David Sterling: Megan, long week — tell me you caught the Greenhouse story.
Megan Skiendel: Caught it and immediately forwarded it to three people. Honestly, Daniel Chait naming this publicly is the thing I can't stop turning over.
David Sterling: Well, here's what grabs me structurally — the 412% increase in applications per recruiter. That's Greenhouse's own platform data. They filled two million jobs in 2024, they're sitting on 175,000 live listings right now, Orianna Rosa Royle publishes the piece in Fortune on July 27, and Chait calls it a doom loop.
Megan Skiendel: Four hundred twelve percent — I mean, what does a recruiter's inbox even look like at that scale?
David Sterling: 254 applicants per job on average. And the thing that made that number possible — AI subscription tools, about $20 a month, letting anyone auto-apply to hundreds of roles. The per-application cost collapsed. That's the event.
Megan Skiendel: Twenty dollars a month is the number that reframes this for me — because that's not a tech-savvy candidate behavior anymore, that's a rational response to a system that candidates already don't trust. And listen, 87% of companies are running AI somewhere in their funnel, 98% of Fortune 500 using ATS with AI filtering. So candidates figured out the math: if AI is almost certainly screening me out anyway, I spray everywhere.
David Sterling: Which is exactly what Chait is diagnosing. The question I need answered — is this a Greenhouse-specific data story, or is it market-wide? Because a CEO whose platform profits when hiring is complex has a structural reason to call the alarm.
Megan Skiendel: That's the gap. And it's the gap we're going to sit in — because the doom loop framing is either a genuine warning from someone with two million filled jobs worth of data, or it's a very well-timed product narrative.
David Sterling: But hold on — that gap you're naming, is it actually a gap, or is it the point? Because the doom loop isn't complicated. Candidates use cheap AI to flood applications. Employers use AI to filter the flood. Rejected candidates apply to even more jobs with even more AI. Every rational individual move makes the collective outcome worse.
Megan Skiendel: Think of it like a spam filter arms race. The more spam gets sent, the better the filters get. The better the filters get, the more spam you have to send to get one message through. Nobody wins — the inbox just gets louder. That's it. That's the whole thing.
David Sterling: Clean. And the recruiter inbox right now — over 400 applications at any given time.
Megan Skiendel: Right — but here's what's actually new versus what the headline overstates. The loop itself isn't new. What's new is the $20 a month. Auto-apply tools at that price point didn't exist at scale before 2024, 2025. That specific price drop is what broke the model — not AI generally, not ATS, not résumé formatting. Twenty dollars.
David Sterling: The per-application cost going to near zero. That's the event.
Megan Skiendel: And then — honestly, this is the number — only 26% of candidates trust AI to evaluate them fairly. Which means 74% are mass-applying while actively assuming the system won't give them a fair read. They're not gaming it. They've just... accepted it's broken, and they're compensating.
David Sterling: Wait — so the loop is inverted on the candidate side. They're not applying more because they're optimistic. They're applying more because they've already assumed rejection.
Megan Skiendel: That's the feedback loop running backwards. And that's what Chait's Greenhouse data can't fully capture — the $20 tool made spray-and-pray cheap, but the 74% distrust figure is what made it feel rational. Those two things together are what's genuinely new.
David Sterling: And that's where the conflict of interest bites hardest — because Chait's diagnosis and Chait's revenue pitch are the same sentence. 'The doom loop is real, you need better AI filtering.' That's also a Greenhouse product brief.
Megan Skiendel: No, I buy that. But hold on — does the conflict disqualify the data?
David Sterling: Not automatically. Look — two million filled jobs in 2024 is a real sample. But the 254-applicants-per-job figure and the 412% recruiter inbox increase are Greenhouse's own proprietary numbers. Not independently audited. So Chait is simultaneously the witness, the analyst, and the vendor selling the solution.
