A few years ago, this would have sounded ridiculous.

You apply for a job.

A few minutes later, you get invited to an interview.

There is no recruiter on the other side.

No scheduling.

No calendar ping-pong.

No waiting three days for someone to find a 30-minute window.

You talk to an AI.

It asks questions about your experience. It follows up. It records your answers. It evaluates them against the role.

And when you're finished, the recruiting team receives a structured summary.

We're already there.

Greenhouse reported this year that 63% of U.S. job seekers surveyed had experienced an AI interview.

LinkedIn has also reported increasing adoption of generative AI across recruiting teams.

And Bullhorn recently released a number that surprised me even more:

92% of candidates who had experienced an AI voice interview rated it as good as or better than a live recruiter interview.

That last number caught my attention.

Because if you spend enough time around recruiters, candidates and founders, you'll hear a completely different story too.

Candidates saying AI interviews feel dehumanizing.

Recruiters worrying about what the AI might miss.

Candidates trying to game automated assessments.

Employers trying to figure out whether candidates are using AI during the interview.

Both sides are bringing AI into the same conversation.

And both sides are becoming suspicious of the other.

I've been thinking about this broader question for a while. When I recently looked at Humand and the idea of building an entire “AI operating system” for HR, I came away with the same concern: adding AI to more stages of work doesn't automatically make those stages better. Read: Humand Raised $66 Million. But Does HR Really Need Another “AI Operating System”?

So I think we're asking the wrong question.

The question isn't:

Should AI interview candidates?

It already does.

The better question is:

Which parts of an interview should we actually trust AI to do?

If you've ever run recruiting at any kind of scale, the attraction is obvious.

Imagine receiving 500 applications.

Maybe 100 are plausibly relevant.

Even if you spend only 20 minutes speaking to each candidate, that's more than 33 hours of interviewing.

And that's before:

Now imagine an AI interviewer can speak to all 100.

At whatever time works for the candidate.

It doesn't care if the candidate wants to interview at 7 a.m. or 11:30 p.m.

It doesn't get tired by candidate 17.

It doesn't have a bad morning.

It doesn't accidentally spend 35 minutes with one candidate and 12 with another.

And theoretically, every candidate gets the same opportunity to answer the same core questions.

That's a powerful proposition.

Especially in remote recruiting.

At HiresLink, we're constantly trying to solve this problem across a large LATAM talent pool: how do you use technology to make sourcing and screening faster without allowing the technology to become the hiring decision itself?

Our current approach combines AI screening with human vetting rather than treating those as substitutes for each other. See how HiresLink approaches LATAM hiring and vetting

That distinction matters.

Because distributed hiring creates real coordination problems.

Different countries.

Different schedules.

Candidates who are currently employed.

Recruiters managing multiple roles.

Hiring managers in the U.S.

Anything that removes unnecessary friction is useful.

And there's another advantage I don't think we talk about enough.

Traditional recruiting is constrained by recruiter hours.

If 400 people apply, most aren't getting a call.

They're getting a résumé review.

Sometimes they're getting a six-second résumé review.

Sometimes they're being filtered by an ATS before a recruiter ever sees them.

So the comparison isn't always:

AI interview vs. wonderful 30-minute conversation with an experienced recruiter.

Sometimes it's:

AI interview vs. no interview whatsoever.

That changes the argument quite a bit.

If AI allows a candidate who would otherwise be rejected based on a résumé to explain why their background makes sense, there is potentially something very valuable there.

Bullhorn's research also suggests candidates value AI when it means faster communication, better matching and a quicker process.

I understand why.

The most common complaint about recruiting usually isn't:

It's the opposite.

Candidates apply.

Then nothing.

They wait.

Maybe they follow up.

Still nothing.

Eventually they assume they were rejected.

If AI can remove some of that silence, I'm all for it.

But this is where I start disagreeing with the more aggressive version of the AI recruiting story.

Efficiency is not the same thing as judgment.

The easiest parts of recruiting to automate are the parts where the answers are relatively objective.

