When AI interviews AI: Why your hiring funnel is broken, and how to fix it

hiring worse.

Why traditional screening has stopped working

There’s a concept in economics called signalling theory, proposed by Nobel laureate Michael Spence. The idea is simple: in a job market, candidates invest in costly signals—degrees, certifications, polished resumes—that are harder for weaker candidates to produce. This cost difference is what lets employers separate strong candidates from weak ones.

AI has collapsed that cost difference to zero. When any candidate can produce a flawless resume for ₹500 a month and pass a coding test with an invisible AI copilot, the signal stops separating anyone. Strong and weak candidates look identical on paper. Economists call this a pooling equilibrium—everyone pools together, and you can’t tell who’s who.

Worse, this triggers what George Akerlof described as a market for lemons. When buyers (employers) can’t distinguish quality, the best sellers (top talent) leave the market because they’re being valued the same as everyone else. What remains is a progressively weaker pool—exactly what many hiring managers are experiencing today.

Add to this Goodhart’s Law—“when a measure becomes a target, it ceases to be a good measure”—and you see the full cascade. Resume keywords became ATS targets, so they stopped measuring skill. Online test scores became AI-assist targets, so they stopped measuring competence. Each metric, the moment candidates learned to game it, lost its informational value.

India amplifies every one of these dynamics. We produce 1.5 million engineering graduates a year, but only 16-26% are employable by industry standards. Our $6.5 billion coaching culture—built on “cracking” exams—has fused with generative AI. HirePro’s analysis of 4.3 million assessments found 30-50% of entry-level candidates cheat during online job assessments. This isn’t a few bad actors. It’s system-wide information collapse.

Hiring the candidate, not the resume

The answer is not better AI detectors. That’s an arms-race you will always lose—the cheating tools evolve faster than the detection tools. The answer is to redesign your hiring architecture so that gaming becomes irrelevant.

Here are four shifts that work:

First, assess the process, not the product. Stop evaluating what candidates produce (code files, written answers) and start evaluating how they produce it. Live pair programming, system design with real-time follow-up probes, and “unexpected questions” that break rehearsed scripts. Genuine expertise handles novelty; AI-assisted performance falls apart under it.

Second, extend the evaluation window. A 45-minute interview is a prediction. A two-week paid trial project is an observation. Contract-to-hire models, hackathon-based hiring, and structured probation periods transform hiring from guessing into knowing. TCS’s National Qualifier Test, conducted across 6,800 centres, cut recruitment cycle time by 66% and classroom training by 80%—proof that better assessment design works at Indian scale.

Third, make AI literacy the test, not the threat. Instead of trying to catch candidates using AI, assess whether they can work effectively with it. Can they prompt well? Can they evaluate and debug AI-generated output? Can they do the work when the copilot is taken away? The question shifts from “Did they cheat?” to “Can they deliver?”

Fourth, build internal talent markets. The most reliable signal of capability is demonstrated performance inside your organisation. Internal talent marketplaces—where existing employees are matched to new roles based on what they’ve actually delivered—bypass the adversarial dynamics of external hiring entirely. Unilever, Schneider Electric, and Standard Chartered already operate these at scale.

The real question

That Bengaluru-to-Delhi flight cost us a few thousand rupees. The real cost was weeks of recruiter time, panel hours, and pipeline capacity wasted on a candidate whose qualifications existed only in AI-generated text.

The question for every CHRO and Talent Acquisition Leader today is not “How do we catch cheaters?” It’s “How do we design systems where cheating is structurally irrelevant?” When AI can produce any signal, the only reliable signal is observed performance under real conditions. The future of hiring belongs to those who build for that truth.

The author is an HR leader.

DISCLAIMER: The views expressed are solely of the author and does not necessarily subscribe to it. will not be responsible for any damage caused to any person or organisation directly or indirectly.