This is the 2026-08 edition, checked 2026-09-07. The data retains its actual observation period. Quarterly estimates and annual surveys do not become measurements of the edition month.
AI diffusion · provider estimate
Working-age population, including people offline. Q1 2026.
147 / 219 economies with eligible evidence · Microsoft AI diffusion · 2026-08.r2
Ranks are among 147 economies in the same source and observation period. Older evidence has no rank.
Country / economy
Value
Actual period
Compare
70.1%
Q1 2026
63.4%
Q1 2026
48.6%
Q1 2026
48.4%
Q1 2026
47.8%
Q1 2026
44.2%
Q1 2026
43%
Q1 2026
42.2%
Q1 2026
42.1%
Q1 2026
41.8%
Q1 2026
39.5%
Q1 2026
39%
Q1 2026
38.1%
Q1 2026
37.8%
Q1 2026
37.3%
Q1 2026
Country comparison · AI diffusion · provider estimate
Choose 2–5 economies. Add another country from the map panel or table.
Economy
AI diffusion · provider estimate
Observation period
Evidence
Korea, Dem. People's Rep.
Not measured
Not measured
No verified, publicly reusable country observation in this edition.
Working-age population, including people offline · Microsoft AI diffusion · 2026-08.r2. No interpolation or replacement with a regional value.
Comparable evidence
Countries in the same measurement lane.
Q1 2026 · Microsoft AI diffusion. Ranked only among covered economies.
Economy
AI diffusion · provider estimate
Evidence
Compare
70.1%
63.4%
48.6%
48.4%
47.8%
44.2%
43%
42.2%
42.1%
41.8%
Top eligible rows in the current region/search. Same metric, measurement profile and source period. 2026-08.r2.
Economic evidence / native study periods
What happens when people use AI?
Different tasks, different results. These studies are outside the country ranking.
Randomized field experiment
Consulting tasks and the limits of assistance
Over 25% faster on tasks within GPT-4’s capability
758 BCG consultants completing realistic knowledge-work tasks. The reported speed gain applies to the tasks within the tested model’s capabilities.
Working paper reported 21 September 2023 · journal publication March 2026
Performance depended on the task; assistance could worsen results beyond the model’s capabilities. This is historical, task-specific evidence, not a national productivity estimate or a claim about today’s models.
16 experienced open-source developers; 246 issues in repositories they knew.
Early 2025 experiment · published July 2025
This setting does not represent all developers or current tools. METR’s February 2026 follow-up identifies selection effects and redesigns the study; do not extrapolate this effect to countries.
Historical backfill retrieved 7 September 2026: 6,588 published World Bank indicator values across 217 economies for available 2023–2025 annual waves, plus 67 enterprise AI survey records from 2023–2024 in a separate seven-category measure. The history view compares native periods and adds the September 2023 BCG consulting experiment, with its March 2026 journal publication. The original 2026-08.r1 snapshot remains unchanged.
Why it may matter
The three-year window is September 2023–August 2026. Complete annual waves overlapping this window retain their original full year. Source definitions, attribution, retrieval dates and fingerprints accompany the history. These are current source vintages of historical observations, not reconstructed atlas scores or what was known at each past date.
What remains uncertain
Coverage differs by indicator and country. Water-stress observations in this window are not published in the connected feed; 2026 annual waves are incomplete and excluded. Microsoft country AI diffusion starts in H1 2025; no 2023–2024 values are inferred. The expanded 2025 enterprise AI basket is separate from 2023–2024, and no cross-basket growth rate is calculated. Comparable country attention and realized net economic benefit remain unavailable.
Take the evidence with you.
Downloads preserve the active region/search, metric, source, period and release. Missing values stay empty.
Quarterly review: 1 January, 1 April, 1 July and 1 October at 08:00 Europe/Zurich, starting 1 October 2026. New editions retain native observation dates; a review does not imply new data from every source.
