AI at work: 8 citation-ready data points for creator briefs
A source-first briefing for creators building an AI-at-work argument: eight usable data points, each paired with its source date, methodology note, citation wording and chart angle. The issue separates employee usage, enterprise deployment, executive strategy and public sentiment so numbers are not mixed across incompatible definitions.
AI adoption is no longer a single clean number. One source counts any work touched by AI, another counts a few uses per year, another measures firm deployment, and platform data measures observed usage rather than survey answers. That is why the safest creator brief should cite the measurement frame, not just the percentage.
Below are eight numbers you can use for scripts, charts, newsletters, decks, and explainers on AI at work. Treat them as separate evidence cards, not interchangeable claims.
Quick scan
| Use this when you want to argue... | Best data point |
|---|---|
| AI work use is growing, but not yet a worker majority | Pew: 21% of U.S. workers say at least some of their work is done with AI |
| A broader definition produces a much higher adoption figure | Gallup: 40% of U.S. employees have used AI in their role a few times a year or more |
| AI adoption numbers vary because surveys ask different questions | Federal Reserve review: worker surveys cluster around 20% to 40% |
| Some organizations are past pilots | Microsoft: 24% of leaders say their companies have deployed AI organization-wide |
| AI is now a workforce strategy, not only a tool rollout | Microsoft: 47% prioritize AI skilling; 45% prioritize digital labor; 33% consider headcount reduction |
| Enterprise API usage leans toward automation | Anthropic: 77% of business API uses show automation patterns |
| Adoption is geographically uneven | Anthropic: Israel's Claude usage index is 7; Singapore's is 4.57 |
| Experts and the public disagree sharply on AI's work impact | Pew: 73% of AI experts vs. 23% of U.S. adults expect a positive impact on how people do jobs |
The cards
1. Worker adoption, narrow definition: 21% of U.S. workers use AI in their job
Data point. In a September 2025 Pew Research Center survey, 21% of U.S. workers said at least some of their work is done with AI, up from 16% roughly a year earlier. Most workers, 65%, still said they do not use AI much or at all in their job. 1
Source date. October 6, 2025.
Methodology note. Pew based the analysis on 5,010 U.S. workers inside a nationally representative survey of 8,750 U.S. adults conducted September 2-8, 2025 through the American Trends Panel. 1
What it can support. Use this for a cautious, mainstream claim: AI is spreading through work, but most workers still report little or no direct use.
Chart suggestion. A two-point slope chart works well: 16% in 2024 to 21% in September 2025. 1
Citation format. "Pew Research Center found that 21% of U.S. workers said at least some of their work was done with AI in September 2025, up from 16% roughly a year earlier." 1
2. Worker adoption, broad definition: 40% of U.S. employees have used AI in their role
Data point. Gallup reported that 40% of U.S. employees said they had used AI in their role a few times a year or more, nearly double the 21% level Gallup measured in 2023. Frequent use, defined as a few times a week or more, rose from 11% to 19% over the same period. 2
Source date. June 16, 2025.
Methodology note. Gallup's quarterly workforce studies are self-administered web surveys of U.S. adults working full time or part time for organizations; respondents come from the probability-based Gallup Panel, and Gallup weights samples for nonresponse and national demographics. 2
What it can support. Use this when the argument is about exposure or occasional use, not daily reliance. It is broader than Pew's "some of their work is done with AI" wording.
Chart suggestion. Use a grouped bar chart with three definitions: "any use a few times a year or more" at 40%, "frequent use" at 19%, and "daily use" at 8%. 2
Citation format. "Gallup reported that 40% of U.S. employees had used AI in their role at least a few times a year, while 19% used it a few times a week or more." 2
3. Measurement warning: workplace AI estimates often land between 20% and 40%
Data point. A Federal Reserve FEDS Note reviewed 16 surveys on AI adoption and found worker-level surveys generally placed workplace AI use in the 20% to 40% range. Firm-level surveys ranged from 5% to about 40%, depending on weighting, wording, lookback period, and whether the question listed specific AI applications. 3
Source date. February 5, 2025.
Methodology note. The authors reviewed surveys from government agencies, NGOs, academics, and private companies, mostly fielded from late 2023 to mid-2024, and recorded respondent count, survey timing, firm vs. worker level, and whether the survey measured AI broadly or generative AI specifically. 3
What it can support. Use this as a footnote whenever a script compares adoption numbers from different sources. It supports the argument that question design can move the headline number.
Chart suggestion. A range plot is better than a single bar: one band for firm surveys at 5% to about 40%, another for worker surveys at 20% to 40%. 3
Citation format. "A Federal Reserve review of 16 surveys found that worker-level estimates of AI use at work typically fell between 20% and 40%, while firm-level estimates varied more widely." 3
4. Organizational deployment: 24% of leaders say AI is deployed organization-wide
Data point. Microsoft's 2025 Work Trend Index says 24% of leaders reported that their companies had already deployed AI organization-wide, while 12% said their companies remained in pilot mode. 4
Source date. April 23, 2025.
Methodology note. Microsoft analyzed survey data from 31,000 workers across 31 countries, LinkedIn labor-market trends, and Microsoft 365 productivity signals; the report also draws on interviews with AI-native startups, academics, economists, scientists, and other specialists. 4
What it can support. Use this for the move from experimentation to rollout. It is a leader-reported organization measure, not a representative worker-use measure.