Megan Skiendel: And the solution is always more filtering. Which he sells. Listen, here's what actually breaks the confident framing — SHRM's 2026 Talent Trends puts AI adoption in HR organizations at 39%. Thirty-nine. The number floating around everywhere else is 87%. Those cannot both be describing the same market.
David Sterling: Right — and that gap is methodological. Vendor surveys pull from clients who already bought in. SHRM is sampling the full HR universe, including mid-market, healthcare, companies where AI means spell-check. So if the actual adoption floor is closer to 39%, the doom loop might be real but narrow — concentrated in tech hiring and high-volume corporate recruiting, not universal.
Megan Skiendel: Which is exactly what Chait's customer base looks like.
David Sterling: Frankly, yes. And then Gartner drops the other shoe — 88% of HR leaders report no significant AI business value yet. So you've got high adoption, no demonstrated value, and a CEO saying the fix is more of the same tool. That's not a warning. That's a sales cycle.
Megan Skiendel: And meanwhile, hireEZ's researchers went through 200,000 actual résumés and found hidden prompt injection text in roughly 1% of them — white-font instructions telling the AI screener to rank the candidate highly. And that 1% is climbing. Picture someone on a Sunday night pasting invisible text into the bottom of her résumé before she submits to Workday. That's not a hack. That's a person who has correctly read that the system isn't evaluating her honestly, and acted accordingly. Which — honestly — is the part that makes the incentive problem even harder, and we'll get to why no one with skin in the game is structurally positioned to fix it.
David Sterling: The 1% is small. But it's the direction that matters — it's climbing. And it confirms the 26% trust figure isn't just survey noise. It's already showing up in behavior.
Megan Skiendel: And that's exactly where Shraddha Sunil and Mudit Saraf land in HBR — June 8, 2026 — work samples, knockout questions. Signals that AI literally cannot mass-produce on a candidate's behalf. That's the structural exit. It exists.
David Sterling: It shifts the cost. Candidate completes a work sample — that's an hour, maybe three. Small employer has to design and grade it. Who absorbs that?
Megan Skiendel: Not the Fortune 500. They can run assessment centers. It's the 200-person company posting on Greenhouse that actually can't.
David Sterling: Right — and the companies already doing this quietly? They won't say so publicly.
Megan Skiendel: Because the moment you announce 'we use work samples now,' you've implicitly admitted the AI tooling you just bought — the ATS, the screening layer — wasn't actually working. That's an expensive thing to say out loud.
David Sterling: So the exit is invisible by design. The incentive to take it privately exists. The incentive to publicize it — that's what doesn't exist.
Megan Skiendel: And meanwhile — I mean, this is the part that actually costs something — the doom loop self-corrects eventually. When mass-applying stops paying off, candidates stop. But 'eventually' is doing enormous work there, because right now it means burned-out recruiters and qualified candidates ghosted by filters that never explained why.
David Sterling: The human cost accumulates at the current equilibrium, not the future one. That's the load-bearing problem.
Megan Skiendel: So watch Greenhouse specifically. If they build verifiable-signal products into the platform — work samples, scored knockouts — that's the tell. That's Chait acknowledging the AI-on-AI arms race has actually hit a ceiling, even if he'll never say it that plainly.
David Sterling: The part I can't resolve — Chait named the doom loop. Greenhouse's business model depends on the AI filtering layer that perpetuates it. Those two facts sit in the same sentence and neither cancels the other out. So the question I'm left with: does anyone with actual power to simplify the funnel have any incentive to do so before the talent market just... fractures? Into companies that can afford to hire well and everyone else.
Megan Skiendel: And the exit requires all three — candidates, recruiters, platforms — to coordinate on doing less. Which is not how competitive markets work. So honestly? I don't know. And I'm not sure the pain has gotten bad enough yet to force a reset. Maybe it has to get worse first.
David Sterling: That's the thread. We don't have an answer. We just have a CEO who profits from the problem, a 412% inbox explosion, and 74% of candidates who've already accepted the system isn't fair. That's where it sits.
Megan Skiendel: Good one to sit with. Thanks for this.
