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.
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.
Each dot marks the weighted maturity of practices within a domain — hover for a brief summary, click for more detail
AI that generates learning and development content and provides personalised course and resource recommendations. Includes microlearning content creation and skill-based learning paths; distinct from adaptive assessment which evaluates knowledge rather than delivering learning.
AI-powered personalised learning is a proven capability with GA tooling, validated ROI, and broad analyst recognition—but rolling it out successfully remains harder than buying it. Platforms like Cornerstone Galaxy and Degreed Maestro serve thousands of enterprises and report returns above 400%, while conversational learning platforms are experiencing 146% YoY adoption growth. Recent peer-reviewed evidence validates effectiveness: a May 2026 meta-analysis of 36 studies (7,229 participants) shows medium overall effect (g=0.499) with strongest results in collaborative learning (g=1.026) and blended settings (g=0.633). July 2026 agentic tutor deployments (Khan Academy's multi-agent Khanmigo, Duolingo's Max Tutor) serving 50M+ learners demonstrate architectural evolution with documented learning velocity improvements (32% over baseline, 41% vs passive video) and engagement gains (28% increase), establishing a new tier of deployment maturity beyond static personalization. The technology question is settled. The implementation and quality question is not: despite 75.5% of enterprises using AI for content generation, only 10% report learning is consistently personalised, and 41% rate AI-generated content as inadequate or poor. The defining tension at this tier is the gap between adoption breadth and implementation depth. July 2026 research documents a critical engagement durability barrier: only 15% of students with Khanmigo access use it regularly despite rapid growth (17x YoY), shifting diagnosis from technology readiness to behavioural adoption—evidence that AI tutoring has "an adoption problem, not a technology problem." Fifty-three percent of L&D leaders cite AI integration complexity as their biggest challenge; 42% of AI learning initiatives stall before reaching production. Critical assessments document quality trade-offs: peer-reviewed analysis (GenAI and the Mirage of Personalised Learning) finds GenAI struggles with complex problem-solving and remains prone to inaccuracies; 78% of AI-generated lesson plans require major adjustments; metacognitive research shows students using generic AI chatbots score 17% worse on exams despite superior practice performance, with 80% forgetting content created with AI assistance. Pedagogical research offers a counterpoint: structured error-analysis pedagogy (where learners identify and correct errors in AI-generated responses) demonstrates large effect sizes (d≥1.6) for both mathematical reasoning and AI literacy development simultaneously, validating that learner capability depends on instructional design intentionality rather than algorithm sophistication. Simultaneously, learners exhibit systematic anti-AI bias in trusting outputs: credibility perception studies show humans detect AI-generated text below 75% accuracy and systematically discount outputs they perceive as AI-authored, regardless of true source—a trust barrier that persists even when quality is equivalent. Agency constraints further limit personalization value: research across AI learning systems finds 75% restrict students to predetermined pathways despite claiming personalization, prioritizing system authority over learner autonomy and limiting the adaptive potential personalization promises. June 2026 data reveals deeper structural barriers: only 4% of organisations deliver role-specific use-case-driven training despite universal tool deployment; Kyndryl's July survey of 1,100 senior leaders found only 23% feel workforce ready for AI despite 57% having deployed broadly (quantifying the readiness-deployment gap); Go1's survey of 950 enterprise leaders identified cross-functional governance as emerging adoption constraint, with 50% of organisations simplifying for "completion rather than effectiveness." A longitudinal analysis of adoption across five technology waves (ERP, e-procurement, cloud, analytics, current AI) documents persistent 65-80% failure rates despite escalating training investment, questioning whether standard L&D interventions address root adoption barriers (institutional trust, context, workflow redesign). World Bank's Peru deployment (6,600 students, RCT across 390 schools) proved the critical insight: "Technology alone rarely changes outcomes; results came from combining it with training and support"—design and change management matter more than model sophistication. Late-June 2026 research sharpens the diagnosis: employee AI confidence fell 18% year-over-year (steepest single-year drop) while deployment rose to 45%, with 56% receiving no AI training and 62% of workers believing leaders underestimate emotional and psychological impacts—suggesting trust deficits and capability gaps outweigh technology maturity. Organisations that succeed pair capable platforms with disciplined change management, content curation, human oversight, and strong governance—treating the technology as infrastructure requiring parallel investment in compliance, pedagogical redesign, continuous improvement processes, and trust-building in workforce readiness. The July 2026 evidence cascade confirms: technology maturity and deployment breadth have advanced, but adoption durability and implementation depth remain the defining constraints at this tier.