Chart suggestion. A small deployment funnel: pilot mode at 12%, organization-wide deployment at 24%, and expected agent integration in the next 12-18 months at 81%. 4
Citation format. "Microsoft's 2025 Work Trend Index reported that 24% of surveyed leaders said their companies had deployed AI organization-wide, compared with 12% still in pilot mode." 4
Open next: Microsoft 2025 Work Trend Index.
5. Workforce strategy: 47% upskilling, 45% digital labor, 33% headcount reduction
Data point. In Microsoft's Work Trend Index, leaders' top AI workforce strategies for the next 12-18 months included AI-specific upskilling of the existing workforce (47%), maintaining headcount while using AI as digital labor (45%), and using AI to reduce headcount (33%). 4
Source date. April 23, 2025.
Methodology note. This comes from the same Microsoft survey and signal mix described above: 31,000 workers across 31 countries, LinkedIn labor-market trends, and Microsoft 365 productivity signals. 4
What it can support. Use this when the story is about strategic trade-offs. The data supports a more precise claim than "AI will replace jobs": leaders are considering several workforce paths at once.
Chart suggestion. A horizontal bar chart with workforce strategies sorted descending: 47% upskilling, 45% digital labor, 44% morale, 33% headcount reduction. 4
Citation format. "Microsoft found that 47% of surveyed leaders prioritized AI-specific upskilling, 45% prioritized digital labor while maintaining headcount, and 33% were considering AI-related headcount reductions." 4
6. Enterprise API use: 77% of business API usage is automation-dominant
Data point. Anthropic's September 2025 Economic Index reported that 77% of business uses in its first-party API traffic involved automation usage patterns, compared with about 50% for Claude.ai users. 5
Source date. September 15, 2025.
Methodology note. Anthropic extended its Economic Index with geographic analysis of Claude.ai conversations and a separate analysis of first-party API traffic; API users access Claude programmatically, which is why the report treats the API cut as a window into business deployment. 5
What it can support. Use this to distinguish consumer chat behavior from enterprise systems. API-based use appears more automated than web-chat use.
Chart suggestion. A paired bar chart: first-party API automation at 77% vs. Claude.ai automation at about 50%. 5
Citation format. "Anthropic's Economic Index found that 77% of business uses in first-party API traffic showed automation patterns, compared with about 50% for Claude.ai users." 5
Open next: Anthropic Economic Index, September 2025 report.
7. Geographic concentration: Israel indexes at 7, Singapore at 4.57
Data point. Anthropic's AI Usage Index compares a country's share of Claude usage with its share of working-age population. Israel led the report's per-capita ranking with an index of 7, followed by Singapore at 4.57, Australia at 4.10, New Zealand at 4.05, and South Korea at 3.73. 5
Source date. September 15, 2025.
Methodology note. The geographic analysis used privacy-preserving analysis of 1 million Claude.ai conversations across 150+ countries and all U.S. states, then adjusted usage by working-age population share. 5
What it can support. Use this for the inequality angle: AI adoption can be high globally while still concentrated in high-income or tech-heavy regions.
Chart suggestion. A ranked lollipop chart of the top five AI Usage Index values: 7, 4.57, 4.10, 4.05, 3.73. 5
Citation format. "Anthropic's AI Usage Index ranked Israel at 7 and Singapore at 4.57, meaning Claude usage was far above what their working-age population shares alone would predict." 5
8. Public vs. expert expectations: 73% vs. 23% on positive work impact
Data point. Pew found that 73% of surveyed AI experts expected AI to have a very or somewhat positive impact on how people do their jobs over the next 20 years. Only 23% of U.S. adults said the same. 6
Source date. April 3, 2025.
Methodology note. Pew surveyed 5,410 U.S. adults from August 12-18, 2024 and 1,013 U.S.-based AI experts from August 14 to October 31, 2024. The public survey was weighted to represent U.S. adults; expert responses were unweighted and representative only of experts who responded. 6
What it can support. Use this for audience-framing claims: expert optimism does not equal public confidence.
Chart suggestion. A diverging bar chart or simple split bar: 73% of AI experts vs. 23% of U.S. adults expecting a positive impact on how people do jobs. 6
Citation format. "Pew Research Center found a 50-point gap between AI experts and U.S. adults on whether AI will positively affect how people do their jobs over the next 20 years: 73% vs. 23%." 6
Open next: Pew's public-and-expert AI survey report.
How to avoid misusing these numbers
Do not merge the 21% Pew figure and the 40% Gallup figure into a single trend line. They measure different things: Pew asks whether at least some work is done with AI, while Gallup counts employees who have used AI in their role a few times a year or more. 12
Do not treat survey adoption and platform-usage telemetry as the same evidence. Surveys tell you what respondents say they do. Anthropic's Claude data tells you what one provider observes inside its own product and API traffic. 5
The cleanest use is to pair each number with the claim it can actually support. If a sentence says "workers use AI," cite Pew or Gallup. If it says "organizations are deploying AI," cite Microsoft. If it says "usage is geographically uneven" or "API usage is automation-heavy," cite Anthropic.
References
- 121% of US workers use AI on the job, up since 2024
pewresearch.org
- 2
- 3Measuring AI Uptake in the Workplace
federalreserve.gov
- 42025: The year the Frontier Firm is born
microsoft.com
- 5
- 6
This story was produced automatically by a channel. One sentence is all it takes for Neodrop to keep producing for you.