The American Productivity Surge Began Before Generative A.I

U.S. companies are producing more with fewer workers, but the gains owe as much to labor scarcity, digitization, remote work and industrial consolidation as they do to artificial intelligence.

Category: Strategy. Written by Jaime Garcia, Founder, SnowRock. Published . 17 min read.

In short

The American economy has entered an unusually strong period of productivity growth.

For several years, companies have increased the amount of output generated by each hour of labor at a pace not sustained in decades. Businesses are operating with leaner teams, more digital infrastructure and broader access to talent. Industries ranging from energy and health care to finance, retail and professional services are producing more without expanding employment at the same rate.

Artificial intelligence is often credited for this shift. The evidence suggests a more complicated story.

Generative A.I. is beginning to change how people write software, analyze information, produce research and perform administrative work. Some businesses are already using it to avoid new hiring or reduce the time required for routine tasks. But the productivity expansion began before large language models were widely deployed. Its deeper causes include the digitization of work during the pandemic, the normalization of remote employment, persistent labor shortages, improved operational software, machine learning, corporate consolidation and years of pressure on executives to increase margins. Artificial intelligence may eventually become the dominant force behind productivity growth. For now, it is an accelerant entering a process already underway.

Why AI productivity growth matters

Labor productivity measures the amount of economic output produced for each hour worked. At the level of an individual company, the concept is intuitive: if a business can serve more customers, manufacture more products or generate more revenue without increasing employee hours proportionately, productivity has improved. Across an entire economy, sustained productivity growth is one of the principal foundations of rising living standards.

A more productive company can theoretically increase wages, reduce prices, improve margins and reinvest in new capacity at the same time, producing more without requiring an equivalent increase in labor, materials or time. That creates the possibility of a broad economic gain: workers can earn more, customers can receive better or less expensive products, companies can generate stronger returns, and governments can collect more revenue without increasing tax rates.

The word theoretically is important. Higher productivity does not determine how the resulting value is distributed. A business can use the gain to raise compensation, reduce prices, hire more people or increase shareholder returns. It can also retain most of the benefit through wider margins and a smaller workforce. Productivity growth creates economic capacity. It does not guarantee shared prosperity.

Digitization Was the First Major Catalyst

The pandemic forced companies to digitize activities that had previously depended on physical offices, paper records, in-person meetings and manual coordination. That transition was abrupt, uneven and frequently inefficient at first. Over time, however, many of the changes became permanent improvements: medical records moved further into digital systems, restaurants adopted cloud-based inventory and scheduling tools, financial institutions automated document processing and customer service, companies replaced in-person administrative procedures with online workflows, and remote collaboration platforms became part of ordinary operations.

The cumulative effect was substantial. Information became easier to locate, share and analyze. Processes that once required repeated human handoffs could be completed automatically. Employees spent less time traveling between offices or attending meetings that did not require physical presence. Digitization also created measurable data around activities that had previously been difficult to observe: a company could identify where a workflow slowed, which locations carried excess inventory or how long employees spent completing a particular process. Once an activity became visible in software, it became easier to standardize and improve.

These gains did not always feel dramatic. They emerged through thousands of modest operational changes: fewer forms, faster approvals, better scheduling, automated reporting and more efficient coordination. Their collective effect helped create the foundation of the current productivity expansion.

Remote Work Expanded the Talent Market

Remote employment also altered the economics of hiring. Before 2020, many companies recruited primarily within commuting distance of an office; a business in a high-cost city competed for workers living nearby or willing to relocate. Remote work expanded that market. Companies could recruit specialists across states or countries, reach candidates with different salary expectations and retain employees whose personal circumstances made relocation impossible.

The productivity benefit did not come solely from people working more hours at home. It came from better matching. A company could hire a stronger candidate from a wider pool. A worker could accept a role better suited to their skills without moving. Employers could build teams around capability rather than proximity.

Remote work also reduced certain forms of organizational friction. Employees gained time previously spent commuting. Distributed teams relied more heavily on written communication and documented processes. Meetings became easier to record, transcribe and summarize. Organizations were forced to make information accessible outside informal office networks. The arrangement introduced new problems, including weaker social cohesion, more fragmented communication and difficulty training junior workers, yet even companies that later required employees to return to offices retained much of the digital infrastructure developed during the transition. The lasting productivity gain may therefore come less from where people work than from how work was redesigned when physical proximity could no longer be assumed.

