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: Your staff have adopted AI for writing, email, and research — but more than half the time it appears to save is being spent checking and fixing its output. The productivity is real for skilled users in narrow tasks; the company-wide return mostly is not.
Almost every organization is now in the pack on these everyday productivity tools — drafting, email, summarizing documents, scheduling. Adoption is near-universal and the tools genuinely work. What separates the leaders from everyone else is not having the tools but governing them: a small group extracts real value from tightly defined, well-supervised tasks, while the majority cannot tell whether they are saving money or just moving the work around. Only about one in eight organizations reports a meaningful business return despite roughly nine in ten using AI. If you are deploying broadly and assuming the savings will show up on their own, you are in the large group that is busy but not better off.
A major study finally put a number on the hidden cost. Researchers at Stanford and UC Berkeley tracked 6,000 office workers and found they gain about 11 hours a week from AI but lose 6.4 of them to "botsitting" — fixing, re-running, and double-checking what the AI produced. Nearly 70 percent admitted shipping work they had not verified. Treat raw "hours saved" claims from vendors as roughly half-true until you measure the checking time too.
The benefit goes to your experts, not your beginners. An analysis of about 400,000 AI work sessions showed that skill — not seniority or job title — predicts who gets value: experts get far more out of each request, while novices hit errors and give up on roughly one in five attempts. Target training and licenses at experienced staff first; broad, untargeted rollouts dilute the return.
Skill erosion is moving from theory to measurement. A separate survey found 39 percent of users say leaning on AI is eroding their own capabilities, rising to 46 percent among the youngest workers, with half admitting they over-rely on it. The cheap-seeming productivity gain carries a long-term capability cost you will not see on this quarter's numbers.
Most leaders cannot see what their own people use. Only 41 percent of HR leaders could name two tools their workforce actually relies on, and nearly a third still think adoption is at the pilot stage. Unmanaged "shadow" AI is already in your building; the first step is finding out what is actually being used.
EU rules tighten on high-stakes language work. From late 2027, AI-assisted translation in healthcare, legal, and other critical settings becomes formally high-risk under the EU AI Act, turning procurement from a cost decision into a compliance one. If you operate in regulated, multilingual contexts, start building the human-review and audit trail now.
Courts are formalizing AI accountability. More than 25 US federal courts now require lawyers to certify and verify AI use before filing, after roughly 900 documented cases of AI inventing facts. Expect this verify-and-attest expectation to spread to other professions; put a named human owner on any AI output that carries legal or financial weight.
The "agentic" wave is the next budget ask — and the next risk. Vendors are pushing software that acts on its own (agentic AI) across email and spreadsheets, yet confidence in these systems has dropped sharply and security researchers keep finding ways to hijack them. Pilot narrowly, demand controls, and do not let an agent touch sensitive data without sign-off.
Better models do not fix the core problem. The cost is verification: fluent AI output hides its mistakes rather than reducing them, so checking does not get cheaper as the technology improves. In regulated work the checking burden can exceed the time saved entirely.
Adoption has run ahead of value, and the gap is structural. Companies keep buying because competitors are, not because the numbers add up — only about one in five finance teams that automated reports can show a clear return. Unchanged workflows and weak measurement, not weak tools, are the bottleneck.
Trust is the quiet casualty. Recipients rate a manager's sincerity far lower when an email reads as AI-written, and audiences spot AI tone within seconds. Used carelessly at scale, these tools can cost you relationships and credibility faster than they save you time.
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.