{
  "id": "personal-effectiveness",
  "label": "Personal Effectiveness",
  "description": "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.",
  "icon": "✨",
  "filters": [
    "growing"
  ],
  "hasSummary": true,
  "hasExecSummary": true,
  "practiceCount": 14,
  "evidenceCount": 2598,
  "practices": [
    {
      "slug": "accessibility-support-for-individuals",
      "name": "Accessibility support for individuals",
      "tier": "good-practice",
      "trend": "steady",
      "blockerType": null,
      "description": "AI tools that support individual accessibility needs including transcription, screen reading, alt-text generation, and voice control. Includes personal accessibility assistance and accommodation tools; distinct from accessibility auditing in product design which tests products rather than assisting individuals.",
      "evidenceCount": 201
    },
    {
      "slug": "brainstorming-and-ideation-support",
      "name": "Brainstorming & ideation support",
      "tier": "leading-edge",
      "trend": "steady",
      "blockerType": null,
      "description": "AI that helps individuals generate ideas, explore possibilities, and think through problems from multiple angles. Includes structured ideation and creative thinking prompts; distinct from content calendar planning which generates topic ideas rather than general brainstorming.",
      "evidenceCount": 176
    },
    {
      "slug": "communication-style-adaptation",
      "name": "Communication style adaptation",
      "tier": "leading-edge",
      "trend": "slowing",
      "blockerType": null,
      "description": "AI that adapts writing tone and style for different audiences — executives, peers, clients — from a single draft. Includes audience-aware rewriting and formality adjustment; distinct from brand-voice workflows which enforce brand rather than personal communication style.",
      "evidenceCount": 179
    },
    {
      "slug": "decision-support-and-reasoning-frameworks",
      "name": "Decision support & reasoning frameworks",
      "tier": "bleeding-edge",
      "trend": "steady",
      "blockerType": null,
      "description": "AI that helps individuals structure decisions, evaluate options, and apply reasoning frameworks to complex choices. Includes decision matrix generation and pro/con analysis; distinct from feature prioritisation which applies frameworks to product decisions rather than general personal choices.",
      "evidenceCount": 200
    },
    {
      "slug": "email-management-drafting-and-triage",
      "name": "Email management — drafting & triage",
      "tier": "bleeding-edge",
      "trend": "steady",
      "blockerType": null,
      "description": "AI that drafts email responses and triages incoming messages by priority, intent, and required action. Includes smart reply generation and priority inbox management; distinct from meeting intelligence which handles spoken rather than written communication.",
      "evidenceCount": 188
    },
    {
      "slug": "personal-knowledge-management-and-organisation",
      "name": "Personal knowledge management & organisation",
      "tier": "leading-edge",
      "trend": "accelerating",
      "blockerType": null,
      "description": "AI that organises personal files, notes, and information and enables semantic retrieval across personal knowledge stores. Includes automated tagging and cross-note linking; distinct from enterprise search which operates across organisational rather than personal knowledge.",
      "evidenceCount": 174
    },
    {
      "slug": "personal-research-and-reading-acceleration",
      "name": "Personal research & reading acceleration",
      "tier": "bleeding-edge",
      "trend": "steady",
      "blockerType": null,
      "description": "AI that accelerates personal research and reading through summarisation, synthesis, and intelligent highlighting of key content. Includes article distillation and research compilation; distinct from deep research tools which autonomously gather sources rather than processing provided ones.",
      "evidenceCount": 187
    },
    {
      "slug": "presentation-creation-and-enhancement",
      "name": "Presentation creation & enhancement",
      "tier": "leading-edge",
      "trend": "accelerating",
      "blockerType": null,
      "description": "AI that creates, structures, and enhances slide presentations from notes, outlines, or spoken input. Includes slide generation and visual enhancement; distinct from event collateral in marketing which produces marketing rather than individual presentations.",
      "evidenceCount": 169
    },
    {
      "slug": "skill-acquisition-and-deliberate-practice-support",
      "name": "Skill acquisition & deliberate practice support",
      "tier": "leading-edge",
      "trend": "steady",
      "blockerType": null,
      "description": "AI that supports individuals in acquiring new skills through structured practice, feedback, and progression tracking. Includes practice exercise generation and personalised feedback; distinct from L&D recommendations which suggest resources rather than structuring practice.",
      "evidenceCount": 181
    },
    {
      "slug": "spreadsheet-and-data-task-automation",
      "name": "Spreadsheet & data task automation",
      "tier": "bleeding-edge",
