Where AI Stands in People & Talent
People & Talent is the most thoroughly instrumented domain in the enterprise, and the most thoroughly disappointed. Across hiring, development, engagement, compensation, and workforce planning, the technology question has been answered: vendor platforms from Workday, SAP, Phenom, Culture Amp, HireVue, Eightfold, Pave, and a long tail of specialists ship general-availability products with documented returns — 300% three-year ROI on skills platforms, 75% time-to-hire reductions, six-figure attrition savings. Adoption is near-universal; roughly 87% of organisations use AI in recruiting, 98.8% of the Fortune 500 run applicant tracking systems, and SHRM puts daily HR AI use at 80%. By every measure of deployment, AI has won in HR. By the only measure that matters to a CFO, it has not. The signature finding of this cycle, repeated across practice after practice, is a single convergent statistic: a ManpowerGroup/Everest Group survey of C-suite and CHRO leaders found 90% have deployed AI in recruiting yet fewer than 5% report transformational outcomes. Gartner says 88% of HR leaders see no significant business value. The gap between what these systems can do and what organisations get from them is the defining structural fact of the domain.
This is why momentum has effectively stalled. Of the seventeen practices we track, every one now sits at "stalled" or "declining" — and as of this scan, the last advancing practice (adaptive assessment) has joined them. That uniformity is not coincidence; it reflects a shared binding constraint. The technology is no longer the bottleneck. Organisational readiness, data governance, manager incentives, employee trust, and a hardening thicket of regulation are. The most telling evidence comes from the practices furthest along: skills mapping has a proven 300% ROI yet 46% of managers actively resist the internal moves it enables; workforce planning has achieved full agentic parity across Workday, SAP, and Anaplan yet only 5% of enterprises report substantial AI returns, and Gartner found that headcount cuts driven by AI deliver identical results whether or not the firm captures value — proving the layoffs are not the payoff. Even the boomerang is now documented: 29% of companies that cut AI-driven headcount (Klarna, IBM, Duolingo among them) rehired for the same roles within six months.
What distinguishes People & Talent from other AI domains is that its failures are legal as well as commercial. Hiring is where algorithmic bias becomes a federal lawsuit. The Mobley v. Workday collective action now covers some 1.1 billion rejected applications, and a June 2026 federal ruling denied Workday's motion to dismiss — establishing that an AI vendor can be a liable "agent," not a neutral tool. A Stanford-led study of four million applications across 156 employers found 10.62% of positions showed adverse impact on Black applicants, with an "algorithmic blackball" effect where one vendor's rejection follows a candidate across employers. No other enterprise AI domain carries this combination of mass deployment, documented discrimination, and crystallising vendor liability. The pattern recurs in adjacent practices: candidate assessment tools now face EU AI Act high-risk classification, the UK ICO has begun writing compliance letters to employers, and HR policy chatbots have been ruled an extension of the company itself, with Germany’s highest HR court holding employers strictly liable for whatever the bot says. It is the reason a mature, ROI-positive technology stack remains, in our assessment, structurally stuck — and why the smartest operators are spending less on new capability and more on the governance, data foundations, and change management that turn capability into outcomes.
What's New, 2026-06-16 to 2026-06-30
This was a fortnight of consolidation rather than rupture, with two genuine shifts. First, the domain's maturity arc flattened completely: adaptive assessment & knowledge testing — the last practice we rated as advancing — moved to stalled, as UK employer surveys (67% worried candidates misrepresent ability using AI) and fresh data-governance litigation (two class actions against i-Ready maker Curriculum Associates over student data) undercut the validity premise faster than the impressive market growth could offset it. Every People & Talent practice is now stalled or declining. Second, and more consequentially, the regulatory picture fractured. Colorado repealed its AI Act (SB 26-189) just weeks before its June 30 effective date, stripping out the bias-audit and impact-assessment mandates that multiple practices had been citing as a hardening compliance force — even as California, Illinois, and NYC enforcement tightened. In parallel, the EU deferred its high-risk HR obligations (recruitment screening, performance management, promotion) to late 2027/2028, leaving only the Article 50 transparency deadline (disclose that users are talking to AI) live for August 2, 2026.