Labor Scarcity Forced Companies to Improve

A tight labor market has been another powerful driver. When unemployment remains low, employers face greater difficulty attracting and retaining workers: wages rise, vacancies remain open longer and businesses become less able to solve operational problems simply by adding staff. That pressure changes investment decisions. A company that can hire inexpensive labor has less incentive to automate a repetitive process, but when labor becomes scarce or costly, the return on better software, equipment and workflow design increases. Employers begin asking whether a job can be simplified, whether a process can be consolidated or whether existing workers can produce more with improved tools.

This mechanism can become self-reinforcing. Higher wages encourage capital investment. Better capital increases output per worker. Higher productivity gives companies more capacity to support compensation or expansion. Strong household income can then sustain consumer demand, giving businesses a reason to continue investing. The cycle can persist until a serious economic shock interrupts it. This is one reason tight labor markets can produce more than wage pressure: they can force companies to confront inefficiencies that were previously hidden by abundant staffing.

Productivity Through Fewer Employees

Not all productivity gains reflect a healthier or more innovative workplace. Some result from layoffs. Technology and finance companies have reduced employment substantially while preserving large shares of their revenue and profit. Professional-services firms have delayed hiring, eliminated junior positions and expected remaining employees to absorb more work. Mathematically, this can raise productivity: if output declines by less than total labor hours, output per worker increases, and the statistic improves even though many employees lose their jobs and those who remain may face heavier workloads.

This distinction matters because doing more with less can describe two very different situations. In one, employees receive better tools and remove low-value work. In the other, a company reduces staff and redistributes the same responsibilities among fewer people. Both can increase measured productivity. Only one necessarily represents a genuine improvement in how work is performed. In practice, many organizations combine the two: they introduce software, restructure teams, remove positions and use the resulting efficiency to maintain output with a smaller payroll.

Artificial intelligence is increasingly part of that calculation. Federal Reserve surveys have found that some businesses are using A.I.-related efficiencies to postpone hiring or avoid filling open positions, and corporate earnings calls across industries have reflected reduced enthusiasm for expanding head count. The first labor-market effect of generative A.I. may be less visible than mass replacement. It may appear as jobs that are never created.

The Oil Industry’s Productivity Transformation

The American oil industry offers a clear example of productivity growth that predates generative artificial intelligence. Companies in the Permian Basin have combined improved drilling technology, longer horizontal wells, better geological data and industry consolidation to extract more oil with fewer rigs and fewer workers. A decade ago, a typical operation might drill approximately two miles vertically and one mile horizontally; improved equipment and techniques now allow operators to extend horizontal sections much farther through productive formations. Each rig can reach more resource. Each well can produce more. The return on labor and capital improves.

At the same time, mergers and acquisitions have concentrated assets within larger operators that can eliminate duplicated staff, standardize equipment and negotiate more favorable supplier terms. The result is a much leaner industry. Oil and gas employment has declined sharply from earlier peaks even as the United States has maintained extraordinary levels of production. Output per worker has risen because the industry changed its technology, physical design and ownership structure.

This pattern appears across the economy. Productivity does not arrive only through revolutionary inventions. It also emerges when businesses refine existing systems, consolidate operations and redesign the relationship between people and capital.

The Professional Economy Is Producing More With Less

Professional and business services have experienced especially strong productivity growth. The sector includes consulting, accounting, legal services, technical work, administration and other information-intensive activities, and annual productivity growth has remained elevated even as employment has weakened. This combination deserves attention. A sector can become more productive because employees are using better tools; it can also become more productive because companies reduce junior staffing while expecting senior employees to produce more. Both dynamics are occurring.

Data analysis, chart production, document review, financial modeling and research have become easier to automate. Work once assigned to analysts or junior professionals can increasingly be completed through specialized software, internal databases and artificial-intelligence systems. A senior economist may now perform analysis and visualization that previously required several junior employees. A lawyer may review large sets of documents with software. A consultant may generate preliminary research and presentation materials without a full support team. The individual professional becomes more productive. The traditional career ladder becomes narrower.

This creates a long-term organizational problem. Junior work often appears inefficient because it can be standardized or automated, but those tasks also train the people who eventually become experienced professionals. If companies eliminate too many entry-level roles, they may improve current margins while weakening the future supply of senior talent. Productivity at the task level can therefore conflict with capability at the institutional level.

