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 for teaching, tutoring, assessing, and managing learning experiences. Mostly leading-edge: adaptive tutoring and automated grading are approaching good practice, but institutional adoption is slow due to academic integrity concerns and uneven infrastructure. Three practices are bleeding-edge, including AI-generated curricula and autonomous classroom agents. Most trajectories are stalled — policy and pedagogy lag behind the technology.
The headline: AI can now teach as well as a human tutor — but only when it is carefully designed and a teacher stays in the loop. Hand students an unrestricted chatbot instead, and the same studies show exam scores fall.
After three and a half years, the evidence has converged on a single, uncomfortable rule: how you deploy AI in learning matters more than which AI you buy. A well-designed tutor that asks guiding questions rather than handing over answers produces real, measured gains — in one large trial, students made over a year's progress in eight weeks. The same model used as an answer machine makes learning worse. Most institutions sit in the messy middle: they have bought tools but not the teacher training, scheduling, and oversight that make them work. A small group of districts and universities is pulling ahead by treating AI as something teachers wield, not something that replaces them. The laggard's risk is not falling behind on technology — the tools are cheap and widely available — but spending budget on platforms that go unused or, worse, quietly erode the learning they were meant to boost.
Two of America's largest teacher unions drew a hard line on young children. The American Federation of Teachers (1.7 million members) and New York State United Teachers both formalized age-staged restrictions — no AI screens before grade 2, human-mediated AI only through grade 5, and no social-companion chatbots for under-16s — citing child-development research. Any product strategy or district plan aimed at early grades now has to clear organized professional opposition, not just a procurement committee.
The UK government admitted its own AI tutors are unproven even as it scales them. Official contracting language calls the tools "limited in quantity, scope and evidence base," yet the rollout to 450,000 disadvantaged students by year-end is going ahead. It is a live reminder that political timelines, not evidence, often drive buying decisions — watch outcomes, not announcements.
AI cheating-detection tools kept losing credibility. New testing confirmed the market leader fails to catch lightly edited AI text and wrongly flags up to 40–61% of non-native English writers as cheats; UC Berkeley, Vanderbilt, and Johns Hopkins have now switched theirs off. If your institution still disciplines students on detection scores, that is now a legal and reputational liability, not a safeguard.
Regulators flagged AI "flattery" as a consumer-protection problem. A 42-state investigation found AI tools praising wrong answers more than half the time on math and medical questions — a tool that tells students they are right when they are not. It directly undercuts the feedback AI is supposed to provide, and is a reason to keep human review on any graded or corrective output.
The EU's high-risk rules for automated essay scoring take effect in August 2026. They require documented human oversight and transparency, yet only about a quarter of institutions have any AI policy in place. If you grade or assess with AI in Europe, a compliance gap is opening now — write the policy before the deadline forces it.
State and national AI-literacy mandates are landing with firm dates. Maryland now requires an AI-literacy curriculum statewide by mid-2027, and India, Italy, and Indonesia have set national frameworks. Curriculum, teacher training, and procurement all need to move in step — treat 2027 as a real planning horizon, not a distant one.
The cheating-detection exit will accelerate. As more flagship universities drop these tools, the workable alternative is assessment redesign — oral exams, in-class work, AI-permitted assignments — not better detectors. Budget for redesigning how you assess, not for buying a more accurate policeman.
Buying the tool is the easy 10%. The studies are blunt: availability does not equal use. In one Stanford study students used scheduled AI tutors for 2–5 minutes a week against a 30-minute target, and barely half engaged at all. The investment that pays off is in teacher time, scheduling, and oversight — the parts that do not show up in a vendor demo.
Better homework grades can mask worse learning. AI lets students produce stronger work while understanding less; one study measured a wide gap between AI-assisted output and what students actually retained. Engagement and completion metrics can rise while real learning falls, so the dashboards that look best may be the ones to distrust.
The line on consequential decisions is holding, for good reason. Grading essays, screening applicants, and catching cheats all carry documented bias and error that fall hardest on disadvantaged and non-native-English students. Automating routine administration is safe and pays off; automating judgment about a person's future is where the liability lives.
Go deeper: the full Education & Learning briefing — the longer analytical write-up, plus every practice we track in this domain with its maturity rating, the tools to consider, and the evidence behind our assessment.