Cornerstone Galaxy anchors the enterprise market at 7,000 organisations and 140 million users; Degreed Maestro, recognised as a 2025 Top HR Product, continues releasing AI-powered personalization features. Real-world deployments validate scale and ROI: Cornerstone University launched SOAR (mobile-first accredited degrees) with 91% persistence rates; Khan Academy deployed Khanmigo with measurable A/B-tested performance gains; Laing O'Rourke achieved 11x faster course production via AI authoring; Workday's internal 3-year deployment across 20,000 employees and 35 countries demonstrated 50% faster course creation and 25% completion lift. Government-backed scale deployments have emerged: SAP's AI-Bilingual Workforce Program (3,000 participants over three years, linked to Singapore's National AI Impact Programme) delivers role-based learning pathways across finance, procurement, HR, supply chain, and manufacturing—signaling public-sector adoption maturity. Enterprise deployments confirm strong cost savings: Docebo serves 3,900+ customers including Booking.com (80% admin overhead reduction), Zoom (2M learners deployed), La-Z-Boy (179% active-learner growth), and SATO (66% turnover reduction); Instacart achieved 612 admin hours saved annually with 82% completion rate via Continu. Market adoption is broad: 87% of L&D teams now using AI tools (up from 34% in 2023) with average ROI of 4.7x within 12 months; IDC documents 536% three-year ROI for AI-powered training platforms; market size reached $440B+ with 26% retention improvement and 45% faster completion on mobile platforms. The microlearning sub-market reached $1.72 billion in 2026 and is projected at $3.10 billion by 2034.
Yet adoption breadth masks implementation quality gaps and workforce readiness barriers. June 2026 data sharpens the diagnosis: ManpowerGroup survey (13,918 workers, 19 countries) found AI usage climbed to 45% but worker confidence fell 18%—the steepest single-year drop—with 56% receiving no AI training and 57% no mentorship. Docebo survey of 2,000 workers found 85% cannot apply what they learned in training to their actual job, with 56% lacking time to learn and 78% training disconnected from actual work tools. Adoption quality research (2,000+ respondents) shows 79% of L&D teams leverage AI but only 35% progressed beyond experimental stage, with 91% unable to fully redefine workflows. A study of 200+ enterprise leaders found 75.5% use AI for content generation, but only 10% experience consistent personalisation, and 41% rate AI-produced content as inadequate or poor. LPI capability survey of 3,575 L&D professionals across 1,874 organisations found AI literacy as lowest-scoring domain (1.54/4) with 71% citing readiness issues. Acorn survey found 77% of executives believe managers are prepared for AI capability development, but only 9% of individual contributors agree (91% say managers unprepared). Quality assessment data shows 78% of AI-generated lesson plans require major adjustments, and peer-reviewed research shows students using generic AI chatbots scored 17% worse on exams despite superior practice performance, with 80% forgetting content created with AI assistance—indicating metacognitive risks. A critical negative signal: i-Ready platform, used by millions, faces lawsuits and lacks peer-reviewed evidence despite widespread adoption. Governance maturity is advancing: Cornerstone Galaxy achieved DISA Impact Level 4 and ISO 42001 (AI Management System standard) by April 2026. However, June 2026 research identifies the core structural gap: only 4% of organisations deliver role-specific use-case-driven training despite universal tool deployment; most are training against outdated job descriptions rather than redesigned workflows. The evidence is clear: technology alone is necessary but insufficient; success requires organizational change management, workflow integration, content curation discipline, and human-centred safeguards rather than algorithmic personalization alone.
— Practitioner webinar (Nelson Sivalingam, HowNow CEO; Talha Faridy, Easygenerator) documents critical gap: efficiency is not effectiveness; 'scaling poor-quality content fast' damages future engagement. Identifies context and author-first design as non-negotiable; completion rate remains ~15% despite platform advancement—emphasizes implementation discipline over technology selection.
— Workday's internal deployment (20,000 employees, 35 countries) over 3 years shows 50% faster course creation, 25% completion rate lift, 86% completion on AI-native course, and 70% reduction in content maintenance—validating platform-based personalized learning at enterprise scale.
— Government-backed large-scale deployment: SAP's AI-Bilingual Workforce Program (3,000 participants over 3 years) delivers role-based learning pathways across finance, procurement, HR, supply chain, and manufacturing, linked to Singapore's National AI Impact Programme—evidence of public-sector scale adoption.
— Peer-reviewed credibility study (PLOS One) shows humans exhibit systematic anti-AI bias: participants less likely to agree with outputs believed AI-generated regardless of true source. Detection accuracy consistently below 75%, highlighting learner trust barriers for AI-personalized content adoption.
— Peer-reviewed conference research (22.5% acceptance rate) reveals critical limitation: 75% of AI learning systems restrict student agency to predetermined pathways despite personalization claims—finding applies across all vendors and suggests current implementations prioritize system authority over learner autonomy.
— IDC study documents 536% three-year ROI for AI-powered training; named deployments (WellSpan Health, appliance manufacturer) achieved 45% onboarding speedup, 20% sales lift, 35% cost reduction, and 92% coaching confidence gain—analyst-backed quantified deployment outcomes.
— Peer-reviewed experimental study (Frontiers in Education) shows GenAI-integrated error analysis pedagogy achieves large effect sizes (d≥1.6) on both mathematical reasoning and learners' ability to detect AI errors, validating instructional design approach to build AI literacy while developing domain knowledge.
— Critical practitioner analysis identifies adoption-effectiveness gap: 87% of L&D teams use AI tools but 74% of companies falling behind on skill development, and Harvard Business School experiment shows GPT-4 reduces task performance 19% outside its capability range—validates that tools alone insufficient without governance, execution capacity, and measurement.