team would be cut in half. What it can't replace is judgment: pushing back on a role definition, reading a candidate in context, or telling a founder something they don't want to hear.
How should we tell candidates we use AI?
Keep it plain. One line in the posting, then more detail in the interview invite: which step uses AI, what it measures, that a person reviews every decision, and how to request an alternative. Something like: "We use automated tools, including AI, to help review applications. A member of our team reviews every decision."
So where do you draw the line?
All of this to say: AI isn't the problem in hiring. Using it without judgment is.
The companies getting this right aren't the ones with the most tools. They're the ones who decided, on purpose, where the software stops and the people start.
That's a lot of what we do at InTandem. We work inside your team, build a hiring process that uses AI where it genuinely helps, and keep senior judgment on every decision that matters. If you're trying to figure out where AI fits in your hiring (or you suspect your current process is quietly losing you great candidates), I'd love to chat. Here's how we work.
