Few people would defend the way hiring works today. Candidates submit dozens of applications and frequently receive no response, recruiters struggle with overwhelming volume, and hiring managers compare people using inconsistent information.
The entire process is frustrating for nearly everyone involved.
That is why the findings from a recent field experiment involving 70,000 job applicants caught my attention. This was not another prediction about how AI might change hiring. The researchers tested it inside an actual recruiting process and measured what happened.
Applicants were randomly assigned to interviews conducted by human recruiters or AI voice agents. In both cases, human recruiters evaluated the interviews and made the hiring decisions.
The AI did not decide who should be hired. It improved the information available to the people making that decision.
The Results Deserve Attention
Applicants interviewed by the AI agent were 12% more likely to receive job offers. Those gains translated into more job starts and higher employee retention, with no decline in the productivity of those hired.
After 30 days, AI-interviewed applicants were 18% more likely to remain employed. That advantage persisted through 120 days, suggesting the results were not simply caused by novelty or a short-term difference in candidate behavior.
The candidate experience did not suffer either. Satisfaction scores were nearly identical between the AI and human interviews, and most applicants who were given a choice selected the AI option.
The explanation for these results may be the most interesting part of the study.
AI interviews were more structured and consistent, but they still adapted to individual candidates. They covered more of the recommended topics and collected more hiring-relevant information.
This is an encouraging example of AI being deployed well. The technology addressed a clear process problem, humans remained accountable for the decision, and the improvement was measured through actual employment outcomes rather than estimated time or cost savings.
Instead of asking AI to imitate human judgment, the process used AI to give humans better information.
But the Interview Is Only One Part of Hiring
As impressive as the results are, they raise a larger operational question: What happens to the value of a better interview when the rest of the hiring process remains broken?
Before reaching an interview, candidates must first survive application triage. Job requirements may be inflated, screening criteria unclear, and recruiters overwhelmed by more applications than they can reasonably review.
If qualified candidates are screened out at the top of the funnel, a better interview cannot help them.
Problems can also appear after the interview. Candidates may encounter repeated rounds, slow approvals, inconsistent evaluations, poor communication, and delays in receiving an offer. Once hired, a weak onboarding process can undermine what appeared to be a strong match.
A company can collect better information and still fail to use it effectively.
This is not a criticism of the research. The experiment had a defined scope, and the results demonstrate that improving the interview created value that continued into employment.
But companies should resist interpreting an improvement in one step as proof that AI has fixed hiring.
Improve the System, Not Just the Activity
This is a common challenge in transformation work. An organization improves one activity without examining the decisions and handoffs surrounding it.
The local improvement may be real, but the constraint can simply move somewhere else. Faster and more consistent interviews will have limited impact if application triage, hiring-manager availability, offer approval, or onboarding becomes the next bottleneck.
Organizations considering AI in recruiting should ask:
- Where are qualified candidates currently being lost?
- Which step is actually limiting the speed or quality of hiring?
- What happens immediately before and after the activity being improved?
- Are we measuring the complete hiring outcome or only one step?
The most important lesson from this research is not that companies should replace recruiters with voice agents. It is that reducing unnecessary variation in how information is collected can help people make better decisions.
That same discipline should now be applied across the rest of the hiring process.
AI made the interview better.
Now we need to fix what happens before and after it.