Perly Consulting │ Beck Eco

The State of Play

A living index of AI adoption across industries — where established practice meets the bleeding edge
UPDATED DAILY

The AI landscape doesn't move in one direction — it lurches. Some techniques leap from experiment to table stakes in a single quarter; others stall against regulatory walls, technical ceilings, or organisational inertia that no amount of hype can dislodge. Knowing which is which is the hard part. The State of Play cuts through the noise with a rigorously maintained index of AI techniques across every major business domain — classified by maturity, evidenced by real-world adoption, and updated daily so you always know where you stand relative to the field. Stop guessing. Start knowing.

The Daily Dispatch

A daily newsletter distilling the past two weeks of movement in a domain or two — delivered to your inbox while the index updates in the background.

AI Maturity by Domain

Each dot marks the weighted maturity of practices within a domain — hover for a brief summary, click for more detail

DOMAIN
BLEEDING EDGEESTABLISHED

HR policy Q&A chatbot

GOOD PRACTICE

TRAJECTORY

Stalled

AI chatbot that answers employee questions about HR policies, benefits, and procedures from organisational documentation. Includes policy RAG and benefit eligibility checking; distinct from enterprise search which serves general rather than HR-specific knowledge needs.

OVERVIEW

HR policy Q&A chatbots have crossed from promising experiment to proven capability. GA platforms from ServiceNow, Moveworks, and Leena AI now handle benefits, leave, compensation, and procedural queries at scale using retrieval-augmented generation against organisational documentation. The tooling works -- deployments routinely show 40-75% reductions in HR ticket volume, and Forrester has validated three-year ROI above 250%. The question facing most organisations is no longer whether the technology delivers, but whether their governance, knowledge bases, and change management are ready to support it. Regulatory complexity is accelerating that readiness gap: U.S. state chatbot disclosure laws and the EU AI Act's high-risk classification impose compliance obligations that demand architectural discipline before deployment. Organisations with mature governance are extracting real value; those without it face litigation exposure that outweighs any efficiency gain.

CURRENT LANDSCAPE

The vendor ecosystem has consolidated around platform incumbents. ServiceNow's March 2026 EmployeeWorks GA (integrating Moveworks' conversational AI) now handles 200M+ employees globally with validated production deployments: Siemens Healthineers (74,000 employees saving 5,000 hours monthly at 91% satisfaction), CVS Health (300,000 colleagues with 50% reduction in live agent chats), and City of Raleigh (98% of initial HR touchpoints routed intelligently). Moveworks achieved FedRAMP Moderate Authorization, opening federal and healthcare deployments with certified compliance infrastructure. Leena AI (400+ customers) released AOP Creator (GA) enabling automated workflow triggering from policy queries. The install base demonstrates production-scale capability: 350+ enterprises using Moveworks with measurable ROI (West Monroe achieved $1.4M annual cost reduction; Procore saved 351K quarterly productivity hours across HR, IT, and payroll functions).

Organisational adoption and governance maturity remain the binding constraints. SHRM's March 2026 conference data on 500+ HR leaders shows AI adoption nearly doubled (26% to 43% in one year), yet only 11% have embedded AI into daily workflows—two-thirds of HR professionals report their organisation hasn't proactively prepared employees to work alongside AI. A CHRO Association survey of 150 CHROs found 91% rank AI as top priority, yet 47% lack clear productivity measurement frameworks. Only 10% of organisations currently use AI for policy recommendations, and 59% of HR leaders report little or no AI deployment.

Regulatory and security risks have become the primary adoption barrier. California's SB 243 (effective January 2026) and companion state laws (New York, Maine, Utah, Nevada, Illinois) introduced private rights of action and disclosure mandates for AI chatbots. The EU AI Act classifies HR chatbots as potentially high-risk with an August 2026 compliance deadline and penalties up to €35 million or 7% of global turnover. March 2026 high-profile failures underscore the stakes: McKinsey's internal AI Q&A chatbot (serving 40,000+ employees, processing 500,000+ prompts monthly) was breached via SQL injection and unauthenticated APIs, exposing 46.5M internal messages. A People Central HR SaaS provider suffered a breach affecting 95,000 employee records through SQL injection vulnerabilities. These incidents demonstrate that governance maturity and security discipline—not product capability—now gatekeep adoption. Organisations pursuing HR chatbot deployments face acute compliance and litigation exposure without rigorous knowledge-base curation, auditability controls, human-in-the-loop validation, and documented bias testing.

