New Delhi, Nov. 1 -- For someone looking for a job, one of the most frustrating experiences is applying for a position and hearing absolutely nothing back. No call. No rejection. Sometimes not even an acknowledgement that the application was received.
As companies increasingly use AI to screen resumes, match candidates to jobs, assess skills and analyse interviews, there is a possibility that some of that silence may now come from an algorithm.
The technology can process thousands of applications in a fraction of the time a recruiting team would need. That is useful. But after spending more than two decades working with HR, payroll and recruiting systems, I have learned that the hardest part of hiring is rarely processing the applications. It is designing a process that works fairly for the people going through it.
Early in my career in India, I worked on a payroll system for a nationalised bank covering more than 62,000 employees. I helped redesign recruiting at a large healthcare employer in California, USA, that receives tens of thousands of applications a year.
When we looked closely at that process, the numbers told us something important. Roughly a third of applicants were abandoning the application midway. Candidates were frustrated.
Recruiters were frustrated. Resume parsing, online portals, and the offer process created their own problems.
Our first response was not to add more automation.
We simplified the application by asking candidates for less information upfront and collecting additional details later. We also rebuilt a manual offer-letter process that had grown to well over a hundred variations.
The experience reinforced something I have seen repeatedly: a hiring problem can begin as a process-design problem long before it becomes a technology problem. That distinction matters as organisations bring AI into recruitment.
AI can make an inefficient hiring process faster. It does not necessarily make it better.
Consider what happens when an AI system rejects a candidate. A recruiter can usually see the person who was shortlisted. But who looks at the person who disappeared from the process?
A candidate who is incorrectly rejected does not show up as a hiring error unless someone deliberately looks for that error.
That is why I would ask organisations adopting AI in recruitment a few basic questions.
Are we checking for false negatives?
Companies often measure how many candidates AI helps identify. They should also examine whom it may be filtering out. Periodically reviewing rejected applications can reveal patterns that a dashboard will never show on its own.
Is the system evaluating skills or simply looking for familiar words?
A candidate may have the right experience but describe it differently from the language used in a job description. This can be particularly important for people changing careers, returning to work after a break, or coming from smaller institutions and less conventional career paths.
A keyword is not a skill.
Can every candidate use the system fairly?
Automated assessments and video-based evaluations can create additional barriers for candidates with disabilities or for people who may not be equally comfortable with the language, format or technology being used. Accessibility should not be something an organisation checks after deploying the system. It should be part of the design.
Perhaps the most basic question is, "If an AI system ranks someone poorly, can a human explain why?"
If the employer cannot explain the decision, it becomes very difficult for either the recruiter or the candidate to challenge a bad outcome.
None of this means organisations should stop using AI in hiring. There is a lot that automation does extremely well. In my own work, I have seen automation eliminate repetitive tasks and give teams back thousands of hours. There is real value in that.
But hiring is different. A job is not simply another transaction in a workflow. For the person applying, it can affect income, family stability, career progression and, in some cases, an entire change in the direction of their life.
That is why the responsibility for an important hiring decision should never disappear behind a piece of software.
India has an opportunity to get this right while AI adoption in recruitment is still developing. The Digital Personal Data Protection Act, 2023 has already brought greater attention to the responsible handling of personal data, while developments such as the European Union's AI Act are putting greater scrutiny on high-impact uses of AI, including employment-related systems.
Indian organisations hiring at large scale should not wait for a major controversy before thinking about oversight.
My own approach would be fairly simple. Keep a named human accountable for consequential hiring decisions. Regularly review rejected applications, not just successful ones. Test whether the system is evaluating actual skills rather than resume language. Check accessibility before deployment. And measure candidate experience alongside metrics such as time-to-hire and cost-per-hire.
The objective should not be to screen a lakh resumes as quickly as possible. It should be to ensure the right person doesn't disappear from the process simply because a machine failed to recognise them. That is why I believe in keeping a "human-in-the-loop." AI is here to help people, not to replace them.
The writer is an HR-technology leader and an author. Views expressed are personal
Published by HT Digital Content Services with permission from Millennium Post.