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

L&D adaptive assessment & knowledge testing

GOOD PRACTICE

TRAJECTORY

Advancing

AI that administers adaptive assessments that adjust difficulty based on responses to efficiently measure competency. Includes item response theory and computerised adaptive testing; distinct from skills assessment in education which maps against external frameworks rather than internal competencies.

OVERVIEW

Adaptive assessment has graduated from methodological novelty to proven operational tool. Computerized adaptive testing (CAT) adjusts question difficulty in real time based on responses, measuring competency in a fraction of the items a fixed-length test requires -- reductions of 40-88% are documented across domains. The practice now operates at scale in corporate talent evaluation, government competency measurement, healthcare outcomes, and K-12 education, supported by a mature vendor ecosystem, GA tooling, and analyst-recognised market growth. Recent algorithmic advances (Bayesian active inference frameworks achieving 30-40% trial reduction) and government-funded research infrastructure (e.g. US IES $3.8M Adult Skills Assessment Project with 20,000-learner validation sample) signal deepening methodological maturity and institutional commitment. The question facing L&D teams is no longer whether adaptive assessment works but how to operationalize it effectively. That operationalization involves managing competing demands: balancing system usability and content relevance (ranked highest in global adoption surveys), ensuring assessment integrity against generative AI circumvention (requiring item design validation and security protocols), and aligning tools with institutional workflows and practitioner capacity. Generative AI remains an active threat—with 88% of undergraduates now using LLMs for assessed work—forcing organisations to rethink assessment design toward scenario-based and contextual formats that maintain validity. These challenges are organisational and design-focused rather than methodological. Adoption is a matter of implementation discipline and assessment design rigor, not proof of concept.

CURRENT LANDSCAPE

The vendor ecosystem is mature and commercially active. SHL dominates corporate talent assessment, reporting 4x candidate throughput gains with AI-assisted adaptive scoring and deploying its Talent Mobility solution -- which evaluates 96 behavioural skills in 15 minutes -- at General Mills, the Royal Navy (350+ leaders), and global financial services firms (10,000+ participants). The Adecco Group runs SHL adaptive assessments across 9 brands in 7 languages. In L&D, STADA pharmaceutical cut SAP training time by 40% using Area9 Lyceum's adaptive platform while achieving near-perfect competency outcomes. Market analysts project the adaptive learning market reaching USD 11.81 billion by 2035 at a 16.81% CAGR; corporate adoption surveys show 89% of L&D leaders shifting to AI-first platforms and 75% finding adaptive learning effective for engagement. Deployment also extends to adult learner populations: University of Massachusetts ASAP partnership (serving 10,000+ adults) demonstrates adaptive assessment at scale in adult literacy/numeracy with ML-driven question sequencing.

Government and education deployments reinforce the breadth. The U.S. IES PIAAC 2023 used adaptive testing for national adult competency evaluation, and the IES is funding a $3.8M 5-year research project (Adult Skills Assessment Project) developing reusable adaptive assessment modules for adult education and community colleges with 20,000-learner validation. Wales runs statutory adaptive assessments for Years 2-9 literacy and numeracy at national scale; India's IIT Council is piloting adaptive testing for the JEE Advanced entrance exam. In healthcare, a hand surgery CAT study (268 patients) replicated full-length patient-reported outcome measures with just two questions and 95%+ correlation, while ML-based CAT now achieves 94-96% accuracy across five languages with roughly 10 items.