Why the A.I. Effect Is Still Difficult to See

Economists remain divided over how much of the recent productivity increase can be attributed directly to artificial intelligence, for several reasons. First, adoption remains uneven: large technology companies, professional firms and sophisticated enterprises may use advanced systems extensively, while many smaller businesses have barely begun, and economy-wide statistics average together organizations at very different stages. Second, companies often need to reorganize before a new technology produces meaningful gains. Purchasing an A.I. tool does not automatically improve productivity; workflows must change, employees must learn when to use it, data must be made accessible, quality controls must be established, and existing systems may need to be replaced or integrated. These complementary investments can take years.

Third, A.I. may initially generate more experimentation than output. Employees spend time testing tools, correcting mistakes and determining where they are reliable, and early usage can create localized benefits without yet affecting national statistics. Fourth, productivity data are inherently noisy: output per hour can change because of inflation adjustments, shifts in industry composition, changes in employment and temporary economic disruptions, so short-term movement may reveal little about the underlying trend. The absence of a clear statistical relationship between A.I. adoption and employment does not prove that the technology has no effect. It may indicate that the effect is early, uneven or occurring through channels that conventional labor data do not yet capture.

Productivity Statistics Can Mislead

Productivity is usually calculated as inflation-adjusted output divided by hours worked, and both sides of that equation are difficult to measure precisely. The output of a factory can be counted in physical units; the output of a consultant, software engineer, nurse or financial analyst is more abstract, and quality improvements may not be reflected fully in prices or revenue. Inflation creates an additional complication: if prices rise sharply, economists must estimate how much of nominal growth reflects actual production and how much reflects higher prices, and errors or volatility in that adjustment can distort the measure. A temporary inflation shock can make real output appear weaker even when workers have not become less efficient.

Changes in the composition of employment can also raise productivity without any individual employee improving. During a downturn, lower-wage or less experienced workers are often laid off first, so the remaining workforce may appear more productive simply because it contains a larger share of experienced employees and high-output industries. Conversely, rapid hiring of inexperienced workers can temporarily reduce average productivity while increasing the economy’s long-term capacity. Productivity figures are therefore most useful when observed over several years rather than a single quarter. The recent trend appears strong enough to be meaningful; its exact magnitude and causes remain open to interpretation.

The Delayed Impact of Transformative Technology

Major technologies often take years to become visible in national productivity statistics. The economist Robert Solow famously observed that computers appeared everywhere except in the productivity data: businesses had invested heavily in information technology, yet the expected economic gains were initially difficult to detect. The gains became clearer only after organizations redesigned themselves around the technology, replacing legacy processes, connecting databases, changing supply chains, training workers and creating entirely new business models. The computer’s value did not come merely from placing a machine on every desk. It came from rebuilding work around computation.

Artificial intelligence may follow the same path, in three stages.

Where Artificial Intelligence Is Already Contributing

Although A.I. is not yet the central explanation for national productivity growth, it is producing tangible gains in specific activities. Software developers use coding assistants to complete routine functions, identify errors and navigate unfamiliar codebases. Researchers use models to organize information, produce preliminary analyses and accelerate literature reviews. Customer-service teams generate responses and summarize interactions. Marketing departments create initial drafts and variations of commercial content. Financial professionals automate portions of modeling, reporting and chart production. Medical systems use machine learning to assist with records, scheduling and administrative tasks.

The gains are strongest where work is digital, repetitive and easy to evaluate. They are weaker where success depends on physical activity, complex human relationships, accountability or extensive real-world context. This unevenness explains why artificial intelligence can appear revolutionary inside one company and almost irrelevant inside another. The technology’s immediate economic effect will be concentrated in particular tasks and occupations; its broader influence will depend on whether those localized gains spread through supply chains, pricing, investment and labor markets.

The Distribution Question

Productivity growth is economically valuable, but its political legitimacy depends on who benefits. For much of modern economic history, rising productivity was associated with rising worker compensation: employees produced more, companies earned more and wages generally increased alongside output. That relationship has weakened. Over the past several decades, productivity has often grown faster than inflation-adjusted compensation, and a larger portion of the resulting value has flowed to profits, highly compensated professionals and owners of capital. When real pay trails productivity, labor’s share of national income falls.

