AI is getting better than humans in more and more jobs.
In one of the biggest natural field experiments of its kind, it reveals that AI voice agents not only outperformed human recruiters in conducting their first job interviews, but also performed. A study that analyzed nearly 70,000 applications showed that interviews using AI have significantly improved employment outcomes, including increased employment delivery rates, initiating new hires, and improving employee retention.
The study was a partnership between University of Chicago researcher Brian Javarian and Luca Henkel of Erasmus University Rotterdam, and PSG Global Solutions, a leading recruitment process outsourcing company. This experiment was conducted in the Philippines for an entry-level customer service position. This is a large amount of advanced industry where efficient and effective recruitment is important.
The 70,000 applicants were included in three groups: those interviewed by human recruiters, those interviewed with an AI voice agent named ANA AI, those who identified their identity at the start of the call, and those who were given the option to be interviewed by either a human or an AI agent. The final employment decision was made by human recruiters who reviewed interview performance (performed by themselves or by AI) and standardized test scores. This design allowed researchers to isolate the effectiveness of AI from the evaluation stage to the interview stage.
The results were surprisingly decisive. Applicants interviewed by AI Voice Agents 12% higher chance Receiving job offers, 18% higher chance Starting work, a 17% higher chance He remained employed for at least 30 days. These findings show that AI could be a more effective tool in the recruitment process than previously thought.
An analysis of interview transcripts revealed that AI is more systematic and effective in collecting relevant information. AI-led interviews are likely to be comprehensive and cover many of the recruiters on the topics needed to make decisions. AI's consistent, structured approach has drawn more valuable linguistic cues from applicants. For example, applicants for AIREDE interviews are engaged in more conversation exchanges. This is a feature that strongly predicts job openings in human-driven interviews when using low-level back-channel cues such as “UHM” or clear questions. Essentially, AI was excellent at teaching conversations and creating information that human evaluators value most.
Interestingly, there was no evidence of backlash against AI. Applicants satisfaction measured by net promoter scores was roughly the same between AI and human interviews. The job acceptance rate was the same. Most surprising, when given a choice, 78% of applicants chose to be interviewed by AI. This study suggests that this may be driven by a positive general attitude towards AI and the convenience of 24/7 scheduling. This study revealed an inquisitive “negative selection” effect. Applicants who choose AI tend to score lower on standardized language and analytical tests, suggesting that lower candidates may feel more comfortable with AI.
After all, human recruiters gave applicants interviewed by AI a significantly higher performance score. But their behavior has also changed. When making final employment decisions for candidates interviewed AI, recruiters placed more weight on standardized test scores and reduced the interview score itself. This suggests that while assessing the information collected by AI, they relied more on objective metrics when they were not conducting interviews themselves.
This study provides compelling evidence that AI can effectively automate the early stages of the employment process, resulting in positive results. Instead of completely replacing human recruiters, AI can free themselves from time-consuming interviews, allowing them to focus on more strategic tasks such as candidate evaluation and decision-making. While initial setups have costs and some applicants may hesitate to interact with AI, the technology is very quickly cost-effective and can significantly improve the quality and efficiency of adoption. Overall, the findings suggest that AI voice agents could be a powerful and practical tool for businesses looking to enhance their recruitment outcomes. And if this happens, it can affect the HR space. Nowadays, many junior employees are tasked with doing these very interviews that seem to be great at AI right now.
.
