Naukri Just Launched an AI Sourcing Platform. Sourcing Was Never the Bottleneck.

Naukri rolled out an AI-powered recruitment suite this month, headlined by AI REX, an autonomous sourcing agent that finds and shortlists candidates with minimal recruiter input. It’s a real product aimed at a real problem, and for TA teams still building longlists manually, it will save time. But watch what it doesn’t touch: once AI REX hands a recruiter a longer, faster-built list of candidates, someone still has to work out which of them can actually do the job. That step — not sourcing — is where high-volume hiring in India has been breaking for years.

The part of the funnel everyone keeps automating

Sourcing tools have gotten good. Job boards, resume aggregators, and now agentic AI sourcing all compete on the same axis: get more candidates in front of a recruiter, faster. That’s a real improvement over a decade ago, and it’s also not the constraint anymore. A ₹30K-CTC campus role in India can pull tens of thousands of applications on its own without any sourcing tool’s help. The problem was never finding candidates. It’s what happens after: a recruiter, or a small team of them, has to work through that list and decide who’s worth a conversation.

We watch this pattern in campus hiring specifically. A single evening drive can bring 3,000 candidates through a structured screening conversation in about two hours — the volume problem was never getting to 3,000 applicants, it was doing something useful with them once they’d arrived. Faster sourcing makes that pile bigger without making the evaluation any faster. If anything, a genuinely good sourcing agent makes the screening bottleneck worse, because it hands the recruiter a bigger stack of resumes to get through in the same amount of time.

Why the fix keeps getting misdiagnosed

It’s an easy mistake to make, because sourcing is the visible, measurable half of the funnel. You can point to a dashboard and show “candidates sourced this week” going up. Screening quality doesn’t show up on that same dashboard — it shows up six months later as attrition, or it shows up right now as a recruiter working punishing hours trying to phone-screen a queue that never gets shorter. We’ve made this argument before about speed versus rejection accuracy: the metric that’s easy to report isn’t always the one that’s actually broken.

What good sourcing plus weak screening looks like in practice

Picture the workflow after a tool like AI REX does its job well: a recruiter now has a noticeably longer list of qualified-looking candidates than they had last quarter. Nothing about their capacity to actually talk to those candidates changed. The same recruiter, running the same manual phone screens at the same pace, hits the same wall they hit before — except now there are more people waiting behind it. Better sourcing without a corresponding jump in screening throughput just moves the bottleneck earlier in the funnel; it doesn’t remove it.

In our own pilot work, structured AI voice screening has cut the HR time spent per candidate by 89% compared to manual phone screens, with auto-rejection running above 60% on campaigns where the volume made manual screening impossible in the first place. That’s not a sourcing number. It’s what happens when the constraint actually gets addressed instead of moved.

The honest caveat

None of this makes sourcing tools pointless — a better longlist is genuinely useful when the team behind it has the capacity to work through it. The mistake is treating a bigger, faster-built candidate list as progress on its own. It’s progress on the easy half of the problem. If your screening capacity doesn’t grow alongside your sourcing capacity, you haven’t fixed anything — you’ve just moved the queue further upstream and made it longer.

What actually scales with a bigger longlist

The fix isn’t slower sourcing, it’s screening that scales the same way sourcing just did. A structured voice conversation with every candidate, run at whatever volume the sourcing tool hands over, does the job a manual phone screen physically can’t once the list gets long enough — it doesn’t run out of hours in the day, and it doesn’t get worse at the job by candidate three hundred the way a tired recruiter does. That’s the actual argument for automating the second half of the funnel, not the first: sourcing tools compete on getting you to a bigger pile faster, and the only thing that pairs well with a bigger pile is a screening layer that doesn’t slow down as the pile grows. Recruiters aren’t the constraint because they’re bad at their jobs. They’re the constraint because a real human conversation, done well, doesn’t compress no matter how good the tool that built the queue in front of them is.

If your sourcing just got a lot faster and your screening didn’t, see what a rejection-first screen looks like on your own volume.

Source: Naukri launches an AI recruitment platform in India for smarter hiring — YourStory