Institutional budget reallocation signals accelerating adoption momentum. Universities now allocate 18-24% of IT budgets to AI assessment tools (up from ~9% two years prior), with adaptive assessment identified as the primary procurement driver ahead of content delivery. Pearson's AI-powered assessment engine serves over 4 million active learners in higher education as of Q1 2026, with institutions reporting 15-22% improvement in student outcome prediction accuracy. Market analysis confirms growth: the global adaptive learning software market stands at $4.5 billion (2024) and is projected to reach $15.0 billion by 2033 at 15% CAGR; the AI adaptive learning platform market is valued at $10.6 billion (2025) with corporate training and L&D representing the largest segment at 38.4% of revenue, projected to reach $83.2 billion by 2034 at 26.1% CAGR. Patent acceleration further signals ecosystem maturity: 22+ filings in adaptive assessment technology patents during 2024-2026, compared to only ~5 in the foundational phase, with India emerging as the dominant innovation jurisdiction. Skillsoft's platform metrics reveal explosive enterprise adoption of AI skills validation infrastructure, with 994% year-over-year growth in AI-related skills validation completions.

Implementation and design barriers remain primary adoption constraints. Practitioners flag design trade-offs -- balancing adaptivity with standardisation for fairness, ensuring content validity, managing privacy and bias risks. Institution-level surveys identify system usability and content relevance as top adoption criteria; technical support emerges as the most critical institutional support factor. Assessment integrity under generative AI pressure requires active design validation: recent peer-reviewed research proposes DIF (Differential Item Functioning) methodology to identify items vulnerable to AI misuse, and critical analysis surfaces design flaws in AI-generated assessment items, necessitating expert review protocols. Standardisation gaps persist in specialised clinical domains, where heterogeneous methods and limited cross-validation slow adoption. Critical implementation research documents that 95% of AI pilot implementations fail to reach production, with the root cause identified as organizational factors (70%) rather than technology (10%) or data quality (20%); success requires explicit outcome accountability, workflow embedding, and change management disciplines. Market perspective remains balanced: adaptive learning excels for structured knowledge domains (compliance, product training, technical certifications) but struggles with interpersonal and emotional skills requiring human feedback, suggesting continued segmentation in adoption patterns.

TIER HISTORY

ResearchJan-2018 → Jan-2018
Bleeding EdgeJan-2018 → Jan-2022
Leading EdgeJan-2022 → Jan-2026
Good PracticeJan-2026 → present

EVIDENCE (123)

— Institutional adoption analysis: AI assessment tools now drive 18–24% of university IT budgets (up from ~9% two years prior); adaptive assessment identified as primary procurement driver; Pearson's engine serves 4M+ active learners; institutions report 15–22% improvement in student outcome prediction accuracy.

Deep Computerized Adaptive TestingResearch Papers

— Peer-reviewed research in Psychometrika on deep learning CAT: achieves high-precision ability estimation (posterior SD=0.4) with average 11.2 items vs. traditional fixed-length tests, demonstrating technical advancement in adaptive testing efficiency.

— IP landscape analysis documenting innovation acceleration: 22+ patent filings in 2024–2026 vs. ~5 in foundational phase; India dominates with 60% of records; identifies five core technology pillars including adaptive assessment engines and ML-driven item generation.

— Critical analysis citing MIT, BCG, McKinsey, IDC research on AI implementation failure: 95% of organizations get zero measurable return; 88% pilot failure rate; BCG's 70/20/10 rule emphasizes organizational change (70%) over technology; identifies success factors: measurable outcomes, workflow embedding, change management, outcome accountability.

— Market research showing ecosystem maturity: $4.5B market (2024) projected to reach $15.0B by 2033 at 15% CAGR; applications across K-12, higher education, corporate training; named vendor ecosystem (SAS, D2L, McGraw-Hill, Wiley, Docebo); regulatory constraints emerging (GDPR, CCPA).

— Cambridge's Adaptive Learn platform (K-8 curriculum) launched as product GA with diagnostic gap identification, progression tracking, and real-time actionable insights for educators; major educational publisher committing to adaptive assessment as core offering.

— 2026 Best Formative Assessment Award finalists: Singapore University AdLeS® adaptive learning system, PARAKH K-12 holistic progress integration, JUZ40 Kazakhstan UNT exam prep with adaptive checkpoints. Named institutional productions across higher ed, K-12, exam preparation.