      "trend": "steady",
      "blockerType": null,
      "description": "AI that automates spreadsheet tasks including formula creation, data analysis, pivot tables, and formatting from natural language. Includes Excel/Sheets AI assistants and formula generation; distinct from natural language to SQL which queries databases rather than manipulating spreadsheets.",
      "evidenceCount": 175
    },
    {
      "slug": "task-prioritisation-and-focus-management",
      "name": "Task prioritisation & focus management",
      "tier": "leading-edge",
      "trend": "steady",
      "blockerType": null,
      "description": "AI that prioritises tasks, manages calendars, and helps protect deep work time through intelligent scheduling and distraction management. Includes priority scoring and calendar optimisation; distinct from project management tools which coordinate teams rather than individual productivity.",
      "evidenceCount": 176
    },
    {
      "slug": "translation-and-cross-language-communication",
      "name": "Translation & cross-language communication",
      "tier": "good-practice",
      "trend": "steady",
      "blockerType": null,
      "description": "AI real-time translation of written and spoken communication for individuals working across languages. Includes live conversation translation and document translation; distinct from content localisation in marketing which adapts campaigns rather than facilitating individual communication.",
      "evidenceCount": 210
    },
    {
      "slug": "writing-assistance-drafting-editing-and-style",
      "name": "Writing assistance — drafting, editing & style",
      "tier": "good-practice",
      "trend": "steady",
      "blockerType": null,
      "description": "AI that assists with drafting documents and improving writing quality through editing, style, and tone suggestions. Includes document drafting from outlines and readability improvement; distinct from content generation in Marketing which targets published content rather than personal professional writing.",
      "evidenceCount": 193
    },
    {
      "slug": "meeting-intelligence-transcription-summaries-and-actions",
      "name": "Meeting intelligence — transcription, summaries & actions",
      "tier": "good-practice",
      "trend": "steady",
      "blockerType": null,
      "description": "AI that transcribes meetings, generates summaries, extracts action items, and tracks follow-up completion. Includes speaker attribution and automated task assignment; distinct from email summarisation which processes written rather than spoken communication.",
      "evidenceCount": 189,
      "crossListed": true,
      "primaryDomain": {
        "id": "research-analysis",
        "label": "Research & Knowledge",
        "icon": "🔬"
      }
    }
  ],
  "summary": "## Where AI Stands in Personal Effectiveness\n\nPersonal effectiveness remains a sharply polarised domain. At one end, writing assistance, translation and accessibility support have become ordinary infrastructure. They ship inside the operating system and the office suite, and their value on bounded tasks is no longer seriously disputed. DeepL, Microsoft Teams and Google's Gemini-powered Live Translate now run real-time interpretation at conference and enterprise scale. The US Census Bureau's first independent federal measurement found that 32% of workers use AI for writing and documentation, saving a median of one to two hours a week. At the other end sit decision support, email triage, spreadsheet automation, research acceleration and task prioritisation. Every major vendor ships capable tooling in these areas, yet very few organisations can show what it is worth. Between the two ends is a middle band: presentation generation, personal knowledge management, style adaptation, brainstorming and practice-based learning. Adoption there is real, but whether the output is good enough is still an open question.\n\nWhat sets this domain apart is that the unit of adoption is the individual, and individual gains leak away. McKinsey's 2026 survey finds that 80% of professionals report personal productivity gains, while only 37% see any contribution to EBIT. That figure is unchanged on last year. Deloitte's survey of 25,000 UK workers finds 63% using generative AI for work, yet 31% report no time saved and only 7% save five hours or more a week. Microsoft reports 30 million paid Microsoft 365 Copilot seats, and Gartner still says firms are struggling to realise the gains and cost savings. The minutes people recover are real. They are absorbed by review, rework, extra volume and scope creep before they reach a ledger. The domain is also unusual in that its outputs are judged socially. A Duke study published in PNAS (N=4,439) found that people who use AI writing help are rated lazier and less competent than peers producing identical work. A Harris Poll of 1,175 US writers found that 60% distrust their own AI-assisted output.