The conversation about AI in talent acquisition has been dominated for two years by two equally unhelpful positions. The first: AI will transform everything, and those who don’t adopt it immediately will be left behind. The second: AI is overhyped, the tools don’t work reliably, and human judgment cannot be replaced.

The senior practitioners in the TalentDNA roundtable had moved past both. They had used the tools. They had measured the outcomes. And they had formed views that were specific, evidence-based, and — in several cases — genuinely surprising.

“AI will not replace talent acquisition professionals. But AI will definitely replace talent acquisition professionals who do not understand AI. That is not a prediction about the future. It is already happening.”

— Senior TA Leader, Roundtable Participant

Where AI is actually moving the needle

The most concrete data point of the evening came from a leader at a global technology firm. When their recruiting team moved from standard outreach to AI-assisted personalised messaging, candidate acceptance rates went from 32% to 53%. Not through a different sourcing strategy. Not through a new ATS. Through a better first message, sent at the right moment, calibrated to the candidate’s context.

This is the version of AI adoption that works: targeted, measurable, and tightly connected to a specific bottleneck in the funnel. Not a platform overhaul. A precision intervention at the point where the drop-off was happening.

A second data point came from a firm that had implemented AI pre-screening agent for technical roles. Since deployment, average time to technical panel had fallen to 14 minutes, and the selection ratio — the proportion of screened candidates who made it through to offer — had improved by 10% in a single quarter.

The 20-minute shortlist

One of the most practical examples of the evening involved no specialised AI tools at all. A leader faced a same-day escalation: source qualified profiles for an international market they had no prior experience in, with no existing pipeline, no local knowledge, and no time.

Their approach: pull 10 to 15 profiles that the hiring manager had previously shortlisted for similar roles. Upload the corresponding interview feedback. Ask an AI assistant to identify the pattern of what was being selected and what was being rejected. Then apply that pattern to screen new profiles from a cold search.

In under 20 minutes, eight qualified shortlists were in the hiring manager’s inbox. The escalation was resolved. The tool used was freely available to anyone with an internet connection. The insight was in knowing how to use it.

Funnel intelligence: seeing what you could not see before

Beyond individual interventions, several leaders in the room were building something more systemic: AI-powered funnel visibility. The idea is straightforward. For every open role, track how many profiles were sourced, how many were submitted, how many were shortlisted, how many were rejected at each stage, and by whom. Then use AI to identify patterns in that data.

Is a particular skill cluster producing a high rejection rate? Is it a sourcing problem — the wrong profiles being submitted — or a calibration problem, where different interviewers are applying different standards? Is the drop-off happening at first screen, at technical assessment, or at the final stage? Each of these diagnoses requires a different intervention. Without the data, TA is firefighting. With it, TA is engineering.

One leader described the ambition clearly: build a system where a recruiter can look at any open role and immediately see where the funnel is breaking. Not after the hire fails. While it is still happening, when something can still be done about it.

The risks the room took seriously

The roundtable was not a technology enthusiasm session. The practitioners in the room had earned their scepticism, and they applied it honestly.

The first risk raised was verification. As AI tools become more capable of generating polished responses, the credibility of candidate-supplied information — assessments, work samples, even video interviews — becomes harder to establish. Several leaders had already encountered candidates using AI to complete technical assessments in real time. The industry’s response frameworks are nascent. The problem is not.

The second risk was infrastructure cost. AI adoption is expensive at scale. The compute, storage, and API costs associated with running AI across a high-volume hiring function are real and growing. Leaders who treat AI as a cost-free efficiency play will encounter the bill eventually. Those who model the economics upfront will make better decisions about where to deploy it and where not to.

The third risk — and the one that generated the most discussion — was dependency. The room had seen what happens when a function becomes over-reliant on any single tool or platform. The question for AI is the same as it was for LinkedIn, for ATSs, and for job boards before that: what happens to your capability when the tool is unavailable, unreliable, or simply wrong? The answer has to be that your professionals can still function. If they cannot, you have built fragility, not efficiency.

“I think about cognitive dissonance. Every smoker believes they will not be in the 5% who get cancer. A lot of people in our industry believe AI disruption will happen to everyone except them. History is not kind to that assumption.”

— Veteran TA Leader, 20+ Years in the Profession

What AI-ready actually looks like

The roundtable converged on a practical definition of what it means for a TA function to be genuinely AI-ready. It is not about tool count or platform subscriptions. It is about three things.

First, literacy: the ability to form a good prompt, evaluate an AI output critically, and recognise when the model is wrong. This is a teachable skill, and organisations that invest in it systematically will outperform those that assume it will develop on its own.

Second, judgment: the confidence to use AI as an input to a decision without outsourcing the decision itself. The practitioners in this room had not handed their hiring decisions to algorithms. They had used algorithms to make better-informed human decisions. That distinction matters enormously.

Third, positioning: a clear view of where in the AI adoption spectrum each member of the team sits — from basic tool usage to workflow orchestration to building agents — and a deliberate plan for moving them along it. The function that is still debating whether to use AI in two years will not be competitive. The function that has invested now in developing genuine capability will be.

We thank the talent leaders and industry professionals who participated in the TalentDNA Leadership Roundtable, including Rajyalakshmi Sadhineni, Shweta Desai, Narane Gundabathula, Chitra Velala, Santhosh Musley, Sheela Kshirsagar, Vinod Kumar Shalva, Sanjay Samuel, Goutham Vishwanath, Santhosh Varghese, Abdul Gafoor, Chanakya Veladi, Sriram Gabbita, Vanitha Lingala and other members of the talent community. The session was moderated by Ramesh Talagampa. The views and takeaways presented in this article are derived from the broader discussion and are intended solely for knowledge-sharing purposes. They should not be attributed to any individual participant or their respective organization.