🔍 Read the full analysis: Knowledge Work Meets Artificial Intelligence on ThorstenMeyerAI.com
Get hardware and tech essentials delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
A November 10 report from ThorstenMeyerAI.com traces the rise of knowledge and information work, drawing on estimates that it accounted for almost 29% of adjusted U.S. GNP in 1958 and 46.4% of the workforce in 1970. It argues that generative AI is aimed at office tasks common in knowledge jobs, while stressing that exposure is not the same as job loss and that overall U.S. employment data shows no clear AI effect so far.
ThorstenMeyerAI.com published a report on November 10 tracing how knowledge and information work grew into a major part of the economy—and why those jobs are now a focus of generative AI. The account cites historical estimates that knowledge-related industries represented almost 29% of adjusted U.S. GNP in 1958, while information workers made up 46.4% of the U.S. workforce in 1970. It also cautions that AI exposure does not establish job loss; the overall U.S. employment data cited in the report shows no clear AI footprint so far.
The report, the final episode of the first arc of the series “Work After Work,” revisits economist Fritz Machlup’s 1962 study, which estimated that the production and distribution of knowledge cost $136.4 billion in 1958, or “almost 29 per cent of adjusted GNP.” Machlup counted spending across education, research and development, media, information machines and information services. The figure measures costs, not the total value produced by knowledge or the contribution of each worker.
Historian of statistics Benoît Godin’s breakdown, as presented in the report, assigns 44% of Machlup’s total to education, 28% to media, 13% to information services, 8% to research and development and 7% to information machines. The report says these knowledge industries grew by 8.8% annually between 1947 and 1958, according to Machlup’s estimates.
It also draws on Marc Porat’s 1977 analysis, which classified workers by the tasks they performed rather than the industry where they worked. Porat estimated that information workers rose from 5.8% of the U.S. workforce in 1860 to 46.4% in 1970. The report gives current occupational shares for managers, professionals and technicians as 62% in Singapore, 52.5% in the United States and 47.9% in Germany, citing the International Labour Organization in 2025. Those figures use a different occupational grouping from Porat’s historical information-worker measure.
Infographic 01 · The number to remember
Fritz Machlup's first count of the knowledge industries, by branch.
US spending on knowledge by branch, 1958 · share of $136.4 bn
Machlup measured what knowledge cost (spending, the way national accounts count), not what it was worth. “Almost 29 per cent of adjusted GNP” is his figure for 1958. The branch values are as tabulated by Godin (2008); the labour shares are Machlup’s estimate as summarised by his publisher.
Sources: F. Machlup, The Production and Distribution of Knowledge in the United States (Princeton University Press, 1962), as tabulated by B. Godin, “The Knowledge Economy: Fritz Machlup's Construction of a Synthetic Concept” (2008); Princeton University Press, book description.
Powered by Thorsten Meyer AI
Why Office Work Is in Focus
Generative AI is being applied to tasks often found in knowledge jobs, including writing, reading, calculating, checking and planning. That puts a large and economically significant category of work at the center of debate about automation. The report’s historical framing is useful because it shifts attention from factories alone to the office tasks that now occupy many workers.
But a task’s exposure to AI is not a reliable count of jobs that will disappear. Employers may use the tools to change how work is done, redistribute tasks or alter hiring, with different outcomes across occupations and workplaces. The report says current U.S. job data shows no clear overall AI footprint so far, while citing evidence that younger workers in the most exposed jobs are falling behind their peers. It does not provide the underlying dataset or specify the scale of that gap in the supplied material.
The distinction matters to workers, businesses and policymakers: historical growth in knowledge work helps explain the reach of AI tools, but it cannot by itself predict the pace or distribution of employment effects. Exposure is a risk signal, not proof of displacement.
Products Worth Considering
Apple Watch SE 3 GPS 40mm Smartwatch
Samsung Galaxy Tab A11+ (6GB RAM, 128GB Storage) – Gray
Logitech MX Ergo S Advanced Wireless Trackball Mouse USB-C – Graphite
As an affiliate, we earn on qualifying purchases.
How Economists Counted Information Work
Management writer Peter Drucker described “knowledge work” in 1959, arguing that people increasingly earned a living through what they knew rather than physical labor. The report says Drucker believed factory-era management methods would not be sufficient for work centered on expertise and information.
Machlup then tried to quantify knowledge production, which cannot be counted in physical units like steel or grain. He used expenditure as a measure, covering industries that create and distribute information and occupations that produce or pass it along. His estimate therefore reflects the cost of producing knowledge, not a direct measure of its economic worth.
Porat’s later framework organized the workforce into four sectors: agriculture, industry, services and information. His task-based approach meant that a clerk working in a steel mill could count as an information worker. The report describes this as a fourth, or quaternary, sector. Its historical figures show information work’s rising share alongside declines in agriculture and, later, industry—but the categories and methods differ from modern occupational statistics.
“almost 29 per cent of adjusted GNP”
— Fritz Machlup, as quoted in the report
What AI Exposure Does Not Show
The source material does not identify which specific exposure estimates it relies on, how “higher-income and clerical jobs” were classified, or what period the estimates cover. It also does not provide the underlying data behind its statement that young workers in highly exposed jobs are falling behind their peers, including the size of the difference or whether AI caused it.
The report’s historical figures are estimates tied to particular definitions and methods, not directly comparable measures across time. Its claim that about half of jobs in rich economies are knowledge jobs is a broad characterization; the supplied occupational figures cover three countries and a particular grouping of managers, professionals and technicians. The scale, timing and employment effects of AI adoption remain unsettled.
Further Episodes and Better Data
The report identifies the next episode as a discussion of what economic statistics measure—and what they miss. That topic follows directly from Machlup’s method: counting what knowledge costs is not the same as measuring what it contributes.
For the AI question, readers will need clearer occupation-level employment data and explicit definitions of exposure, alongside evidence tracking changes over time. Until those are available, claims about broad job losses or gains should be treated cautiously. The source provides no forecast, confirmed employment total attributable to AI, or specific timetable for when effects may become clearer.
Key Questions
What is the news development?
ThorstenMeyerAI.com published an episode on November 10, 2026, connecting the historical rise of knowledge work with the kinds of office tasks generative AI may affect.
What did Machlup estimate?
Machlup estimated that U.S. spending on producing and distributing knowledge reached $136.4 billion in 1958, or almost 29% of adjusted GNP. The figure measures estimated costs, not the total value of knowledge.
Does AI exposure mean a job will be eliminated?
No. Exposure means some tasks may be affected by AI; it does not establish that an employer will remove a job. The report says overall U.S. job data shows no clear AI footprint so far.
How much of the U.S. workforce did Porat classify as information workers?
Porat estimated that information workers accounted for 46.4% of the U.S. workforce in 1970. His classification was based on what people did at work, not simply the industry that employed them.
What evidence about younger workers does the report cite?
The report says young workers in the most AI-exposed jobs are falling behind their peers, but the supplied material does not give the dataset, size of the gap or evidence that AI caused the difference.
Source: ThorstenMeyerAI.com