— Survey of 382 HR professionals: 94% use assessments, 51% AI-enabled; only 22% 'very confident' ethical use; 'shadow AI' governance gap reveals one-third operate AI systems they cannot audit; rising candidate cheating and fairness concerns critical adoption friction.

HISTORY

  • 2018: Adaptive assessment moves from academic research to early commercial deployment. Methodological advances confirm CAT efficiency (40-99% item reduction with maintained accuracy). Corporate and healthcare deployments confirmed (Bombardier, medical outcome measurement). Market attention grows with significant vendor investment ($30M+ for Area9) and Gates Foundation support. Adoption barriers identified: technological, pedagogical, and organizational challenges limit institution-level take-up despite positive individual outcomes.
  • 2019: Market consolidation accelerates as SHL acquires Aspiring Minds, integrating AI-powered assessment into mainstream talent workflows. NEJM Knowledge+ platform validated with gold-standard award for physician exam preparation impact. Academic research expands CAT application across psychiatry, behavioral health, and psychology domains. K-12 adoption gains: multiple US states fully transition to CAT for statewide assessments. Underlying adoption barriers shift from validity concerns to integration complexity and organizational readiness.
  • 2020: Production-scale CAT deployments confirmed in healthcare (patient-reported outcomes at 19,523+ patient scale with maintained accuracy). Methodological advances in cognitive diagnostic testing improve classification algorithms. Tooling maturity increases with open-source Concerto platform enabling broader non-expert development of adaptive assessments. Adoption remains uneven across institutions despite growing evidence base and vendor consolidation.
  • 2021: Continued methodological innovation in cognitive diagnostic CAT with new item selection algorithms addressing practical constraints. Open-source tooling maturity deepens with EduCAT library providing integrated methods (IRT, machine learning) for CAT development. Academic and implementation research solidifies CAT efficacy across healthcare and educational contexts, though adoption barriers persist in organizational deployment.
  • 2022-H1: Deployments expand across behavioral health and corporate recruitment. CAT-SS validated for military veteran suicide risk screening (305 subjects, 11-item median, 107-second administration). SHL deployed adaptive assessments at 75,000-person BPO scaling to 7,000 monthly candidate evaluations with 32% CSAT improvement. Methodological research advances CAT application in psychological measurement and online testing frameworks. Vendor ecosystem consolidation and partnerships (Area9/NEJM) signal sustained market maturity despite persistent adoption barriers in organizational integration.
  • 2022-H2: Methodological research expands CAT application into special education and clinical outcome measurement. Simulation studies confirm CAT efficiency in inclusive settings; clinical deployments demonstrate burden reduction (patient surveys from 11 to 2 questions). Within-testlet adaptation algorithms improve classification accuracy. Vendor ecosystem matures with structured certification programs for adaptive content development. Critical counterbalance: SHL (leading vendor) publishes analysis documenting that validity claims in talent assessment are often overestimated, raising caution about predictive accuracy despite widespread deployment.
  • 2023-H1: Government-scale deployment confirmed: U.S. PIAAC 2023 uses adaptive assessment for national adult competency evaluation. Corporate talent assessment expands globally: Adecco Group (world's #2 HR provider) deploys SHL adaptive assessments across 9 brands in 7 languages. Healthcare adoption deepens in rehabilitation settings with PROMIS CAT feasibility validation. K-12 state-level deployment continues with WIDA ACCESS multistage adaptive testing for English language learners. Research advances in neurocognitive assessment methodologies. New adoption barrier emerges: ChatGPT and generative AI tools threaten assessment validity in talent testing; SHL documents AI's ability to artificially inflate scores, requiring organizations to rethink assessment design and security.
  • 2023-H2: Continued expansion of production deployments. Heineken implements SHL adaptive assessment for graduate recruitment (18,000 candidates, 175% ROI). Healthcare research extends CAT to medical professional certification (China national exam with 121-item bank). K-12 user acceptance validated in Japanese schools; adoption barriers in state testing persist due to cost and comparability concerns. Governance frameworks emerge: SHL documents AI risks including model decay and algorithmic bias in adaptive hiring assessments. By year-end, CAT is operationalized at enterprise scale but faces new challenges from generative AI tools circumventing assessment validity.
