Skip to content

Chapter 21

What comes next, and what to do on Monday


The chapter in a minute

  • Every trend here carries a label (measured, forecast or estimate), with who said it and when. Check the date before you trust one, and ask who made a prediction and what they assumed.
  • The Ladder Loop, the three ways a system can act and your checklist card all work on tools that don't exist yet.
  • Read one trusted source once a month, and let the other headlines go by. The dated details are on the companion page, each with the date it was checked.
  • One small loop task a day, checked in minutes, is a good way to start on Monday. Save what worked each evening.
  • Your next rung is your choice: Books 2 to 5 in any order, or Book 6 or 7 if it fits your work.

Try it now

Your Monday plan

0 of 11 steps done

Step 1 of 11

Write down three ordinary jobs waiting for you this week, like Yusuf's inbox, form and rota.

Progress is saved in this browser only.

Try this

Try this, from the book

Choose one trusted source, and keep the companion page to hand for the dated details. Set a reminder on your phone or calendar for the same day each month to read your source, and let the rest go by. Pick the day now, while it's on your mind! To undo it, delete the reminder.

Tools and settings, as of a date

The book points to the ‘What's new’ section of the companion page. Here it is in full, with the date each detail was checked. Open it on the companion page

Chapter 21 points here: "the dated figures behind each trend and the date each was checked". Read again on 2 November 2026.

The book teaches kinds of tool and a routine, and leaves product names and the latest figures here, so it can outlast the tools in it. Use this section once a month: read what's new in your trusted source, then compare it with the dated details below. Each trend is marked as measured, estimated or forecast.

Data centre electricity (forecast)

  • IEA, April 2026: demand "grew by 17% in 2025", with use of 485 TWh in 2025 "roughly doubling" to 950 TWh in 2030 in the central projection, about 3% of the world's electricity. The IEA says it "will continue to update its projections regularly." Page 17 of the same report puts 2025 use at "slightly more than 1.5%" of world electricity.
  • IEA, April 2025: about 415 TWh in 2024 (about 1.5% of world electricity). The Base Case reached 945 TWh by 2030, "just under 3%". By 2035 the range across four cases spans 700 to 1,700 TWh. The four cases are the Base Case, Lift-Off (faster uptake), High Efficiency (stronger efficiency gains) and Headwinds (slower uptake, bottlenecks).
  • No newer IEA release was listed on 2 November 2026.

Exposure of jobs (estimate)

ILO, May 2025: one in four jobs (24%) has some exposure, one in three in high-income countries (exposure is 11% of employment in low-income countries and 34% in high-income countries). The 2025 figures are lower than the 2023 ones. The ILO says it is not a forecast of jobs lost.

Productivity (forecast)

OECD economists, June 2025 (OECD Artificial Intelligence Papers, No. 41), read through the abstract and the authors' own summary because the OECD site refused automated access: across slow, medium and fast adoption scenarios, AI adds 0.4 to 1.3 percentage points a year to labour productivity growth in high-exposure countries such as the US and UK. Gains elsewhere in the G7 are "up to 50% smaller". Italy and Japan 0.2 to 0.8 points. There is no figure for Ireland. A working paper, not peer reviewed. The AI Index (page 220) sets it beside a survey of 6,000 executives in four countries that found little realised gain so far.

Agents (measured)

Stanford HAI, 2026 AI Index: agent success on OSWorld, a benchmark of real computer tasks, rose from about 12% to about 66%, and agents "still fail roughly 1 in 3 attempts on structured benchmarks". Documented AI incidents rose to 362, up from 233 in 2024. Agent deployment "remains in single digits across nearly all business functions" (paraphrased in part). Shared with "Agent options" and "Agent safety".

Expert and public forecasts (forecast)

Forecasting Research Institute, as reported in the 2026 AI Index (pages 373 and 374): across 68 forecasts, "experts consistently predicted much faster AI progress than the public". The median expert expects generative AI to assist 18% of US work hours by 2030; the public expects 10%. US only. Read only as reported in the AI Index; the institute's own report was not opened.

The "jagged frontier" example

The 2026 AI Index puts it this way: "AI models can win a gold medal at the International Mathematical Olympiad but cannot reliably tell time". Its example: Gemini Deep Think earned a gold medal at the 2025 maths Olympiad, "yet the top model reads analog clocks correctly just 50.1% of the time" (top takeaways, page 9 of the report PDF (opens in a new tab)). The technical chapter (PDF page 96) gives the detail from ClockBench, a test of 720 questions on 180 clock faces: people read the clocks correctly 90.1% of the time, while the top model, GPT-5.4 High, reached 50.6% in March 2026, and its wrong answers were typically one to three hours out, against three minutes for people. The report gives 50.1% in one place and 50.6% in the other; the 50.6% is the one tied to a named test and date. Read on 2 November 2026.

The next AI Index

The next edition is expected around April 2027. When it is out, the agent and incident figures above will be updated from it.

Checked: 2 November 2026.

The one-week plan page

A blank copy of Table 21.1 (the one-week plan) is in the companion PDF. See "The companion PDF".

Sources

  1. Macroeconomic productivity gains from Artificial Intelligence in G7 economies (opens in a new tab) · Filippucci, Francesco and Gal, Peter and Laengle, Katharina and Schief, Matthias, 2025
  2. Generative AI and Jobs: A Refined Global Index of Occupational Exposure (opens in a new tab) · Gmyrek, Pawe\l and Berg, Janine and Kami\'nski, Karol and Konopczy\'nski, Filip and \Ladna, Agnieszka and Nafradi, Balint and Ros\laniec, Konrad and Troszy\'nski, Marek, 2025
  3. The 2026 AI Index Report (opens in a new tab) · Stanford Institute for Human-Centered Artificial Intelligence, 2026
  4. Energy and AI (opens in a new tab) · International Energy Agency, 2025
  5. Key Questions on Energy and AI (opens in a new tab) · International Energy Agency, 2026