TIER HISTORY

ResearchJan-2020 → Jan-2021
Bleeding EdgeJan-2021 → Jan-2023
Leading EdgeJan-2023 → Apr-2025
Good PracticeApr-2025 → present

EVIDENCE (102)

— Critical analysis debunking cost-reduction claims; real deployment examples (Klarna, Alibaba, Vodafone) with measured failure rates demonstrate that chatbot ROI requires three preconditions—absence signals implementation risk.

— HR-specific adoption metrics including concrete chatbot deployment rates and performance benchmarks; provides current evidence of organizational HR chatbot adoption and operational outcomes.

— Production HR chatbot (Eva) deployed on Azure handling multi-country policies with four retrieval strategies (HyDE, step-back, query rewrite, standard search), LLM reranking, and live agent handoff. Architectural best practices for HR policy Q&A at scale.

— Critical adoption barrier signal: large-scale survey evidence of severe employee AI tool resistance; directly demonstrates implementation challenge for HR policy Q&A chatbot user adoption and acceptance.

— SHRM survey of 1,908 HR professionals: 92% of CHROs expect further AI integration; 87% forecast greater HR process AI adoption in 2026. Large-sample research establishing executive adoption intentions for HR AI systems including chatbots.

— Critical finding: only 17% of HR AI implementations 'highly successful'; organizations following change management 2.6x more likely to succeed. Only 43% used change management; 57% of core HR systems have zero AI functionality.

— Capgemini-deployed Gen AI chatbot for Nortura (food producer) serves multilingual 24/7 HR support across diverse global workforce with policy/benefits Q&A while maintaining data security; demonstrates production deployment at international scale.

— BMJ Open study: 49.6% of medical chatbot responses problematic; hallucinations, poor references, and false confidence are standard. Empirical evidence of Q&A chatbot failure modes directly applicable to HR policy Q&A deployment risks.