Do you have experience with Salesforce?

How many years?

Are you comfortable working U.S. hours?

What compensation range are you considering?

Have you managed a team?

Have you worked in healthcare?

When could you start?

AI can handle a lot of that.

Probably better than making a recruiter repeat those questions 40 times.

But good recruiters aren't just collecting answers.

They're noticing things.

Someone says:

A recruiter might ask:

What did you personally own?

The candidate explains.

Something doesn't quite line up.

So you ask another question.

Then another.

Suddenly you realize they participated in the project, but didn't lead it.

That's not necessarily dishonesty.

People naturally simplify their experience when describing it.

Understanding the difference requires context.

Or consider the opposite.

A candidate gives a weak answer to the first question.

An automated scoring system might downgrade them.

An experienced recruiter might realize they're nervous, change how they ask the question, and uncover a very strong candidate five minutes later.

That's interviewing.

The question creates the conversation. It isn't the conversation.

One thing recruiting teaches you very quickly is that what someone can do and what someone actually wants to do next are completely different questions.

A résumé can tell you someone spent six years managing operations.

It doesn't tell you they're completely tired of managing operations.

Someone can be an excellent engineering manager and desperately want to become an individual contributor again.

Someone can technically meet every requirement while being totally wrong for what the founder actually needs.

And sometimes you discover this from one tiny comment.

Then you listen.

Not for keywords.

For the story.

That's where recruiting becomes much closer to judgment than filtering.

And I don't think we should be in a hurry to automate that away.

I expected the research on this to be fairly predictable.

It wasn't.

Greenhouse surveyed 2,950 job seekers across the U.S., UK, Ireland, Germany and Australia.

Among U.S. candidates who had experienced AI evaluation:

70% said the use of AI wasn't clearly disclosed before their most recent AI interview.

Only 18% said most employers had explicit, clear AI policies.

And 38% said they had already withdrawn from a hiring process because it included an AI interview.

That sounds terrible for AI interviews.

But then look at Bullhorn's research.

Among candidates who had actually completed an AI voice interview:

92% rated the experience as good as or better than a live recruiter interview.

And 93% of candidates who had a positive AI experience said they'd work with that staffing firm again.

Another AI interviewing company, Classet, reported that 87.5% of more than 10,000 post-interview ratings collected between January and August were positive.

Now, we should be careful here.

Companies selling recruiting technology obviously have an interest in AI adoption.

These aren't controlled experiments that settle the debate.

But taken together, I think they're showing us something important.

Candidates don't necessarily hate talking to AI.

They hate being treated badly by hiring processes.

And AI can either improve that problem or make it much worse.

63%
of U.S. job seekers surveyed by Greenhouse had experienced an AI interview.

38%
had withdrawn from a hiring process because it included one.

92%
of candidates surveyed by Bullhorn who had experienced an AI voice interview rated it as good as or better than a live recruiter interview.

26%
of candidates surveyed by Gartner said they trusted AI to evaluate them fairly.

Sources: Greenhouse, Bullhorn and Gartner.

Candidates can like an AI interview while distrusting AI hiring decisions.

Those are not the same thing.

And I think a lot of companies are going to confuse them.

An AI interview can be convenient.

It can be conversational.

It can happen immediately.

It can ask relevant follow-up questions.

A candidate might genuinely prefer that experience to waiting five days for a rushed recruiter to call them.

That doesn't mean the candidate wants an algorithm deciding whether they deserve a job.

Gartner found only 26% of candidates trusted AI to evaluate them fairly.

That's the gap I'm interested in.

Candidates seem much more comfortable with:

“AI is helping me through this process.”

than:

“AI is deciding whether I'm good enough.”

Those two experiences may look almost identical from the employer's side.

They feel very different from the candidate's side.

And that distinction should probably guide how we build these systems.

This debate gets framed strangely sometimes.

Companies using AI = concerning.

Candidates using AI = innovation.

Or the reverse.

The reality is both sides are using it.

Candidates use AI for:

We're entering a recruiting environment where:

AI writes the job description.

