A Recruiting Platform Just Launched With ‘Zero Automated Decisions.’ Here’s Where That Breaks.

On July 14, 2026, recruiting platform uRecruits launched version 2.0 with a specific, quotable pitch: AI scores and organizes candidates, but a recruiter makes every advance, reject, or approve decision. No button exists to auto-reject anyone. “The best part of what we’re building is that AI doesn’t reject or approve anybody, that stays with people,” CEO Thomas Alexander told the press (PR Newswire, July 14, 2026). The launch came out of a 150-team beta and is positioned squarely as a response to tightening AI hiring law, referencing Illinois’s amended Human Rights Act, NYC’s bias-audit requirements, and the ongoing Workday discrimination litigation.

It’s a coherent answer to a real compliance pressure, and it’s worth taking seriously rather than dismissing as marketing. It’s also an answer built for a hiring volume that doesn’t resemble campus recruiting in India, and the gap between those two contexts is worth naming precisely, because it’s the same gap a lot of AI-hiring compliance advice glosses over.

The math “a recruiter reviews everything” doesn’t survive

A 150-team beta customer base for a general recruiting tool is plausibly hiring in the dozens or low hundreds of candidates per role. At that volume, routing every single AI-scored candidate through a human for a final call is not just feasible, it’s good practice. There’s no meaningful cost to a recruiter confirming a handful of decisions the AI already leaned toward.

Campus hiring in India runs at a different order of magnitude. A single opening at a mid-size Indian enterprise routinely pulls 800 to 3,000 applicants; enterprise drives run 9 to 14 of these per year. In a structured screening pilot we ran, 1,099 interviews got completed across active campaigns, with the system auto-rejecting 60%+ of candidates who failed the same knockout criteria a recruiter would apply anyway, just applied consistently in the first 90 seconds of a call instead of somewhere between minute six and burnout. Routing every one of those clear-fail conversations through a human reviewer doesn’t make the process safer. It makes the queue the size it was before AI screening existed, just with an extra click added to each rejection.

What “human in every decision” is actually protecting against

The compliance pressure uRecruits is responding to is real and worth taking at face value. Illinois’s disclosure law, NYC’s Local Law 144 bias audits, and the pending Workday litigation are all, at their core, asking the same question: can a candidate find out an algorithm was involved in rejecting them, and is there a person accountable for that outcome. “A human clicks reject on literally everyone” is one way to satisfy that. It’s not the only way, and for high-volume hiring it’s not the practical one.

The alternative that scales: auto-decide the clear top and bottom of the distribution, and route only the genuinely ambiguous middle band, typically 40-55% of a screened pool, to a person for real judgment. That’s not “zero human involvement.” It’s human involvement concentrated where a human’s judgment actually changes the outcome, instead of spread thin across thousands of decisions that were never close calls. We’ve written before about what that middle-band routing does to a recruiter’s actual day: the short version is that a recruiter reviewing a stack of obvious auto-rejects learns nothing they didn’t already know from the score, while a recruiter reviewing a queue of genuinely ambiguous cases is doing the job a human was actually needed for.

The honest tradeoff, stated plainly

There’s a real argument for “a human touches every single decision,” and it’s not solely about compliance optics: it catches edge cases an auto-decide threshold will occasionally miss, and it keeps a person accountable in a way that’s easy to explain to a regulator or a candidate who asks. That argument gets stronger as volume drops and weaker as it climbs. At 150 beta customers hiring at corporate scale, it holds. At 3,000 campus applicants for 40 openings in a single drive, insisting on it doesn’t protect candidates any better, it just guarantees the recruiter reviewing the back half of that queue is exhausted, inconsistent, and making worse calls than the auto-decide threshold would have on the clear-cut cases.

The right design question isn’t “does a human touch every decision.” It’s “does a human touch every decision where their judgment would actually change the outcome.” Those are different standards, and conflating them is how a genuinely good compliance instinct turns into a bottleneck that doesn’t make anyone safer.

If you’re weighing where the line between automated and human-reviewed should sit for your own hiring volume, talk to us — we can show you what that split looks like against your actual applicant numbers rather than a generic recommendation.