What Artificial intelligence (A.I.) Is Actually Doing to the Economy
Artificial intelligence is already changing how work is done. Yet a basic question remains remarkably difficult to answer: what is A.I. actually doing to the economy right now?
Some data suggest A.I. is costing people their jobs. Other data show that companies adopting A.I. are hiring faster than their peers. Economists, policymakers and workers are all trying to make sense of these conflicting signals—and the stakes are high.
Rising Anxiety, Limited Clarity
In public discussion about A.I., one emotion emerges repeatedly: fear.
Workers describe “mixed feelings” and “love–hate relationships” with new tools. A writer reports that A.I. improves her output even as she believes it is taking her job away. Others say A.I. has already displaced them and forced a reconsideration of their careers.
Polling captures the same unease. Around 70 percent of Americans believe A.I. will lead to fewer jobs. Economists increasingly suspect that A.I. could become the defining economic story of this decade, potentially overshadowing concerns about tariffs or oil prices when future analysts look back from the 2030s.
The paradox is that anxiety is rising faster than evidence. A clear, data‑driven picture of A.I.’s impact on jobs and wages has not yet emerged.
Why the Numbers Lag Behind Reality
One reason for the uncertainty is that the economic measurement systems in use today were not designed for technologies that spread as quickly as modern A.I.
The U.S. monthly jobs report, for example, tracks how many jobs are added or lost. It does not provide a clean category for “tech,” and certainly not for “A.I.” Tech work is scattered across older industrial groupings: information (which also includes newspapers and media), professional services, manufacturing and more. There is no single line item that allows analysts to see what is happening specifically to A.I‑related jobs.
The same limitation applies to workers viewed as particularly vulnerable: recent graduates in A.I‑exposed occupations. Official data does not follow their fortunes month by month in a systematic way. The survey infrastructure and category design largely reflect an economy of decades past.
Governments cannot reasonably invent new measures for every emerging technology, so researchers have turned to private‑sector datasets. Payroll processors such as ADP, hiring platforms such as LinkedIn and Indeed, and company‑level data all now play a role in attempts to track A.I.’s effects.
These sources are valuable, but they do not tell a single, coherent story:
Some rigorous studies, based on detailed occupational and skill data, detect job losses among entry‑level workers in roles highly exposed to A.I.
Other studies, using similarly careful methods, find that firms adopting A.I. most aggressively are expanding their workforces faster than competitors.
The result is a confusing picture in which A.I. appears associated with both job destruction and job creation, depending on the lens. That ambiguity makes it difficult to draw firm conclusions about net impact.
A.I. as Corporate Scapegoat
Corporate communication adds another layer of complexity. Firms increasingly cite A.I. when announcing layoffs and restructuring.
Amazon, for instance, has cut tens of thousands of jobs while referencing A.I. as part of the rationale. Block, the payments company, has also used A.I. language in explaining workforce reductions. At the same time, investors reward companies that describe ambitious A.I. strategies; their stock prices rise and venture capital flows more readily.hbr+1
This environment creates clear incentives. Executives who believe their organizations overhired during boom years, or who face business slowdowns, may prefer to attribute cuts to “A.I.‑driven productivity” rather than to misjudged demand or management errors.
Economists therefore approach such claims with skepticism. In some cases A.I. is genuinely one factor among many driving restructuring. In others, it serves as a convenient narrative, rebranding ordinary cost‑cutting as technological progress.
This mix of genuine technological change and strategic storytelling further muddies attempts to isolate A.I.’s true economic impact.
Two Emerging Facts
Despite the noise, two broad observations are increasingly accepted.
First, A.I. is spreading very quickly. The technology itself is improving at a rapid pace, and adoption is expanding across sectors far beyond traditional tech. Marketing teams, customer service centers, law firms, media organizations and manufacturers are all experimenting with generative A.I. tools.
A recent statement signed by about 200 economists warned that A.I. could trigger an economic transformation larger than the Industrial Revolution, but compressed into a much shorter period. If that assessment proves even partially correct, the current moment marks the early phase of a historic shift.
Second, measurable labor‑market effects remain subtle. Unemployment, in aggregate, is not spiking. Official data do not yet show the kind of large, concentrated job losses that would be expected if A.I. were already wiping out entire sectors.
Taken together, these points suggest an economy in transition: technology moving ahead of measurement, and early impacts that are real but not yet dramatic at the macro level.
The J‑Curve of Technological Change
Economists often describe the rollout of major technologies using a “J‑curve.”
When a transformative technology arrives, productivity frequently falls before it rises. Initially, organizations experiment; workers learn new tools, workflows are redesigned, systems are rebuilt. Mistakes are common. The eventual gains appear only once the technology is embedded and processes are optimized.
This pattern was visible in the 1990s with the internet and more recently with video conferencing during the pandemic. Early adoption tended to be clumsy and inefficient. Over time, these tools became integral to everyday work.
A.I. appears to be in that early “scoop” at the bottom of the J. For many users, forcing a generative system to perform a task is still slower than doing it manually. Journalists, analysts and office workers often try A.I. tools, then revert to traditional methods when results disappoint.
Over the coming years, individuals are likely to become more fluent, while organizations reorganize jobs and processes around A.I. That is when productivity gains—and more visible disruptions—will begin to show up clearly in the data.
The critical questions are about pace: how quickly the economy moves up the J‑curve, who benefits, and who bears the costs.
Two Historical Precedents: Internet and China Shock
To understand possible futures, many economists look back at two different episodes of large‑scale economic disruption: the rise of the internet and the “China shock” in trade.
