🔍 Read the full analysis: Inside The 80% Service Surge In AI on ThorstenMeyerAI.com
Get the little things that make your day 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
Thorsten Meyer’s Nov. 9 report connects the large U.S. service workforce to Baumol’s cost disease, in which wages rise even when some services are hard to make more productive. It cites a study where an AI assistant helped 5,172 customer-support agents resolve 15% more issues per hour on average, but whether such gains will reduce staffing or shift work is unclear.
A study of 5,172 customer-support agents found that an AI assistant increased the number of issues they resolved per hour by 15% on average, according to Thorsten Meyer’s Nov. 9 report on AI and the service economy. The result offers a measured example of AI improving productivity in work often delivered through human interaction, but it does not show whether companies will cut jobs or how broadly the gain will apply.
Meyer frames the result through Baumol’s cost disease, an economic idea about work that is difficult to make faster without changing the service itself. A string quartet still requires four musicians to perform a piece that takes about 45 minutes; improving factory productivity, by contrast, has historically allowed more goods to be made with fewer labor hours. In many services, labor is part of what customers receive, so productivity gains can be harder to achieve.
The report says services accounted for 86% of U.S. nonfarm payroll jobs in 2025, based on the author’s calculation using annual averages from the Bureau of Labor Statistics. The measure excludes farm workers and the self-employed. Separately, International Labour Organization estimates put services at 79.6% of all U.S. workers in 2025. These figures use different definitions and should not be treated as interchangeable.
Price comparisons in the report illustrate the uneven pattern. Since 2000, the official quality-adjusted price of televisions fell 98%, while hospital services rose 274%, college tuition 188% and car repair 143%. Those figures do not establish a single cause for rising service prices. Meyer also notes that haircuts and car repairs became more expensive, suggesting that health insurance and student loans alone do not explain the trend.
Infographic 02 · Baumol's cost disease
Why the workers moved to services
Baumol's mechanism in four steps, and what it looks like in US jobs.
- 1966Baumol and Bowen: the quartetA 45-minute Schubert quartet still takes about three hours of human work
- 1967Two kinds of workWhere labour is a tool, machines cut it again and again; where labour is the product, they can’t
- WagesPay rises everywhereTeachers, nurses and musicians must earn more too, or they leave
- ResultCosts rise, people moveIf we keep buying the same mix, more and more workers end up in the slow sector
Service industries, share of US nonfarm payroll jobs
US jobs, annual average 2025 (millions)
Payroll shares exclude farm workers and the self-employed (BLS establishment survey). Health spending: national health expenditure as a share of gross domestic product.
Sources: W. J. Baumol, “Macroeconomics of Unbalanced Growth”, American Economic Review 57(3), 1967, BLS Current Employment Statistics (own calculation of shares), CMS, National Health Expenditure, Table 1.
Powered by Thorsten Meyer AI
AI Tests the Limits of Service Productivity
The support-agent study matters because it documents a productivity gain in a task-based service setting, where AI can assist workers without necessarily replacing the entire interaction. If similar tools help staff handle more cases in the same amount of time, employers could reduce waiting times, serve more customers or lower labor costs. The study result alone does not establish which outcome will follow.
That distinction is central to the broader labor question. AI may take over some tasks, increase the output expected from each worker, or create demand for new forms of work. A reported 15% average gain in one study is not evidence of an equivalent reduction in staffing, nor does it show that all service jobs can be automated. The outcome will depend on the type of work, the quality employers require and how businesses use the extra capacity.
The stakes extend beyond customer support. Services account for a large share of U.S. employment, and many are costly precisely because they require time and personal attention. A technology that changes the amount of labor needed could affect prices, access and jobs. But the material presented here does not yet show economy-wide effects on any of those measures.
Products Worth Considering
Hyperice Hypervolt 2 Handheld Percussion Massage Gun
Withings Sleep – Sleep Tracking Pad Under the Mattress
LEVOIT Vital 200S-P Smart Air Purifier for Large Rooms
As an affiliate, we earn on qualifying purchases.
