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 individual productivity, communication, organisation, and self-directed learning. The most polarised domain: writing assistance and meeting summarisation are good practice, but nearly half the practices are bleeding-edge — personal AI agents, life planning, and autonomous scheduling lack reliable implementations. Most trajectories are stalled, reflecting a gap between consumer hype and sustained daily utility.
The headline: New research finds AI makes people faster, more confident, and measurably less accurate — and they cannot tell. Budget for the checking, not just the tool.
This is the corner of AI that touches everyone's working day: drafting email, writing documents, building slides, prioritizing tasks, doing research. Nearly every organization has these tools now — Microsoft alone reports over 30 million paid Copilot seats — and almost none can show what they returned. A small group extracts real value by redesigning how work flows around the tool and by measuring one or two specific workflows properly; the majority buy licenses, watch a third of staff use them, and renew on faith. That gap is closing on its own terms this year, because finance teams have started demanding proof rather than anecdotes. If you cannot name the workflow, the baseline and the delta before your next renewal, you are in the majority.
Microsoft posted its biggest Copilot quarter — and the return gap widened anyway. Paid seats passed 30 million with net additions doubling, alongside large agentic deployments (software that acts on its own without being prompted) at Atos, EY and Chow Tai Fook. Yet only 35.8% of licensed employees use Copilot regularly, against 83.1% for ChatGPT, and a separate survey found 95% of organizations reporting no measurable return despite 80% adoption. Weight your renewal case on measured outcomes in one or two workflows, not on seat counts.
AI advice cut people's accuracy from 27% to 9% while nearly tripling their confidence. The same peer-reviewed study found the recognition of uncertainty collapsing from 44% to 3%, and three education trials this fortnight showed the same pattern: high satisfaction, faster task completion, zero actual learning. Where your teams use AI to decide rather than draft, build in a step that forces someone to state their own conclusion before seeing the AI's.
Independent testing put error rates at 8–23% across the six leading legal AI platforms. Harvey, CoCounsel, Lexis+, Westlaw, Spellbook and Ironclad all hallucinated — confidently made things up — at those rates on research and contract review, with vendor accuracy claims found to be self-referential. Fabricated citations then turned up in PwC's own AI-generated client reports, the second Big Four firm caught after KPMG. Ask vendors for third-party accuracy testing, and treat its absence as an answer.
The verification bill finally got a number. A survey of 6,000 workers found AI saving eleven hours a week and taking 6.4 hours back in checking and fixing — a real but modest net gain — while 69% admitted shipping unreviewed work anyway. Fund review capacity as an explicit line in the rollout, or the net gain quietly disappears into rework.
Finance teams are killing the "saves time" business case. Among 830 IT decision-makers, time savings fell 5.8 points as the leading justification for AI purchases, replaced by demands for direct profit-and-loss evidence. Contracts renewing in 2027 will need a named workflow, a baseline measured before the tool arrived, and a delta — start collecting that now, because you cannot reconstruct a baseline retrospectively.
EU rules are turning translation and accessibility into market-entry requirements. The AI Act, Medical Device Regulation, Machinery Regulation and Accessibility Act converge through late 2027 to make multilingual and accessible output mandatory for selling into Europe. Meanwhile 45% of workers say accessibility is absent or unclear in their organization's AI policy. Put both on the procurement checklist this quarter rather than next year.
Your personal tools are becoming plumbing for AI assistants. Superhuman shipped an MCP server — a connector standard for AI tools — that lets Claude and ChatGPT read, draft and send mail directly; note-taking and calendar products are moving the same way. Decide now which systems you are willing to let an assistant act inside, and write it down, before someone connects them for you.
The verification tax lands hardest on your best people. An analysis of 2,771 verified reviews found 88% of buyers are casual users who genuinely save time — while the 12% who write professionally report that review overhead cancels the gain outright. The tool helps most where the stakes are lowest.
Your data is the brake, not the model. Two-thirds of enterprises have delayed or cancelled Copilot deployments over fears it will surface confidential internal files, with 75% of C-level respondents hesitant. No model upgrade fixes this; it is a permissions and information-governance project, and it needs an owner outside IT procurement.
Individual wins do not scale to the organization. Individual task gains are reported as high as 5x, yet only 29% of organizations see significant return. What separates them is organizational design, not tooling: at Ropes & Gray, Citigroup and Mars, internal champion networks drove roughly twice the sustained usage of top-down mandates.
Go deeper: the full Personal Effectiveness 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.