AI helps source the candidate.

AI writes the candidate's résumé.

AI evaluates the résumé.

AI prepares the candidate for the interview.

AI conducts the interview.

AI summarizes the interview.

And eventually a human receives a beautifully structured page telling them what two AI systems thought about each other.

At some point you have to ask:

Where did we actually meet the person?

That's the part I don't want recruiting to lose.

I wrote about this tension before when looking at AI screening tools: automation can absolutely remove repetitive work, but if you automate the parts where recruiters are supposed to exercise judgment, you're changing the job rather than simply making it faster. Read: AI's Hiring Hijack — Bots Screening Your Next Star (Or Stealing the Show?)

AI doesn't magically make a bad hiring process good.

If your company doesn't know what it's evaluating, automation just makes the confusion faster.

Suppose three hiring managers have completely different definitions of a “strong candidate.”

One values startup experience.

Another values enterprise experience.

One thinks five years of experience is essential.

Another doesn't care.

Nobody agrees on what questions actually matter.

Now add AI.

Congratulations.

You've automated ambiguity.

Before companies automate interviews, they need to be very clear about:

Otherwise the technology gives you a false sense of precision.

A candidate gets a 72.

Another gets an 84.

Those numbers look scientific.

But if the criteria underneath them are weak, the precision is meaningless.

This is something we deal with constantly when running direct-hire searches at HiresLink.

The difficult part isn't always finding people.

It's calibrating what the company actually needs before you start evaluating them. Our direct-hire process starts with defining the role, seniority and requirements before candidates are sourced and vetted. See HiresLink's direct-hire process

AI doesn't remove that work.

If anything, AI makes good calibration more important because bad criteria can now be applied to hundreds of candidates very quickly.

One argument for AI interviews is consistency.

And there's real value there.

Human interviewers absolutely have biases.

We like people who communicate like us.

People who went to schools we recognize.

People whose career paths make immediate sense.

People who remind us of previous high performers.

Even something as simple as interviewing someone at 9 a.m. versus interviewing them after six back-to-back meetings can affect the conversation.

Structure helps.

AI can help enforce structure.

But consistency isn't automatically fairness.

What happens when someone:

The U.S. Equal Employment Opportunity Commission has specifically warned that algorithmic hiring tools can unintentionally screen out people with disabilities and that employers may need to provide alternative assessment formats or reasonable accommodations.

That doesn't mean don't use AI.

It means the company deploying the system still owns the decision.

You can't outsource accountability to software.

No.

But I would let it conduct a lot of them.

And that's an important distinction.

If I were designing a hiring process today, I'd think about it roughly like this:

Potentially even a conversational first screen for high-volume roles.

Those are different jobs.

Trying to make one technology perform all three is where I think companies will get into trouble.

I actually think AI makes good recruiters more valuable.

Because once software handles:

and:

the recruiter has nowhere to hide.

Their value has to come from something else.

Understanding the business.

Understanding people.

Knowing when the obvious candidate isn't the right candidate.

Knowing when the non-obvious candidate deserves another conversation.

Convincing a great person to take a role.

Telling a founder their expectations are unrealistic.

Understanding why someone who looks perfect on LinkedIn isn't moving forward.

Finding the question that changes an interview.

That's recruiting.

AI should remove the administrative work surrounding those moments.

It shouldn't remove the moments.

They'll be the ones that know exactly where not to use it.

If an AI interview means a candidate gets a conversation today instead of a résumé rejection next week, that's an improvement.

If AI means every candidate receives consistent questions instead of wildly different interviews, that's an improvement.

If AI means recruiters spend less time scheduling and more time actually recruiting, that's an improvement.

But if AI means candidates never speak to anyone, don't understand how they're being evaluated, receive an automated rejection and leave wondering whether a human being ever looked at their application...

We've probably optimized the wrong thing.

The goal isn't an automated hiring process.

The goal is a better hiring process.

Sometimes those will be the same thing.

Sometimes they won't.

And knowing the difference may become one of the most important skills recruiters have.

— Julián