From Schubert to the Service Economy
Economists William Baumol and William Bowen examined the economics of performing arts in their 1966 book, Performing Arts: The Economic Dilemma. Their example of the Schubert quartet captured a productivity problem: the same live performance takes roughly the same combined musician-hours, even as wages elsewhere rise. Baumol developed the idea in a 1967 paper, distinguishing activities where technology can steadily increase output per labor hour from work where labor is closely tied to the service itself.
The distinction is not absolute. Retail changed through supermarkets and self-service, while technology made some kinds of travel more efficient. But services have grown for decades, not only since manufacturing employment began to decline. Meyer’s BLS series puts services at 62% of nonfarm payroll jobs in 1939, 72% in 1979 and 86% in 2025. The report’s figures show that the employment shift was already underway well before the U.S. manufacturing job peak in June 1979.
Baumol’s explanation is that wages in less productive fields tend to rise alongside wages in more productive ones; otherwise, workers could leave for better-paid jobs. When output per worker does not rise at the same pace, labor costs can take up a growing share of the cost of providing a service. That theory helps explain a pressure on costs, but it does not account for every price increase. Meyer contrasts U.S. health spending, at 18.0% of the economy, with Germany’s 12.4%, showing that the same broad economic forces can operate alongside different national spending levels.
“A half hour horn quintet calls for the expenditure of 2½ man hours in its performance.”
— William Baumol, quoted from his 1967 paper
What the 15% Gain Cannot Tell Us
The supplied account does not identify the customer-support study’s authors, publication date, research setting or exact definition of an issue resolved. It reports the sample size and average productivity result, but does not say how outcomes varied between agents, how the AI was deployed or whether the result persisted over time. Those details are needed to judge how readily the finding applies to other workplaces.
It is also unclear whether the productivity gain changed staffing, wages, customer satisfaction or service quality. More issues resolved per hour does not by itself establish that customers received better service or that fewer workers were needed. The available material gives no economy-wide estimate of AI’s effect on service employment, prices or the cost pressures described by Baumol’s theory.
The employment percentages also describe different populations: the BLS figure concerns nonfarm payroll jobs, while the ILO estimate covers all workers. The report’s price comparisons do not isolate the causes of each change, and the contrast between U.S. and German health spending does not explain the difference by itself.
Evidence Needed Beyond Customer Support
The next test is whether the reported result can be reproduced in other customer-support settings and in other services, including fields where personal judgment, physical care or teaching is central. More details about the study’s methods and setting would help readers assess what the 15% figure measures and whether it applies beyond the agents examined.
For employers and workers, the key developments to watch are how organizations use productivity gains: to handle more demand, shorten response times, change job duties or reduce staffing. Those choices will show whether AI complements service workers or substitutes for some of their work. The report does not provide a timeline for broader adoption or predict the balance of those effects.
At a wider level, future evidence would need to track output, costs, service quality and employment together. Until those measures are available, the support-agent study is a specific productivity finding, not proof that AI will resolve the service sector’s rising costs or displace its workforce.
Key Questions
What did the customer-support AI study report?
As summarized by Thorsten Meyer, an AI assistant increased the number of issues resolved per hour by 15% on average among 5,172 customer-support agents. The supplied material does not provide the study’s authors or full methods.
Does the study show that AI will eliminate service jobs?
No. The result measures issues resolved per hour, not job losses. The report does not say whether employers reduced staffing or how the productivity gain affected workers.
What is Baumol’s cost disease?
It describes a pressure on costs in work where productivity is difficult to increase. If wages rise alongside those in more productive industries while output per worker changes little, providing the service can become more expensive.
How large is the U.S. service workforce?
The report gives two measures for 2025: services made up 86% of nonfarm payroll jobs, using the author’s BLS calculation, and 79.6% of all workers, according to ILO estimates. The measures have different coverage.
What remains unknown about AI’s effect on services?
The material does not establish whether productivity gains will reduce prices, improve service quality, change wages or cut jobs. It also does not show whether the customer-support result generalizes to other service industries.
Source: ThorstenMeyerAI.com