HISTORY

  • 2021: VC funding ($30M+ for Leena AI, Paradox, Aisera) validated HR chatbot market; ServiceNow reported 80% L1 support reduction across departments including HR; peer research identified intrinsic employee motivation as key adoption driver; identified critical failure modes (intent misunderstanding, loops, handoff failures) as barriers to broader adoption.
  • 2022-H1: Moveworks launched Moveworks for HR (GA) with deployment at Solidigm (1,000+ employees); Kimberly-Clark's HR chatbot received 2.5x more questions than traditional channels; research identified specific design factors (bidirectionality, social cues) driving successful interactions; limitations in intent classification and emotion handling remained common, suggesting technology still immature despite growing vendor investment.
  • 2022-H2: Leena AI recognized as market leader (G2 analyst report); practitioner deployments in professional services demonstrated production viability but required narrow scoping and organizational change; market growth reports indicated sustained momentum; critical assessments documented enduring limitations (NLP brittleness, organizational resistance, unrealistic expectations) as barriers to mainstream adoption.
  • 2023-H1: Moveworks published Forrester ROI study (256% three-year ROI, $2.2M HR cost savings) confirming financial viability; ServiceNow expanded Virtual Agent with 18 pre-configured HR topics; academic research and practitioner case studies reinforced implementation criticality; market consolidated around specialist vendors and platform incumbents, with viable deployments limited to narrow, well-scoped use cases.
  • 2024-Q1: ServiceNow released HR Virtual Agent updates (Washington DC release); Leena AI maintained niche adoption (99 customers, 0.1% market share); Air Canada tribunal ruling established corporate liability for chatbot errors; NYC government chatbot audit revealed dangerous accuracy failures in legal/compliance domain, raising stakes for HR policy applications; compliance experts warned of million-dollar exposure from HR chatbot errors.
  • 2024-Q2: ServiceNow released Q2 2024 platform enhancements (Virtual Agent topic migration, Sensitivity Detection) advancing product maturity; Gartner survey showed 38% of HR leaders piloting/implementing GenAI with HR chatbots as top use case; however, significant adoption friction emerged—Paychex survey found 41% of employees prefer less AI and 71% uncomfortable with AI-led HR; U.S. Department of Labor issued field bulletin warning AI cannot substitute for human oversight in wage/hour compliance, establishing acute regulatory liability for HR chatbots handling policy decisions; Czech corporate survey showed 54% deploying or planning AI in HR but confirmed employee preference for human contact and EU AI Act high-risk classification.
  • 2024-Q3: IBM's AskHR case study revealed critical organizational barriers—initial CSAT collapse (-35) from poor change management, but full recovery to 94% query handling and 40% HR budget reduction after strategic redesign; Moveworks achieved Forrester Leadership across 19 criteria; ServiceNow achieved 37% faster case resolution through internal policy knowledge search; employee adoption sentiment improved (TriNet: 66% use AI for HR, 1 in 3 prefer AI to humans) but remained polarized (Paychex: 41% prefer less AI); HR leadership endorsement strengthened (Forrester: 75% deem AI-powered HR essential for future investment, 66% report satisfaction); peer research confirmed persistent technical limitations alongside functional value, confirming market bifurcation between successful narrow deployments and broader automation blocked by accuracy, compliance, and trust barriers.
  • 2024-Q4: ServiceNow released November 2024 enhancements (LLM Topics, multi-turn Q&A, Sensitivity Detection, multilingual support) advancing platform maturity; market adoption reached 70%+ of large enterprises handling 80,000+ HR queries monthly with 40% onboarding time savings, while HR professional adoption climbed to 94% despite governance gaps (40% lack policies); Leena AI confirmed scale (400+ customers, 70% self-service); legal analysis surfaced persistent risks (bias, discrimination, privacy, unreliable advice), emphasizing that functional capability had decoupled from organizational and regulatory readiness.
  • 2025-Q1: Vendor consolidation accelerated—ServiceNow announced $2.85B acquisition of Moveworks (March 2025), signaling market concentration around platform incumbents; real-world deployments expanded with Inspire for Solutions Development (450 employees, IBM watsonx), e2open (4,000+ employees, 75% HR question reduction), and EverBank (40% task reduction), confirming production viability across scales. However, critical implementation barriers emerged: independent reviews showed Moveworks at 6.4/10 satisfaction with only 81% recommendation rate; ScreenMeet analysis revealed Virtual Agent deflection rates below 15% vs. promised 50% due to "Done Gap" in knowledge bases. EU AI Act compliance deadline (August 2026) began reshaping governance requirements with €35M/7% turnover penalties for non-compliance, signaling regulatory maturity of the practice but raising adoption friction for cautious organizations.
  • 2025-Q2: BambooHR and e2open deployments demonstrated production viability at scale (30% and 75% ticket reductions respectively); ServiceNow Yokohama release added GA-level KPI tracking for KB-driven HR case resolution; vendor platform maturity continued advancing (Moveworks Agent Studio claimed 95% production success, Forrester analyst validation). However, adoption expansion stalled at organizational level—Vlerick survey showed 59% of HR leaders with little/no AI use, only 25% using chatbots; HR Acuity benchmark showed 44% of organizations report no AI use in employee relations and <10% leverage AI for policy recommendations. Employee sentiment remained polarized (66% use AI per TriNet, but 41% prefer less AI per Paychex). Leadership recognized governance gaps (92% of HR leaders per G-P report) ahead of August 2026 EU AI Act compliance deadline. The inflection shifted from platform capability to knowledge-base completeness and organizational readiness barriers.
  • 2025-Q3: Moveworks reached 5 million platform users and 90% company-wide rollout; Globe Telecom deployment (Leena AI) achieved 75% HR ticket self-service resolution. Product enhancements matured (multi-turn Q&A, multilingual support); however, organizational adoption stalled—HR Acuity showed <10% of organizations leverage AI for policy recommendations. Critical risks surfaced: employment law analysis warned of hallucination-induced compliance liability (avg. $40K+ per claim); security researchers documented prompt injection vulnerabilities enabling sensitive data extraction. Regulatory pressure intensified with August 2026 EU AI Act compliance deadline reshaping investment priorities. The inflection clarified: the practice consolidated toward narrow-scope, high-confidence deployments (standardized benefits, routine policy questions) rather than mainstream adoption, with organizational readiness and knowledge-base completeness as primary adoption barriers.
