🔍 Read the full analysis: AI And Productivity: Questions Worth Asking on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Thorsten Meyer’s Nov. 12 article revisits the history of productivity measurement as a way to assess AI’s economic impact. It reports substantial gains in one study of customer-support agents, but says executives estimate AI has added about 0.2 percentage points a year to productivity growth so far; the broader effect remains uncertain.
AI’s economy-wide productivity effect remains modest in current estimates, even as one study cited by Thorsten Meyer found a 15% productivity gain among customer-support agents using the technology. In an article published Nov. 12, Meyer compares those early workplace results with executives’ estimate that AI has contributed about 0.2 percentage points a year to productivity growth so far, and with past waves of technology that took time to appear in national statistics.
Meyer defines labor productivity as output per hour worked and argues that its long-run growth rate shapes how quickly living standards can rise. In his account, productivity growing at 2% annually doubles in about 35 years; at 1%, doubling takes about 70 years. The calculation illustrates why a seemingly small difference in annual growth can matter across generations, though it is not a forecast for any particular country or household.
The article draws on economist Robert Solow’s 1957 analysis of U.S. output per man-hour from 1909 to 1949. Solow attributed 87.5% of the increase to technical change and 12.5% to increased capital use, such as more machinery and buildings per worker. Meyer notes that the technical-change portion included more than equipment: better methods, organization and skills can also raise output.
For today’s AI debate, Meyer contrasts reported gains within individual workplaces with a broader estimate. A study mentioned in the article found 15% higher productivity for support agents, while executives put AI’s contribution to productivity growth at about 0.2 percentage points annually so far. The article does not identify the study or executives in the provided material, so the figures should be read as reported estimates, not as a settled measure of AI’s overall effect.
Infographic 01 · The sector clock
Productivity moves the sector clock
US farms produced nearly three times their 1948 output in 2023, with only about 9% more inputs in total. Over the same decades, farming's share of US jobs fell from about one in eight to 1.5%.
- Field12.1%
- Factory29.9%
- Services58.0%
Lebergott · own calculation
- Field1.5%
- Factory18.9%
- Services79.6%
ILO estimate
Farm output +210% (1948-2023, 1.52% a year); all inputs +0.12% a year (about +9% in total); total factor productivity +1.40% a year (USDA ERS). 1950: services = remainder, incl. unemployed.
Sources: USDA ERS, Agricultural Productivity in the U.S. (updated 25 Aug 2026); US employment shares: S. Lebergott, in US Census Bureau, Historical Statistics of the United States (1975; 1950, own calculation, see episode 2); ILO modelled estimates via World Bank (2025).
Powered by Thorsten Meyer AI
Why Economy-Wide Gains Matter
The gap between a productivity gain at one firm and a measurable rise across an economy matters because the two figures describe different things. A tool may help a particular team complete more work, but a national productivity measure also reflects how widely the tool is adopted, whether it changes processes, and whether the additional output is captured in economic statistics.
Meyer’s historical comparison cautions against assuming that a new technology will quickly lift national growth. The article says U.S. nonfarm business productivity grew about 2.8% annually from 1948 to 1973, then about 1.4% from 1973 to 1995, even as computers spread. It later rose to roughly 3.1% annually from 1995 to 2004. Those periods show that the timing and size of technology’s economy-wide impact can differ from the first visible workplace applications.
For workers and businesses, the distinction affects expectations about whether AI will translate into higher output, changes to jobs or broader economic growth. The figures cited here do not establish how benefits are distributed, whether gains persist, or what happens to employment. They do support a more measured question: whether gains reported in particular tasks spread across organizations and eventually show up in productivity data.
Products Worth Considering
Logitech MX Mechanical Wireless Illuminated Keyboard, Tactile Quiet, Graphite
LG 27US500-W 27-Inch 4K UHD IPS Monitor
Vari Ergo Electric Standing Desk, 54×26, Light Wood/White
As an affiliate, we earn on qualifying purchases.
Computers and the Solow Paradox
In a 1987 book review, Solow observed that the computer age seemed visible everywhere except in productivity statistics. The comment became known as the Solow paradox. Meyer places it against the slowdown in U.S. productivity growth after 1973 and the later acceleration from 1995 to 2004, when information technology was spreading through businesses.
Meyer cites a 2000 estimate by Federal Reserve economists Stephen Oliner and Daniel Sichel that information technology use and computer production accounted for about two-thirds of a one-percentage-point step-up in productivity growth between the first and second halves of the 1990s. The comparison is historical, not proof that AI will follow the same path. The article also uses U.S. farming as an example of long-run productivity change: farm output in 2023 was nearly three times its 1948 level, while total inputs had risen about 9% over the period, according to the figures it attributes to the U.S. Department of Agriculture.
“You can see the computer age everywhere but in the productivity statistics.”
— Robert Solow, as quoted in a 1987 New York Times Book Review review
What AI Data Cannot Yet Show
The article’s account does not establish how the reported 0.2-percentage-point estimate was calculated, which executives supplied it, or what period it covers. It also does not name the study reporting a 15% gain among support agents or provide details about its methods, sample or comparison group. Those omissions limit what can be concluded from the figures.
It remains unclear how much AI will raise productivity across the whole economy, how quickly adoption will spread, and whether early task-level gains will endure. The cited historical computer episode offers context, but it cannot predict AI’s trajectory. Meyer’s material does not provide a current official national productivity estimate that isolates AI’s contribution.
Watch Broader Productivity Measures
The next test is whether reported workplace gains become broad, sustained changes in output per hour. Readers can follow official productivity releases and compare them over time, while keeping in mind that national statistics do not usually identify the contribution of a single technology on their own.
Further evidence would need to clarify which tasks and industries benefit, how the estimates are measured, and whether efficiency gains produce more output rather than simply changing how work is recorded. Until those details and longer-run data are available, the article’s figures point to early gains in some settings, but not a settled economy-wide result.
Key Questions
What does productivity mean in the article?
Labor productivity is output per hour worked: how much value workers produce in an hour.
How much has AI raised productivity so far?
Meyer reports an estimate from executives of about 0.2 percentage points a year of productivity growth so far. The article material does not identify the executives or explain the estimate’s method and time window.
Does the 15% figure describe the whole economy?
No. It refers to a gain reported in a study of customer-support agents. A result in one work setting is not the same as an economy-wide productivity increase.
What was the Solow paradox?
It was Solow’s observation that computers appeared widely in workplaces while productivity statistics had not yet shown a corresponding broad acceleration. U.S. productivity later rose in the period Meyer cites, but that history does not guarantee AI will have the same effect.
Source: ThorstenMeyerAI.com