The rest of the cycle deepened existing themes rather than changing them. The Mobley v. Workday discovery ruling extended vendor-liability doctrine; the UK ICO's "Recruitment Rewired" audit wrote compliance letters to 16 of 30-plus employers and called out firms that label automated decisions "decision support" to dodge oversight; Germany's OLG Hamm ruled companies strictly liable for HR chatbot hallucinations regardless of training-data quality. On the demand side, India's IT sector active openings hit a 28-month low with entry-level postings down 44% year-on-year — the clearest signal yet that AI is structurally compressing junior hiring — while Wipro and TCS announced plans to train tens of thousands of staff on Claude, signalling that the workforce-redesign conversation is finally moving from headcount cuts to capability building. Stability here is itself the signal: the domain is settling into a durable plateau where capability keeps improving and organisational outcomes do not.
Key Tensions
The 5% problem: deployment without value. The most-cited number across this domain is the gap between adoption and return. ManpowerGroup/Everest Group found 90% of leaders deploy AI in recruiting but fewer than 5% report transformational outcomes; Gartner puts the figure at 88% of HR leaders seeing no significant business value. SHRM data shows cost-per-hire and time-to-hire have both increased over the three-year AI adoption period. The binding constraint is not capability but workflow redesign — 95% of organisations have not redesigned jobs despite 84% reporting all roles are changing — and most organisations chase quick wins instead.
Litigation has reclassified the vendors. Hiring AI is the only enterprise software category where the tool maker is being sued as the employer's agent. Mobley v. Workday (≈1.1 billion applications, ADEA collective certification, June 2026 dismissal denied) and Kistler v. Eightfold (FCRA exposure for undisclosed candidate scoring, customers including Microsoft and Morgan Stanley) have shifted liability onto vendors. A peer-reviewed FAccT study found recruiters mirror biased AI recommendations 90% of the time — meaning the "human-in-the-loop" safeguard regulators rely on does not actually work, making governance failure the core compliance risk.
The regulatory ground is shifting under both feet at once. This is no longer a story of steadily hardening rules. Colorado repealed its landmark AI Act weeks before it took effect; the EU pushed its high-risk HR obligations out to 2027/2028. Yet California, Illinois, and NYC are tightening, the UK ICO is auditing and writing compliance letters, and Germany's highest HR court has imposed strict liability for chatbot errors. The result is a fragmenting patchwork where 57% of HR professionals are unaware of the rules that apply to them — making compliance a moving target rather than a checklist.
Trust is collapsing on both sides of the desk. Candidate confidence in AI evaluation sits at 26%; 66% of US adults say they will not apply to employers using AI hiring; offer-acceptance rates in AI-assessed processes fell from 74% in 2023 to 51% in 2026. Inside the building, ManpowerGroup recorded the steepest annual drop in worker AI confidence on record (down 18%) even as usage rose, and a Prudential study found a 27-point gap between employer enthusiasm (78%) and employee confidence (51%). The EU AI Act's August 2026 emotion-recognition ban will eliminate mood detection and voice-based sentiment analysis across European operations entirely.
The signal that AI was meant to read is being corrupted by AI. Hiring has become an arms race. Roughly 40% of resumes are now AI-drafted, 92% of hiring leaders say AI-generated applications are commonplace, and only 33% remain confident resumes reflect real skills. Candidate-side fraud is escalating — 41% admit using prompt injection to bypass screening, 20–45% of remote tech-role applicants are fraudulent, and Gartner projects one in four candidate profiles will be fake by 2028. The tools deployed to find signal are simultaneously generating the noise that buries it.
DEI analytics is the cautionary tale the rest of the domain may follow. Here the tooling peaked precisely as the willingness to use it collapsed. Fortune 500 public DEI disclosure fell 65% year-on-year; S&P 500 disclosure in 10-Ks dropped from 97% to 55%; IBM settled a DOJ False Claims Act case for $17.1M over diversity-linked bonuses and demographic targeting. Measurement infrastructure has not disappeared — 92% of the Fortune 100 still run active DEI departments — it has gone behind closed doors, with a larger cohort migrating analytics into confidential internal channels invisible to regulators. It is the clearest example in the domain of the central truth: the constraint on HR AI is never whether the technology works, but whether organisations are willing and able to act on what it tells them.