  • 2025-Q4: Regulatory compliance became the primary inflection point. New state laws (California SB 243 effective Jan 2026, New York, Maine, Utah, Nevada, Illinois) introduced private rights of action and disclosure requirements for AI chatbots, raising litigation exposure. Analyst reports (Capstone DC) documented acute risks from Mobley v. Workday age discrimination case and EU AI Act high-risk classification (€35M or 7% turnover penalties). Ecosystem partnerships matured—Interact and Leena AI integrated agentic AI capabilities enabling policy Q&A with automatic action triggering. Real-world deployments continued: Moveworks secured multinational infrastructure company with 28,000+ employees across 53 countries. HarmonyHR analysis confirmed that successful implementations require process discipline first, with AI agents positioned as complementary to human oversight. Critical assessments (inFeedo) documented continued limitations: emotional context gaps, complex issue handling barriers, and need for human escalation. By end-Q4 2025, regulatory maturation and litigation risk had overtaken product capability as the primary adoption barrier; organizations pursued only highest-confidence narrow-scope deployments with rigorous governance and change management.
  • 2026-Jan: Regulatory framework stabilized with California SB 243 and companion state laws (New York, Maine, Utah, Nevada, Illinois) now in effect, establishing disclosure requirements and private right of action for AI chatbots. Vendor platforms advanced (Moveworks FedRAMP-authorized across regulated sectors including healthcare and financial services; ServiceNow HRSD with sentiment analysis and KB-driven metrics). Workplace AI adoption broadened (Gallup: 12% daily, ~6/10 of AI users rely on chatbots for admin tasks), though HR-specific adoption remained cautious. Governance frameworks matured (Pesync compliance matrix, ercel guidance for EU AI Act high-risk classification with August 2026 deadline). Security risks crystallized through case studies (Eurostar prompt injection); governance failure and regulatory liability emerged as equal barriers to product capability. Deployments remained narrow-scope and high-confidence with mandatory human-in-loop controls.
  • 2026-Feb: Regulatory framework operationalized with state chatbot laws now active (California SB 243 private right of action, New York/Maine/Utah/Nevada/Illinois disclosure mandates). Vendor consolidation advanced: Moveworks achieved FedRAMP Moderate Authorization for federal/healthcare deployments; ServiceNow launched EmployeeWorks integrating Moveworks' conversational AI; Leena AI released AOP Creator (GA) enabling automated workflow triggering from policy queries. Real-world accuracy failures reinforced adoption caution: NYC's MyCity chatbot shut down (Feb 4) for providing illegal advice (e.g., withholding tips), demonstrating deployment risks directly applicable to HR policy automation. Industry guidance (AI Tribune) positioned HR policy Q&A as viable "good first project" but noted 80%+ of companies gained no measurable productivity from AI yet. Organizational readiness remained weak: HR-specific adoption continued stalling with <10% leveraging AI for policy recommendations. The decisive inflection clarified: regulatory compliance and organizational readiness (not product capability) now gatekeep adoption; deployment remained confined to large enterprises with sophisticated governance disciplines.
  • 2026-Mar: ServiceNow EmployeeWorks reached general availability (March 2026) with documented customer deployments achieving measurable HR impact: Siemens Healthineers (74K employees, 5K hours/month saved, 91% satisfaction), CVS Health (300K colleagues, 50% chat reduction), City of Raleigh (98% initial touchpoint resolution). Moveworks reached 350+ total customer base with specific HR use case evidence; FedRAMP certification enabled healthcare/federal sector deployments. However, critical governance and security failures emerged: McKinsey's internal AI Q&A chatbot (40K employees, 500K+ monthly prompts) breached via SQL injection/unauthenticated APIs (46.5M messages exposed); People Central HR SaaS provider leaked 95K employee records through SQL injection vulnerabilities (salaries, bank accounts, emergency contacts exposed). These high-profile breaches demonstrated that governance maturity and security discipline—not product capability—remain the binding adoption constraint. SHRM conference data (500+ HR leaders) confirmed adoption bifurcation: AI usage doubled (26% to 43%) but embedding into workflows stalled (11%). CHRO Association survey (150 CHROs) revealed 91% prioritize AI but 47% lack productivity measurement frameworks. EU AI Act chatbot penalties (€35M or 7% global turnover, effective August 2026) sharpened the compliance urgency for organizations deploying policy Q&A systems. The inflection solidified: deployment success depends entirely on organizational and regulatory readiness (knowledge-base curation, auditability, bias testing, human-in-the-loop controls) rather than platform feature parity.
  • 2026-Apr: Enterprise-scale deployment evidence expanded: Johnson Controls' agentic HR assistant (100K+ global employees) achieved 30-40% call volume reduction on routine policy and onboarding queries, while an NHS Foundation Trust (5,000+ employees) completed a structured governance-first deployment using a 15-person HR/OD validation working group. Hallucination risk received renewed scrutiny—the Halluhard benchmark found Claude Opus 4.5 with web search hallucinated in ~33% of multi-turn cases, and Klarna's reversal of its 700-person HR/CS AI replacement highlighted agentic failures in complex handling. The EU AI Act August 2026 enforcement deadline (€35M or 7% turnover penalties) and SHRM's finding that 80% of HR professionals use genAI daily but governance and change management lag behind in high-judgment tasks confirmed that regulatory and organisational readiness remain the binding deployment constraints.
  • 2026-May: New production case studies confirmed deployment viability at scale—Coretus achieving 85% resolution across 14 countries, a Microsoft Eva multi-strategy RAG deployment with live agent handoff on Azure, and a Capgemini Nortura multilingual deployment. Critical cost-benefit analysis (Crisp, examining Klarna, Alibaba, and Vodafone deployments) debunked vendor ROI claims by documenting real-world failure rates and establishing that three preconditions must be met before chatbot ROI materialises; SHRM's survey of 1,908 HR professionals confirmed organisational readiness gaps persist alongside growing deployment scale.