\n\nThe arc has therefore moved from generation to verification. The tools can now draft the email, build the deck, write the formula and summarise the paper. The binding questions are two. Can the organisation afford the human checking that makes the output safe? And does heavy use wear down the judgement doing the checking? Where organisations build that verification layer deliberately, momentum is building: post-editing pipelines in translation, governed vaults in knowledge management, supervised side panes rather than in-cell formulas in Excel. Where tools are licensed and left to spread on their own, adoption either stalls at the pilot stage or turns into unmeasured and often unsanctioned use. Deloitte found that 31% of UK workers use AI without their employer's knowledge.\n\n## What's New, 2026-09-13 to 2026-09-27\n\nThe clearest movement this fortnight was upward, in two middle-band practices. **Presentation creation** gained both breadth and enterprise pull. Menlo Ventures' survey of 5,067 US adults put presentation building at 52% AI use. Templafy reports that enterprise use of its document agents rose from roughly zero in October 2025 to 53% by June 2026, with median pitch-deck completion falling from two hours to eight minutes. On 25 September Microsoft folded PowerPoint creation and editing into a redesigned Copilot, and Gamma launched an Enterprise tier. The quality caveats have not gone away. Gamma's own creative lead conceded that much of the category produces \"AI slop\". Independent export tests still find charts and icons turned into flat images and text boxes overflowing. A post-mortem on Tome, which shut in April 2025 with 20 million users but roughly $2 million in annual recurring revenue, is a reminder that usage and a viable business are not the same thing. **Personal knowledge management** shifted from individual tinkering towards governed infrastructure. The obsidian-tc server exposes 163 governed capabilities with folder-level access controls and audit logs. Eli5 built a company \"second brain\" that uses structured indexing instead of vector search and requires human approval on every write. The limits are just as explicit: AI-maintained wikis reportedly hit a maintenance wall at around 100 pages, and 69% of users in one survey call themselves digital hoarders.\n\nElsewhere, the new evidence sharpened known limits rather than moving them. Microsoft formally retired the =COPILOT() worksheet function on 14 September, conceding that non-deterministic output cannot live inside a spreadsheet. In the same fortnight, Handshake's ATLAS-Finance benchmark found the best model, Claude Opus 5, passing just 12.3% of expert finance tasks. Against that, Vena found 86% of 364 finance professionals already using AI inside Excel. On decision support, a US Special Operations Command Pacific (SOCPAC) intelligence analyst's AI turned an intelligence synthesis into a fabricated nuclear-weapons claim, and the error was caught only minutes before a ship boarding. On brainstorming, a Procter & Gamble field experiment with 791 R&D staff found that AI raised idea quality by 9.6% but cut selection of the best option from 50% to 37%. On reading, a Harvard Business Review study of more than 1,000 participants found that easier AI-mediated access reduced what people remembered. In translation, DeepL ran live voice interpretation across more than 1,000 sessions at Salesforce's Dreamforce conference. A 71,262-segment Welocalize study found that AI with human post-editing outperformed direct LLM translation, while medical-device content concentrated the severe errors. New accessibility and skills evidence fitted the existing picture. Level Access found that 80% of accessibility professionals use AI, but those with governance and training in place are about six times as likely to benefit. Workera found that 56.4% of regular AI users get no work time to upskill.\n\n## Key Tensions\n\n- **Individual gains that never reach the ledger.** Across the domain, self-reported time savings coexist with flat organisational returns. In McKinsey's survey, 80% report personal gains but only 37% see EBIT impact. Orbit Media found 92% of bloggers using AI but only 14% reporting strong results. Bain calls this the micro-productivity trap: task-level savings that are reabsorbed into approval queues and extra work, never turned into margin.\n\n- **Faster ideas, weaker judgement.** The evidence increasingly separates generating from deciding. Procter & Gamble's R&D experiment found better ideas but worse selection. Atlassian found AI speeds ideation by 43% but lowers ownership and originality, and peer review beat AI brainstorming on peak originality. HEC Paris analysis shows that when humans override a more accurate AI, the combined result (76%) barely beats the human working alone (74%) and falls well short of the AI's 90%. Wharton's agency-decay analysis ties heavier reliance to weaker critical thinking.\n\n- **Autonomy pushed forward at the platform, pulled back at the edge.** Microsoft made Copilot in Outlook agentic, and 32% of US consumers in Menlo's survey let AI act without approval. Yet Google's own developer guidance recommends draft-only email scopes, and Microsoft retired in-cell AI formulas because their output is not reproducible. Vendors are shipping autonomy and building in retreats from it in the same month.\n\n- **Accuracy ceilings bite where stakes are highest.