Top 10 Evidence Items
90% of Companies Use AI in Hiring. Fewer Than 5% Are Seeing It Work (adoption-metric) — The single statistic that defines the domain's stall: mass deployment coexisting with near-zero transformational return is not a technology problem but a governance and change-management one, and this ManpowerGroup/Everest Group survey of 80 C-suite and CHRO leaders makes that explicit. https://investor.manpowergroup.com/news-releases/news-release-details/90-companies-use-ai-hiring-fewer-5-are-seeing-it-work
California Federal Court Denies Workday's Motion to Dismiss in Mobley v. Workday (news-coverage) — The June 2026 ruling establishing that an AI vendor is a liable "agent" under FEHA (not a neutral tool) is the legal event that reclassifies the entire hiring-AI vendor category, giving the 1.1-billion-application class action its most dangerous teeth yet. https://blogs.duanemorris.com/classactiondefense/2026/06/24/california-federal-court-grants-in-part-and-denies-in-part-workdays-motion-to-dismiss-in-mobley-v-workday/
AI-Powered Recruitment Tools Are More Likely to Lead to Racism and Systemic Exclusion (research-paper) — Stanford HAI's empirical study of 4 million applications across 156 employers finding 10.6% of positions with adverse impact on Black applicants, and an "algorithmic blackball" that follows candidates across employers, is the peer-reviewed foundation beneath every discrimination claim in the domain. https://gigazine.net/gsc_news/en/20260625-ai-hiring-tools-racial-bias/
Resume Screening, Fast and Slow: (Biased) AI Recommendations' Influence on Human Decision Making (research-paper) — The FAccT 2026 paper showing recruiters mirror biased AI recommendations 90% of the time demolishes the "human-in-the-loop" defence that regulators and vendors routinely cite; governance failure is the compliance risk, not the absence of a human reviewer. https://arxiv.org/abs/2606.22213
80% of Companies Cut Jobs for AI in 2026. The Returns Never Showed Up. (adoption-metric) — Gartner's survey of 350 large enterprises decoupling headcount cuts from value creation is the workforce-planning analogue of the recruiting 5% finding: AI-driven layoffs are being executed on speculation, not demonstrated replacement, with high-ROI firms doing the opposite. https://www.linkedin.com/pulse/80-companies-cut-jobs-ai-2026-returns-never-showed-up-rahul-soin-3jswe
Recruitment Is Broken. Automation and Algorithms Can't Fix It. (opinion) — SHRM's 2025 benchmarking data showing cost-per-hire and time-to-hire both increased during the period of heaviest AI adoption is the most direct refutation of the efficiency promise that drove HR AI spending; the arms-race dynamic between AI-optimised resumes and AI screening bots is the mechanism. https://www.shrm.org/topics-tools/news/hr-trends/recruitment-is-broken
If an AI Chatbot Misleads You, Who Is to Blame? (opinion) — Legal expert analysis establishing that AI agents are legal agents of the deploying organisation, with Air Canada and the OLG Hamm Germany ruling as precedents, closes the gap between "the bot said it" and corporate liability — critical context for every HR policy chatbot deployment. https://www.theguardian.com/commentisfree/2026/jun/24/ai-errors-companies-responsibility
EU AI Act Article 50: What Internal Chatbots Must Display by August 2, 2026 (industry-report) — With the high-risk HR obligations deferred to 2027/2028, Article 50 transparency disclosure is the one live EU deadline in this scan cycle; this compliance guide makes concrete what deployers of HR chatbots must do in weeks, not years. https://www.k-ai.ai/en/news/eu-ai-act-article-50-internal-chatbot-disclosure-deployer-august-2026/
Lawsuit Challenges i-Ready's Student Data Practices (news-coverage) — The class action against Curriculum Associates covering 112K+ students for improper data collection and third-party sharing is the specific event that moved adaptive assessment from advancing to stalled this cycle, illustrating how data-governance litigation undercuts validity claims faster than market growth can compensate. https://next-13-sage.vercel.app/axios/0877a838-b7a7-4d12-b762-73dc0389c1c5
Understanding AI Compliance Obligations Following Colorado's Scaled-Back Legislation (industry-report) — Colorado repealing its landmark AI Act weeks before its June 30 effective date, while Connecticut, Illinois, and California tighten, is the definitive illustration of the regulatory patchwork tension: 57% of HR professionals unaware of applicable rules are navigating a landscape that is simultaneously loosening and tightening in different jurisdictions. https://staffinghub.com/employment-laws-and-regulations/ai-hiring-compliance-staffing-2026/