  • 2024-Q1: Methodological maturity deepens through research surveys bridging psychometric and machine learning perspectives. Area9 Lyceum demonstrates sustained impact in higher education (20-point retention, 13-point pass-rate gains across 30M+ students). Special education equity research validates CAT for students with special needs; shows reduced assessment length and bias with maintained accuracy. K-12 state-level adoption continues (California, Hawaii, Oregon, Arkansas, Utah) though organizational barriers (cost, data interpretation) persist. Simulation studies refine optimal CAT configurations (30-item fixed tests, 0.35 standard error). New challenge documented: K-12 implementation gaps remain despite methodological proof; adoption constrained by institutional cost and comparability concerns rather than validity.
  • 2024-Q2: Market maturity signals deepen: analyst projections forecast USD 1.88B adaptive learning market by 2028 (24.04% CAGR); corporate adoption surveys show 31% of L&D teams using AI for personalization (39% in US). Talent assessment deployment scales: SHL reports 4x candidate throughput gains in technology hiring with automated AI-assisted scoring. Methodological innovation advances: cognitive design systems reduce item calibration costs. Critical implementation evidence emerges: K-12 teacher study finds tools improve engagement and efficiency but reveal training gaps, data reliability concerns, and misalignment with grading systems—exposing friction beneath market growth. Adoption barriers shift from validity proof to organizational integration and practitioner capacity.
  • 2024-Q3: International and governmental adoption accelerates: OECD PISA programme operationalizes Highly Adaptive Testing (HAT) methodology for global educational assessment; Wales deploys adaptive assessments nationally serving 120,000+ students (5M assessments to date); medical education validates CAT in 1,432-student crossover study. Healthcare applications mature: COPD patient-reported outcomes CAT achieves 59-74% item reduction with maintained reliability in clinical settings. Corporate L&D adoption deepens: survey of 200 UK L&D leaders shows 75% finding adaptive learning effective for engagement and 63% for retention. Instructor-level deployment evidence confirms real-world impact: higher education instructors report exam scores rising from 68% to 90% with adaptive quizzing. The practice demonstrates operationalization across government, corporate, higher education, and healthcare sectors with sustained momentum despite implementation complexity.
  • 2024-Q4: Technical innovation and product maturity advance CAT ecosystem. Machine learning frameworks (BanditCAT, AutoIRT) deployed in high-stakes testing contexts (Duolingo English Test) for item calibration and adaptive administration. Product-level maturity signals expand: ACER launches curriculum-agnostic PAIS Adaptive for international schools; Future Centre delivers language CAT with 80% test time reduction across 20+ languages. Deployment evidence from organizational pilots: Westnetz (energy sector) achieves 100% competency on safety training with 65% sub-45-minute completion via Area9. Assessment equity research deepens: SHL documents practice test effects (58% deductive reasoning, 2x numerical reasoning gains) with no differential racial impact, advancing inclusive assessment evidence base. By year-end, adaptive assessment is operationalized globally across talent selection, educational assessment, healthcare outcomes, and specialized language testing; technical and organizational maturity are established; challenges remain in maintaining assessment integrity amid generative AI tools and ensuring practitioner capacity in K-12 and enterprise L&D contexts.
  • 2025-Q1: Corporate L&D deployment expands internationally with named implementations; healthcare CAT methodologies advance with clinical stopping rule optimization. STADA pharmaceutical company deploys Area9 adaptive SAP training platform, achieving 40% time reduction with near-perfect competency outcomes. Healthcare CAT research refines stopping rules (40-88% test reduction depending on method) balancing reliability and patient burden in clinical settings. Vendor investment in equity deepens: SHL launches 2025 Inclusive Assessment Research Program annual report documenting expanded neurodiversity and disability research. Methodological caution emerges: systematic review of substance use disorder CATs identifies heterogeneity in stopping rules and limited validation, highlighting standardization gaps across specialized clinical domains. Adoption signals expand geographically with positive receptivity to CAT among higher education leaders in developing regions (South Africa), though implementations remain concentrated in developed markets. The practice consolidates operational maturity with growing methodological sophistication, but faces persistent challenges in standardization across clinical domains and maintaining assessment integrity against AI-assisted test circumvention.