** Frontier models reach 76.7% on Vals AI's Excel modelling benchmark but only 64% on numerical accuracy. They answer 59.7% of Taste-Bench's decision-fork questions correctly, however much reasoning budget they are given. Speech recognition accuracy falls from 82% to 3.9% for profoundly impaired speech. Asylum claims have been undermined by mistranslated pronouns and colloquialisms. The gap is widest exactly where errors are costliest.\n\n- **The human layer is now the product.** The deployments that work are defined by the checking built around them, not by the model. Examples include AI post-editing in translation, human approval on writes in Eli5's knowledge base, and formal editing that roughly doubles blogging performance. NATO now disqualifies job applications prepared with AI tools, and EU AI Act transparency duties exempt content only where a named human takes editorial responsibility. Accountability is being written into procurement.\n\n## Top 10 Evidence Items\n\n1. **AI Productivity Gains and EBIT Impact: McKinsey and Deloitte 2026 Survey Data** (adoption-metric) — Gives the domain's central paradox in one figure: 80% report personal gains while only 37% see EBIT impact, unchanged year on year. https://www.marketscale.com/industries/software-and-technology/80-of-workers-say-ai-helps-but-only-37-see-ebit-impact\n2. **Deloitte UK AI Workplace Survey: Uneven Payoff and Shadow AI (31% Zero Time Savings, 7% Save 5+ Hours)** (adoption-metric) — Shows the leakage of individual gains at population scale: 31% report no time saved, 7% save five hours or more, and 31% use AI without employer knowledge. https://www.cfo.com/news/shadow-ai-use-booms-in-new-deloitte-uk-survey/830747/\n3. **Duke PNAS credibility penalty: AI writing-assistance users rated lazier and less competent despite identical output (N=4,439)** (opinion) — Supports the claim that outputs in this domain are judged socially: users are rated lazier and less competent despite identical work. https://www.metaintro.com/blog/ai-sounding-writing-work-credibility-review\n4. **When AI Becomes a Teammate** (opinion) — Shows the shift from generating to verifying: AI improved idea quality but weakened selection of the best option, from 50% to 37%. https://stratify.academy/en/blog/when-ai-becomes-a-teammate\n5. **SOCPAC Intelligence Failure: AI Fabricated Nuclear Weapons Claim, Nearly Triggered Ship Boarding** (news-coverage) — The starkest failure on file: a fabricated claim caught minutes before a ship boarding shows what unverified decision support risks. https://www.techtimes.com/articles/327796/20260920/us-military-almost-boarded-chinese-ship-over-ai-hallucinated-nuclear-claim.htm\n6. **Microsoft Excel: COPILOT Function Retirement (September 14, 2026)** (tutorial) — A vendor retiring in-cell AI formulas because output is not reproducible is a concrete case of autonomy being pulled back at the edge. https://support.microsoft.com/nl-nl/excel/functions/copilot-function\n7. **AI Post-Editing at Scale: What a 71,262-Segment Study Reveals About the Future of Translation** (research-paper) — Best evidence that a deliberate human verification layer works: AI with human post-editing beat direct LLM translation across 71,262 segments. https://www.welocalize.com/insights/ai-post-editing-at-scale-what-a-71262-segment-study-reveals-about-the-future-of-translation/\n8. **Tome AI presentation tool shutdown: 20M users, $300M Series B valuation, ~$2M ARR, April 2025 discontinuation** (opinion) — Reminds readers that mass adoption is not a viable business: 20 million users but about $2 million in recurring revenue, behind this fortnight's presentation surge. https://start-wise.io/blog/what-happened-to-tome-ai/\n9. **Eli5 company second brain: avoiding vectors via structured indexing** (case-study) — Illustrates governed knowledge management, with human approval on every write, as the deployment pattern that works. https://www.eli5.io/insights/second-brain-for-business-outline-claude-mcp\n10. **Human Override Paradox: How Human Authority Erases AI's Accuracy Advantage** (opinion) — Shows that checking cannot be assumed to help: human override of a more accurate AI barely beats the human alone, which undermines the assumption that human review always improves results. https://www.hec.edu/en/dare/tech-ai/giving-humans-final-say-can-make-ai-decisions-worse",
  "execSummary": "**The headline:** Your people's AI time savings are real, but they are not reaching the bottom line. The companies getting value pay for the human checking. Buying more licenses does not do it.\n\n### The Picture\n\nMost companies now have AI writing, email and meeting tools in the hands of staff, and staff use them: 63% of UK workers in Deloitte's survey of 25,000 use generative AI for work. The gains mostly stop at the individual. McKinsey finds 80% of professionals report personal productivity gains, while only 37% of organizations see any contribution to operating profit, the same as a year ago. A small group is pulling ahead by building the review layer on purpose: human editors on AI drafts, sign-off on anything AI writes into shared systems, and AI kept beside the numbers rather than inside them. The rest have unmeasured and often unsanctioned use: 31% of UK workers use AI without their employer knowing.\n\n### This Fortnight\n\n- **Microsoft retired its AI formula inside Excel cells on 14 September, conceding that the answers were not reproducible enough for spreadsheets.