At least seventeen Fortune 500 CHROs have rolled back, paused, or materially restricted their AI-powered resume-screening tooling in the first quarter of 2026. ENTRA's count, built from a combination of vendor disclosures, CHRO conference remarks, internal-comms leaks, and direct conversations with three sitting CHROs at companies in the count, includes employers across banking, retail, industrials, and healthcare. The vendors affected span the obvious names — Workday, SAP SuccessFactors's AI module, Eightfold, Paradox, HireVue's resume layer — and several smaller players. The rollbacks are not coordinated. They are converging on the same conclusion from different directions.
The conventional read of this trend, which has begun appearing in HR-press coverage, is that the rollbacks are about regulatory exposure: New York City's Local Law 144, the EU AI Act's high-risk-system designation for hiring tools, and the cascading state-level bias-audit requirements in California, Illinois, and Maryland. That reading is incomplete. Regulatory exposure is the cover story. The actual driver, per every CHRO we've spoken to in the last six weeks, is signal collapse.
The signal-collapse story works like this. From mid-2024 onward, the proportion of inbound applications carrying AI-generated content grew from a single-digit percentage to a majority. By the start of 2026, the most-cited internal benchmark we've heard — repeated in three separate CHRO conversations — is that roughly 70 to 80 percent of resumes received for white-collar roles at large enterprises now show evidence of AI authoring or AI-assisted polishing. The resumes are grammatically clean, structurally consistent, keyword-rich, and indistinguishable from each other. The screening AI, trained to identify those exact features, ranks them all in the top decile. The funnel produces a monoculture.
That monoculture is the failure mode. A Fortune 500 talent leader at a top-five US bank described the dynamic to us in March: "We were running 40,000 applications through the screener and getting 8,000 'top tier' candidates back. Three years ago we'd have gotten 1,200. The model wasn't broken. The input was. We were optimising AI for AI."
Three rollback patterns are emerging across the seventeen companies in our count.
The first pattern: removal of resume-screening AI from senior roles entirely, while retaining it for high-volume hourly hiring. The logic is that the signal-collapse problem is most acute in white-collar professional applications — where AI tooling is most accessible — and least acute in hourly retail, warehouse, and frontline service roles where applicants are submitting through phone-first or kiosk-based flows that limit AI authoring. JPMorgan's commercial-banking talent organisation has reportedly moved in this direction; Walmart's corporate-headquarters hiring stream is reported to be evaluating a similar split.
The second pattern: replacement of resume-AI with structured behavioural signal capture. Several of the rollbacks we've tracked involve substituting the resume-screening layer with a short, async work-sample task — usually a 20-to-40-minute exercise designed to surface domain reasoning rather than credentials — sent to every applicant who passes a basic eligibility filter. The work-sample answers are then graded by humans against a rubric. The work is more expensive per applicant; the funnel converges materially faster on the candidates who can actually do the job.
The third pattern: founder-and-leader-led re-introduction of human review at the top of the funnel. The most striking version of this comes from a healthcare-system CHRO we spoke with in early April, who described re-staffing a team of seven resume reviewers — explicitly framed as a "humans first, then AI" funnel design — and reported a 23 percent improvement in early-stage offer-acceptance rate within ninety days. That number is one data point and we'd want to see it stabilise across multiple companies before treating it as benchmark, but the directional claim has been echoed across three other CHRO conversations.
The named CHROs willing to go on the record about this transition are still few. Lisa Buckingham, formerly of Lincoln Financial and now consulting across multiple Fortune 100 talent transformations, has been the most public voice. Diane Gherson, the former IBM CHRO, has written about the structural problem on her newsletter. Most of the conversations are still happening privately, at conferences, between peers.
The vendor-side response is bifurcating. The first-generation resume-AI vendors — the ones whose product is fundamentally a parsing-plus-ranking layer — are repositioning toward "candidate experience" and conversational tooling. The second wave — Mercor, Karat, and a small set of Verified Profile-style vendors — are positioning around the inverse claim: that the resume itself is the obsolete artefact, and that the future of screening is structured signal collected through video, code, or async work product. The Mercor pitch in particular has resonated in CHRO conversations: per a senior recruiter at one of the largest US technology employers, "the resume is dead. We just haven't told the candidates yet."
Forecast: by the end of 2026, fewer than half of the Fortune 500 will be running their professional-role resume funnels through the same first-generation AI tooling they were using at the start of 2024. The rollbacks will continue. What replaces them will not be a return to the pre-AI funnel — that ship has sailed — but a layered system of work-sample tasks, structured async exercises, and Verified Profile-style portable credentials. The CHROs who get this transition right in 2026 will own the hiring narrative for the rest of the decade. The ones who don't will be hiring AI-written resumes by AI-trained screeners, and wondering why the candidates they hire don't match the candidates they pictured.
For a structural look at how Anthropic compressed the senior-research funnel without leaning on resume AI, see How Anthropic restructured its talent stack. For the broader picture on what the senior-IC compensation reset is doing to the same funnel, see the OpenAI compensation reset briefing.