  • 2025-Q2: Talent mobility and L&D deployment momentum accelerates. SHL launches Talent Mobility solution integrating adaptive behavioral skills assessment (96 skills in 15 minutes) with predictive redeployment analytics; Williams Energy deploys SHL adaptive assessment for graduate rotational employee development. Healthcare CAT validation deepens: hand surgery study (268 patients) confirms CAT achieves 95%+ correlation with full-length PROM using median 2 questions vs. 10, advancing clinical adoption. K-12 adaptive assessment expands: Bellevue School District adopts i-Ready adaptive platform for K-8 math and literacy (including Spanish dual-language). Early childhood applications broaden: EuleApp© CAT validated for German kindergarten early literacy assessment (307 children). Implementation barriers surface: critical analysis documents persistent challenges in adaptive design trade-offs (model selection, content validity, standardization vs. adaptivity, ethical risks). The practice deepens operational maturity with expanded deployment examples and product innovation, but implementation complexity and design trade-offs remain primary adoption friction points alongside methodological standardization gaps in specialized clinical domains.
  • 2025-Q3: Methodological innovation addresses question selection bias with ACM CIKM 2025 conference research proposing debiasing framework for CAT algorithms. Corporate L&D adoption momentum confirmed: 89% of L&D leaders shifting to AI-first platforms with 93% worker demand for adaptive training; survey of 556 L&D professionals shows adaptive learning systems in 'Selective' adoption tier (20-50% planned uptake) driven by learner impact. K-12 implementation barriers persist: New Jersey's NJSLA-Adaptive fall 2025 field test highlights regulatory compliance and transparency concerns in large-scale deployment. Practice continues operationalization with strong L&D market signals but faces persistent K-12 regulatory challenges and ongoing implementation complexity.
  • 2025-Q4: Clinical CAT research advances continue with peer-reviewed studies extending adaptive testing to suicide risk and paranoia assessment longitudinally. Healthcare applications demonstrate CAT's capability for intensive clinical monitoring. Educational guidance from major assessment providers (ACER) documents methodology for supporting diverse learner needs. Practice consolidates clinical and educational applications with expanding methodological sophistication; organizational readiness and implementation complexity remain the primary barriers to broader K-12 adoption despite proven methodological validity.
  • 2026-Jan: Market momentum accelerates across educational and corporate sectors. Curriculum Associates launches i-Ready Inform, a streamlined adaptive K-8 assessment for 1M+ educators. Market analysts project continued growth: K-12 testing market to $32.4B by 2030 (9.5% CAGR) and adaptive learning market to $11.81B by 2035 (16.81% CAGR), signaling sustained investment in adaptive platforms. SHL continues production deployment of adaptive talent assessment across global enterprises. Government-level adoption signals emerge: India's IIT Council proposes adaptive testing for JEE Advanced national entrance exam with 2026 pilot. Critical balance: peer-reviewed systematic review documents persistent validity challenges (item bias, test security) that require ongoing research investment. Ecosystem demonstrates operational maturity with expanding market investment, but implementation obstacles and methodological standardization gaps remain primary adoption constraints.