** In the same fortnight, a test of expert finance tasks built with bankers from Morgan Stanley and UBS found the best AI passed only 12.3%, even though a separate survey found 86% of finance staff already use AI inside Excel. Any AI-assisted model going to clients or the board needs a second person checking the numbers.\n\n- **Presentation-building AI moved into the mainstream: Microsoft folded PowerPoint creation into a redesigned Copilot on 25 September, and Gamma launched an enterprise tier.** Templafy reports median pitch-deck time falling from two hours to eight minutes, but independent tests still find charts turned into flat images and overflowing text, and Gamma's own creative lead admits much of the category produces \"AI slop\". Budget for design and fact review, because a faster deck is not necessarily a better one.\n\n- **A US military analyst's AI turned an intelligence summary into a fabricated nuclear-weapons claim, which was caught minutes before armed personnel boarded a ship.** The output looked like an authoritative report and moved through command channels without challenge. Any AI-generated briefing that feeds an executive decision needs a named person who checks the sources as well as reading the summary.\n\n- **A Procter & Gamble experiment with 791 R&D staff found AI improved idea quality but cut how often teams picked the best option, from 50% to 37%.** Other research this fortnight points the same way: AI widens the options but weakens the choice between them. Use it to generate options and keep the final choice with people.\n\n- **DeepL ran live AI voice interpretation across more than 1,000 sessions at Salesforce's Dreamforce conference.** A separate study of 71,262 translated segments found that AI followed by human editing beat AI alone, and that severe errors clustered in medical-device content. Live interpretation is ready for events and internal meetings. Regulated or legal content still needs a human editor.\n\n### Coming Up\n\n- **Email AI is becoming agentic (acting on its own without being prompted): Microsoft's Copilot in Outlook now sorts mail and reschedules meetings by itself.** Google's own developer guidance tells builders to let AI draft email but not send it, and 32% of US consumers already let AI act without approval. Set a default rule that AI drafts and people send before this reaches your staff's inboxes.\n\n- **Rules on who answers for AI output are hardening, and they point to a named human.** Since 2 August, EU AI Act transparency duties exempt AI-generated content only where a named person takes editorial responsibility, and NATO now disqualifies job applications prepared with AI. Decide who signs off AI-generated material you publish or submit, and expect clients' requests for proposals to start asking.\n\n- **Microsoft now has 30 million paid Copilot seats, and Gartner says firms are struggling to realize the promised gains and cost savings.** One adoption review found 64% of Copilot licenses go unused, which it put down to poor workflow integration and training rather than product quality. At your next renewal, tie seat counts to measured use and named workflows, not headcount.\n\n### What's Hard About This\n\n- **Time saved on individual tasks gets absorbed before it reaches the ledger.** Deloitte finds 31% of UK workers save no time at all, and the minutes others save go into review, rework, extra volume and approval queues, which Bain calls the micro-productivity trap. Capturing the value means taking steps out of the work itself, which no tool does for you.\n\n- **Heavy use weakens the judgment that is supposed to catch AI's mistakes.** In a BCG survey, 90% of senior leaders said employees have stopped checking AI's work, and a Harvard Business Review study found easier AI access reduced what people remembered. The human-in-the-loop (a person who reviews each AI output) only protects you if that person is still paying attention.\n\n- **AI-assisted work carries a social penalty that better tools will not remove.** A Duke study of 4,439 people published in PNAS found that people who use AI writing help are rated lazier and less competent than peers producing identical work. Your disclosure policies and the way managers reward work need to account for that, or staff will simply hide their use.",
  "headline": "Your people's AI time savings are real, but they are not reaching the bottom line. The companies getting value pay for the human checking. Buying more licenses does not do it.",
  "execSummarySections": [
    {
      "id": "the-picture",
      "title": "The Picture",
      "body": "Most companies now have AI writing, email and meeting tools in the hands of staff, and staff use them: 63% of UK workers in Deloitte's survey of 25,000 use generative AI for work. The gains mostly stop at the individual. McKinsey finds 80% of professionals report personal productivity gains, while only 37% of organizations see any contribution to operating profit, the same as a year ago. A small group is pulling ahead by building the review layer on purpose: human editors on AI drafts, sign-off on anything AI writes into shared systems, and AI kept beside the numbers rather than inside them. The rest have unmeasured and often unsanctioned use: 31% of UK workers use AI without their employer knowing."