Artificial intelligence could widen that divide. The technology may allow companies to reduce head count, weaken the bargaining power of certain workers and capture more output through software and intellectual property owned by the firm. It could also create the opposite effect: if A.I. makes workers substantially more capable and labor markets remain tight, employees may negotiate higher compensation because each person generates more value, and new companies and occupations may emerge around capabilities that did not previously exist. The distribution will not be determined by technology alone. It will depend on labor-market conditions, corporate governance, competition, tax policy, education, worker bargaining power and who owns the systems producing the gain. A productivity boom can coexist with widespread economic insecurity. The statistic does not resolve the social question.

Better Productivity or Greater Work Intensity?

Another unresolved issue is whether employees are becoming more efficient or simply working harder. Remote and digital work have made it easier to measure output and coordinate tasks; they have also weakened the boundaries around the workday. Employees can respond to messages at all hours, attend more meetings and move rapidly between assignments, and software can track activity, deadlines and performance with greater precision. A company may record higher output per employee because technology removed unnecessary work. It may also record higher output because fewer people are working more intensely.

The difference matters for sustainability. A system that improves processes can support productivity over many years; a system that depends on overwork eventually produces burnout, turnover, mistakes and declining quality. Executives should therefore examine more than output per head. They should track error rates, employee retention, customer satisfaction, absenteeism and the amount of work requiring correction, because short-term efficiency can conceal long-term operational damage. The strongest productivity gains are those that reduce friction without exhausting the people who remain.

The Next Phase of the Productivity Cycle

The American economy now has several forces working in the same direction. Labor remains relatively scarce. Companies have invested heavily in digital infrastructure. Remote and hybrid work have expanded access to talent. Industries have consolidated. Software continues to automate administrative activity. Artificial intelligence is beginning to reduce the time required for knowledge work. Together, these conditions create the potential for productivity growth to remain elevated.

The outcome is not assured. A recession could reduce investment. Trade conflict could raise costs and disrupt supply chains. Weak competition could allow companies to retain productivity gains without lowering prices or increasing wages. Poorly implemented A.I. systems could create errors and complexity rather than efficiency. The labor market could also weaken enough that companies return to hiring inexpensive workers instead of investing in better systems. Productivity growth depends not simply on technological potential but on economic incentives: companies improve when they have both a reason and the capacity to invest.

The More Accurate A.I. Narrative

Artificial intelligence has become the dominant explanation for nearly every change in corporate strategy and employment. That narrative is premature. The United States was already experiencing a significant transformation in how work was organized before generative A.I. became widely available: pandemic-era digitization, remote work, labor scarcity, automation and consolidation had begun raising output per worker across major industries. A.I. is entering that environment at an important moment, after companies spent years moving information, communication and operations into digital systems, and those earlier investments make it easier to deploy intelligent tools now. The technology may therefore achieve adoption faster than previous innovations because much of the necessary infrastructure is already in place. Artificial intelligence did not create the entire productivity surge. It inherited one.

But the chronology matters. The central economic question is what happens when the existing trend meets a technology capable of automating not only routine physical or administrative work, but significant portions of analysis, communication and software production. The answer could be a sustained period of stronger growth. It could also be a period in which corporate output rises while employment opportunities and labor’s share of income decline. Productivity tells us how much the economy produces. It does not tell us whether the system producing it is equitable, resilient or sustainable. That will be decided by how companies deploy the technology, how workers adapt and how the gains are ultimately shared.

The SnowRock read

We find this the most clarifying way to think about AI in a business, because it lowers the temperature. If the productivity surge started before generative models arrived, then AI is not a magic lever that transforms a company on contact. It is the newest entrant in a decade-long project of digitizing work, and it pays off on the same terms the rest of that project did: only after the workflow is redesigned around it, the data is made reachable, and someone owns the result.

So the advice we give an operator reading this is to resist both the panic and the hype. You are probably not behind, and buying a tool will not, by itself, move your numbers. Pick the one or two tasks where the work is digital, repetitive and easy to check, redesign the process around the model rather than bolting it on, and measure the result in a line your finance team already tracks. The companies that win the next phase will not be the ones that adopted AI earliest. They will be the ones that reorganized around it most honestly, and shared enough of the gain to keep the people who make it work.