The Internet: Gradual, Diffuse Transformation
In the 1990s, the internet reshaped the economy. Dial‑up connections, early web portals and online communication formed the foundation for what later became smartphones, social media, e‑commerce and cloud computing.
The internet both created and eliminated jobs.
New roles emerged in software development, web design, online advertising and IT.
Older roles shrank or disappeared, including typists, travel agents and some bank clerks and customer service workers.
Yet the internet era is not remembered primarily as a period of mass unemployment. The main reason is that change was gradual and broadly spread.
Companies did not dismiss entire typing pools in a single day. Travel agents did not lose all their clients at once. Instead, technology slowly replaced certain tasks, and jobs evolved accordingly.
Workers had time to respond. Some retired at the end of their careers. Younger people could see the direction of the economy and choose different training routes. New industries grew fast enough to absorb many of those leaving declining ones.
For all its turbulence, the 1990s are often described as a decade of opportunity as well as disruption. If A.I. follows a similar trajectory, it could raise productivity and living standards while giving workers and communities time to adapt.
The China Shock: Rapid, Concentrated Dislocation
A different pattern emerged with the “China shock” of the late 1990s and early 2000s.
Deeper trade integration with China led to a surge of low‑cost imports, particularly in manufacturing. The impact was swift and geographically concentrated. Specific regions and industries experienced sharp employment losses.
Hickory, North Carolina, offers a notable example. Once a major furniture manufacturing center, the region faced intense competition from Chinese producers. Local factories closed, and tens of thousands of jobs vanished in a relatively short period.
The damage did not remain confined to factory floors:
Retailers lost customers.
Restaurants and service businesses saw demand fall.
Schools and public services faced budget strain.
Housing markets weakened, trapping residents in place.
Beyond economics, communities struggled with higher rates of addiction, social fragmentation and political discontent. Many residents felt abandoned by policymakers and the broader economy.
The China shock was fast and concentrated. Workers had little time to retrain or relocate, and alternative opportunities nearby were often scarce.
If A.I. were to eliminate large categories of work in a similarly rapid and geographically focused way, the social consequences could be profound.
Possible Paths for A.I.
The future of A.I. may combine elements of both patterns.
Some technologists imagine a world where machines perform most work while humans enjoy leisure supported by new forms of income. Most economists view that scenario as speculative. Past technological revolutions, from the Industrial Revolution onward, generated predictions that work would end; none have yet produced such an outcome.
The more immediate concern is the nature of the transition. If A.I. adoption proceeds gradually, with new roles and industries emerging as older ones shrink, the economy may adjust without extreme dislocation. The process would still be painful for many individuals, but manageable overall.
If adoption is rapid and heavily concentrated in certain sectors or regions, the transition could resemble the China shock: swift job losses, limited time for adjustment, and serious social strain.
At present, available data do not reveal which pattern will dominate. Different sectors may experience different dynamics, and the balance may shift over time.
The Policy Response: Early and Incomplete
Policy discussion is underway but remains in early stages.
Governments and think tanks are beginning to examine A.I.’s potential impacts, and public hearings are being held. However, there is no consensus yet on a comprehensive framework for managing the transition.
Economists and policy experts often highlight three priorities:
Improved measurement.
Modern tools are needed to track A.I.’s impact in close to real time, including:Which occupations are changing most rapidly.
Which regions are experiencing displacement.
Which workers are at greatest risk and how they are faring.
Strengthening existing safety nets.
The pandemic revealed weaknesses in unemployment insurance systems, while trade adjustment programs for workers hurt by globalization often failed to reach those most affected. Many experts argue that existing protections must be modernized and expanded to avoid leaving workers exposed during A.I‑driven shifts.Exploring new, A.I‑specific instruments.
Several more radical ideas are now part of mainstream debate:Sovereign wealth funds that hold stakes in A.I‑intensive companies, allowing gains to be shared more broadly.
Variants of universal basic income that provide a stable floor of support in the face of more volatile employment.
These concepts remain far from implementation but indicate the scale of change some policymakers anticipate.
Guidance for Individuals: Uncertainty as a Given
For policymakers, the emerging agenda is relatively clear: measure better, reinforce safety nets, and consider innovative tools. For individual workers and families, actionable guidance is much less definite.
During the 1990s, a reasonably coherent narrative existed. The message was: pursue higher education, move toward knowledge‑intensive and tech‑related fields, and the chances of economic security will improve. That advice did not guarantee success, but it offered a direction.
Today, the situation is more ambiguous. A.I. will almost certainly create new jobs, but the exact nature of those roles remains unclear. It is difficult to say which degrees, skill sets or training programs will be safest bets over a multi‑decade horizon.
Parents question what to tell their children about career choices. Mid‑career workers wonder whether to specialize in A.I‑related skills or diversify away from areas that seem heavily exposed. Older employees sometimes decide to retire earlier rather than adapt to rapidly changing tools and expectations.
This uncertainty is not a temporary communications problem; it reflects genuine unknowns about how A.I. will reshape work. The technology is here, and adoption is accelerating, but the detailed map of future occupations and industries does not yet exist.
For now, the most accurate conclusion is that A.I. is moving faster than the systems designed to measure and manage its impact. Its eventual economic footprint may be enormous, but the current evidence shows only subtle early effects. Whether the transition resembles the gradual internet revolution or the disruptive China shock—or some combination of both—will depend largely on the speed, concentration and distribution of change.
In the meantime, workers, businesses and governments alike will have to operate under a high degree of uncertainty about what comes next.
Comments
Post a Comment