AI is getting better than humans at more and more kinds of jobs.
In one of the largest natural field experiments of its kind, a new study reveals that an AI voice agent not only matched but outperformed human recruiters in conducting initial job interviews. The research, which analyzed nearly 70,000 applications, found that using AI to conduct interviews led to significantly better hiring outcomes, including higher job offer rates, more new hires starting, and improved employee retention—all without sacrificing applicant satisfaction.
The study was a partnership between researchers Brian Jabarian of the University of Chicago and Luca Henkel of Erasmus University Rotterdam, and PSG Global Solutions, a major recruitment process outsourcing firm. The experiment was conducted in the Philippines for entry-level customer service positions, a high-volume, high-turnover industry where efficient and effective recruitment is critical.
The 70,000 applicants were put into three groups: applicants interviewed by a human recruiter, applicants which were interviewed by an AI voice agent named Anna AI, which disclosed its identity at the start of the call, and applicants who were given the choice to be interviewed by either a human or the AI agent. The final hiring decision was made by a human recruiter who reviewed the interview performance (either conducted by themselves or the AI) and a standardized test score. This design allowed the researchers to isolate the AI’s effectiveness at the interviewing stage from the evaluation stage.
The results were surprisingly definitive. Applicants who were interviewed by the AI voice agent had a 12% higher likelihood of receiving a job offer, an 18% higher likelihood of starting the job, and a 17% higher likelihood of remaining employed for at least 30 days. These findings demonstrate that AI can be a more effective tool in the hiring process than previously thought.
An analysis of the interview transcripts revealed that the AI was more systematic and effective at gathering relevant information. AI-led interviews were more likely to be comprehensive, covering more of the topics recruiters needed to make a decision. The AI’s consistent and structured approach also elicited more valuable linguistic cues from applicants. For example, applicants in AI-led interviews engaged in more conversational exchanges—a feature that strongly predicts a job offer in human-led interviews—while using fewer low-signal backchannel cues like “uhm” or asking clarifying questions. In essence, the AI was better at guiding the conversation to produce the information human evaluators valued most.
Interestingly, there was no evidence of a backlash against the AI. Applicant satisfaction, as measured by the Net Promoter Score, was nearly identical between the AI and human interviews. Job offer acceptance rates were also similar. Most strikingly, when given the choice, 78% of applicants chose to be interviewed by the AI. The study suggests this may be driven by positive general attitudes toward AI and the convenience of 24/7 scheduling. The study did uncover a curious “negative sorting” effect: applicants who chose the AI tended to have lower scores on standardized language and analytical tests, suggesting that lower-scoring candidates may feel more comfortable with an AI.
Human recruiters, as it turned out, gave significantly higher performance scores to applicants interviewed by the AI. However, their behavior also changed. When making a final hiring decision for an AI-interviewed candidate, recruiters placed more weight on standardized test scores and less on the interview score itself. This suggests that while they valued the information collected by the AI, they relied more heavily on objective metrics when they hadn’t conducted the interview themselves.
This study provides compelling evidence that AI can effectively automate the initial stages of the hiring process, yielding positive results. Instead of fully replacing human recruiters, AI can free them from time-consuming interviews so they can focus on more strategic tasks like candidate evaluation and decision-making. While the initial setup has a cost and some applicants may be hesitant to interact with an AI, the technology becomes cost-effective very quickly and can significantly improve the quality and efficiency of recruitment. Overall, the findings suggest that AI voice agents could be a powerful and practical tool for businesses looking to enhance their hiring outcomes. And if this happens, it can have impacts in the HR space, where currently many junior employees are tasked with conducting these very interviews that AI now seems to excel at.
.