    },
    {
      "id": "this-fortnight",
      "title": "This Fortnight",
      "body": "- **Microsoft retired its AI formula inside Excel cells on 14 September, conceding that the answers were not reproducible enough for spreadsheets.** In the same fortnight, a test of expert finance tasks built with bankers from Morgan Stanley and UBS found the best AI passed only 12.3%, even though a separate survey found 86% of finance staff already use AI inside Excel. Any AI-assisted model going to clients or the board needs a second person checking the numbers.\n\n- **Presentation-building AI moved into the mainstream: Microsoft folded PowerPoint creation into a redesigned Copilot on 25 September, and Gamma launched an enterprise tier.** Templafy reports median pitch-deck time falling from two hours to eight minutes, but independent tests still find charts turned into flat images and overflowing text, and Gamma's own creative lead admits much of the category produces \"AI slop\". Budget for design and fact review, because a faster deck is not necessarily a better one.\n\n- **A US military analyst's AI turned an intelligence summary into a fabricated nuclear-weapons claim, which was caught minutes before armed personnel boarded a ship.** The output looked like an authoritative report and moved through command channels without challenge. Any AI-generated briefing that feeds an executive decision needs a named person who checks the sources as well as reading the summary.\n\n- **A Procter & Gamble experiment with 791 R&D staff found AI improved idea quality but cut how often teams picked the best option, from 50% to 37%.** Other research this fortnight points the same way: AI widens the options but weakens the choice between them. Use it to generate options and keep the final choice with people.\n\n- **DeepL ran live AI voice interpretation across more than 1,000 sessions at Salesforce's Dreamforce conference.** A separate study of 71,262 translated segments found that AI followed by human editing beat AI alone, and that severe errors clustered in medical-device content. Live interpretation is ready for events and internal meetings. Regulated or legal content still needs a human editor."
    },
    {
      "id": "coming-up",
      "title": "Coming Up",
      "body": "- **Email AI is becoming agentic (acting on its own without being prompted): Microsoft's Copilot in Outlook now sorts mail and reschedules meetings by itself.** Google's own developer guidance tells builders to let AI draft email but not send it, and 32% of US consumers already let AI act without approval. Set a default rule that AI drafts and people send before this reaches your staff's inboxes.\n\n- **Rules on who answers for AI output are hardening, and they point to a named human.** Since 2 August, EU AI Act transparency duties exempt AI-generated content only where a named person takes editorial responsibility, and NATO now disqualifies job applications prepared with AI. Decide who signs off AI-generated material you publish or submit, and expect clients' requests for proposals to start asking.\n\n- **Microsoft now has 30 million paid Copilot seats, and Gartner says firms are struggling to realize the promised gains and cost savings.** One adoption review found 64% of Copilot licenses go unused, which it put down to poor workflow integration and training rather than product quality. At your next renewal, tie seat counts to measured use and named workflows, not headcount."
    },
    {
      "id": "whats-hard-about-this",
      "title": "What's Hard About This",
      "body": "- **Time saved on individual tasks gets absorbed before it reaches the ledger.** Deloitte finds 31% of UK workers save no time at all, and the minutes others save go into review, rework, extra volume and approval queues, which Bain calls the micro-productivity trap. Capturing the value means taking steps out of the work itself, which no tool does for you.\n\n- **Heavy use weakens the judgment that is supposed to catch AI's mistakes.** In a BCG survey, 90% of senior leaders said employees have stopped checking AI's work, and a Harvard Business Review study found easier AI access reduced what people remembered. The human-in-the-loop (a person who reviews each AI output) only protects you if that person is still paying attention.\n\n- **AI-assisted work carries a social penalty that better tools will not remove.** A Duke study of 4,439 people published in PNAS found that people who use AI writing help are rated lazier and less competent than peers producing identical work. Your disclosure policies and the way managers reward work need to account for that, or staff will simply hide their use."
    }
  ],
  "url": "https://www.thestateofplay.ai/domain/personal-effectiveness",
  "license": "CC BY 4.0",
  "licenseUrl": "https://creativecommons.org/licenses/by/4.0/",
  "generatedAt": "2026-10-01"
}