  • 2026-Feb: Technical and deployment momentum expands across assessment domains. Machine learning-based CAT advances outperform traditional IRT with 94-96% accuracy on developmental assessments using ~10 items; reinforcement learning frameworks enable algorithmic innovation in item selection. Corporate talent assessment scales further with named deployments at General Mills, Royal Navy (350+ leaders), and major financial services firms (10,000+ via 360 programs). UK education sector documents adaptive testing adoption (Wales Years 2-9 at national scale; Scottish National Standardised Assessments; AQA exam board analysis of benefits and realistic barriers). Healthcare CAT research deepens equity and clinical feasibility. Critical counterbalance: practitioner analysis surfaces institutional embedding as overlooked adoption constraint—technical strength alone fails without alignment with curriculum, pedagogy, procurement, and organizational change management. Ecosystem reaches critical juncture: proven effectiveness across domains meets persistent implementation complexity and design-fairness trade-offs.
  • 2026-Mar: L&D market adoption signals accelerate. Cornerstone OnDemand Sapient Insights survey (4,600+ organizations, 71 countries) confirms adaptive learning and adaptive assessments moving from experimentation to everyday use across multinational organizations. Enterprise skill assessment deployments expand: named platforms (SkillMatrix, IBM SaaS, Google Talent AI, Microsoft Skills-First) report 28% cycle-time reduction, 80% skill-gain improvement, and 22% increase in non-traditional hiring with maintained retention. Assessment industry shows cautious AI adoption: e-Assessment Association 2026 survey documents sector prioritizing focused AI support (item generation, automated marking, analytics) over full-automation, with significant governance and bias concerns leading adoption decisions. Generative AI tutoring integration accelerates: 62% of top edtech platforms now include LLM-powered tutors by March 2026 (vs. 18% in 2022), enabling real-time formative assessment at scale. K-12 practitioner frameworks emerge: schools deploy adaptive assessments with explicit bias-monitoring cadence (weekly checks, monthly demographic analysis), operationalizing equity governance alongside technical maturity. The practice consolidates operational maturity across corporate L&D (moving to standard practice), education (government and school-level deployments), and healthcare, with implementation discipline and governance frameworks becoming primary adoption levers rather than methodological validation. The adaptive learning market reached USD 5.8B (2025), projected to USD 29.7B by 2034 at 19.5% CAGR, with LLM integration as the primary driver of growth in the current cycle.
  • 2026-Apr: Deployment evidence broadened across adult education and assessment research: the University of Massachusetts ASAP adaptive program serves 10,000+ adult learners with ML-driven question sequencing, a US IES-funded $3.8M project targets a 20,000-learner validation sample for adult literacy CAT modules, and a Bayesian active inference framework achieved 30-40% trial reduction over fixed-design baselines. Assessment integrity under AI pressure advanced as a discipline: peer-reviewed DIF methodology now allows practitioners to identify assessment items vulnerable to chatbot misuse, tested across chemistry diagnostics and university entrance exams. Critical quality concerns for AI-generated items (weak distractors, convergence errors) were documented by NUS medical educators, reinforcing the need for expert review protocols alongside AI-assisted item generation.
  • 2026-May: Institutional adoption momentum accelerates with structural budget reallocation toward adaptive assessment. Universities now dedicate 18–24% of IT budgets to AI assessment tools (up from ~9% two years prior), with adaptive assessment identified as the primary procurement driver over content delivery; Pearson's AI-powered assessment engine serves 4M+ active learners across higher education institutions (Q1 2026 snapshot). Market acceleration confirmed: adaptive learning software market at $4.5B (2024) projected to $15.0B by 2033 (15% CAGR); AI adaptive learning platform market at $10.6B (2025) with corporate L&D segment largest at 38.4% of revenue, reaching $83.2B by 2034 (26.1% CAGR). Patent ecosystem signals rapid innovation: 22+ filings in 2024–2026 vs. ~5 in foundational phase; India emerges as dominant jurisdiction. Skillsoft enterprise platform metrics show 994% YoY growth in AI skills validation completions. Methodological advances: Cambridge-published peer-reviewed CAT using deep learning achieves high-precision ability estimation (posterior SD=0.4) with average 11.2 items. Critical implementation research documents 95% failure rate of AI pilots with organizational factors (70%) identified as root cause, emphasizing change management and outcome accountability over technology as success levers. Ecosystem consolidates: proven operational effectiveness meets persistent implementation complexity as primary tier-defining tension.