Behind the Rise of AI Agents Replacing Human Recruiters

AI voice agents have beaten human recruiters in a head-to-head test involving 67,000 job interviews, delivering results that are rocking the recruitment market. 

Researchers from the University of Chicago Booth School of Business and Erasmus University Rotterdam find that AI-led interviews generate 12% more job offers and keep staff on the books 17% longer.

For companies struggling with recruitment costs and staff turnover, those numbers matter.

The experiment focuses on entry-level customer service roles in the Philippines. Job seekers were randomly assigned to be interviewed by AI voice agents, human recruiters or given their pick. The bots made clear they were artificial from the outset.

Humans still made the final hiring calls based on how candidates performed in interviews and standardised tests. But the AI systems proved better at sticking to script, covering more ground and asking the right questions. When conducting thousands of interviews, human fatigue becomes a real problem.

“The AI spoke less and prompted the interviewee to speak more,” says Brian Jabarian, one of the study’s authors at Booth’s Roman Family Center for Decision Research, where he examines how technology is changing workplace productivity.

The study offers some of the hardest data yet on whether corporate AI investments actually pay off, yet not everything went smoothly.

The problems from the study and what they imply 

Technical glitches hit 7% of AI interviews, while another 5% of candidates hung up rather than talk to a robot. 

Applicants also found the AI “significantly less natural” than human conversations – which is hardly surprising given current voice technology limitations.

Yet among those who stuck with it, 70% rated their AI interview positively compared to about half who spoke with humans. That caught the professional recruiters off guard – as most expected the technology to flop.

When given the choice, nearly 80% of candidates opted for AI interviews. 

The convenience factor played an additional role – because candidates could schedule calls whenever suited them. The study also found candidates held generally positive views about AI, which likely boosted take-up rates.

The results surprised even the researchers – as social interaction has long been seen as a human forte, something machines struggle with despite advances in natural language processing.

“I was quite surprised the AI voice agent was as good as it was at collecting data through social interaction,” Brian says.

Why short term success may not lead to long term success

However, the study shows how better performance doesn’t automatically mean better business cases. 

While AI interviews could be scheduled faster, human recruiters needed twice as long to wade through the results – and those efficiency gains vanished quickly.

The findings echo recent research from MIT showing 95% of enterprise AI projects failed to deliver measurable returns despite billions in corporate spending.

Whether AI recruitment tools make financial sense depends on circumstances, such as small firms in regions where recruiters earn modest wages might struggle to justify the upfront costs. 

Whereas large corporations processing thousands of applications in expensive labour markets could see real savings.

The quality angle matters too. If AI really does identify better candidates who stick around longer, companies save on replacement costs – particularly valuable in sectors like call centers where staff turnover runs high.

"We have to move from the ‘possible’ discourse to the hard-data discourse, so we don’t lose our rationality,” Brian says.