How Might AI Change the Workplace? Evidence from Corporate Executives
March 25, 2026
Note: The graphs and analysis in this blog are based on a research paper that can be found here.
The advent of sophisticated artificial intelligence (AI) technology has fueled intense speculation over its potentially transformative role in the workplace. Exhaustive press coverage cites AI's potential to increase worker productivity while simultaneously changing the size and composition of the workforce.
To better understand AI's implications for labor productivity, workforce composition, and employment—topics of perennial interest to researchers and policymakers—we present a study using new data from a survey of more than 700 corporate executives across a wide range of sectors, geographies, and firm sizes. The data provide early evidence on firms' current and expected AI investments and their implications for productivity and labor markets: Although investment in AI and related software has increased rapidly in recent years (see figure 1), the technology is still too nascent for the effects on workforce outcomes to be easily observed in official statistics.
Overall, companies report divergent experiences with AI. A majority invested in AI in 2025, and a much larger share expect to invest in AI in 2026. On average, business executives report labor productivity gains and anticipate further increases. Strikingly, especially in light of recent slowing aggregate job growth, we find little evidence that firms have experienced or anticipate near-term AI-driven employment declines, even as AI could reshape task allocation. However, companies—especially larger firms—anticipate a shuffling of the workforce, with a shift away from routine clerical jobs toward more skilled technical jobs and tasks.
The data are gathered from two waves of surveys of financial executives. The primary survey was fielded from November 11 to December 16, 2025, and produced 603 responses from panel members in The CFO Survey (a quarterly survey of CFOs that the Atlanta Fed conducts in partnership with the Richmond Fed and Duke University). We sent additional surveys to senior financial decision-makers whose firms are members of Financial Executives International (FEI) and/or NASDAQ, as well as alumni of Duke University in the finance and tech industries. These supplemental surveys ran from mid-December to mid-January 2026 and collected 145 responses.
Eighty percent of firms will invest in AI during 2026, and large firms invest the most
Nearly 60 percent of responding firms invested in AI in 2025. This includes 80 percent of large firms and around half of small firms in the sample. The percentage of firms investing in AI is expected to grow substantially, with more than 80 percent of firms expecting to invest in 2026 (see figure 2). The largest change was among small firms, where 30 percent more will invest in 2026 despite not investing last year.
The amount of investment in AI varies widely, though large firms have outspent (and expect to outspend) smaller firms by a substantial margin (see figure 3). For context, around 30 percent of large firms expect to invest more than $1 million in AI in 2026 (compared to 1 percent of smaller firms). Meanwhile, nearly 60 percent of smaller firms plan to invest less than $20,000 in AI in 2026 (compared to 14 percent of larger firms).
Productivity improvement—not cost reduction—was the main driver of AI investment
We asked companies what motivated them to invest in AI. They ranked topics on a scale of 0 (not a motivation) to 4 (an extremely important motivation). Figure 4 shows that productivity- and efficiency-related objectives were bigger motivations to invest in AI than cost reduction was. Indeed, among the highest-ranked categories on average were improving production efficiency and labor productivity, while lower-ranked options include reducing labor and nonlabor costs. CFOs also indicated that innovation (developing new products) and demand (better reaching and serving customers) motivated their companies to invest in AI.
We also asked CFOs what benefits their company had experienced from investing in AI in 2025, and what benefits they anticipate in 2026. Figure 5 shows that the primary motivators for AI investment—improving production efficiency, enhancing decision-making speed, and enhancing output—are also where the largest AI benefits have been realized. In aggregate, small and large companies alike expect larger effects in 2026 than they experienced in 2025.
In addition, aggregated across all firms, there is little evidence that AI usage has meaningfully affected the total number of employees or costs, nor do business executives anticipate shrinking employment and costs this year: firms reported a negligible impact from AI on employee headcounts in 2025, and the average impact on 2026 employment levels is also close to zero (see figure 6). That said, one subgroup of firms—namely, large companies—expects to reduce employment by 0.8 percent in 2026 as a result of AI.
The key takeaways are that amid swift AI adoption and sizable reported gains in productivity, firms have not decreased their headcounts due to the incorporation of AI—nor do they plan to decrease them in the near term.
To more precisely estimate firm-level productivity effects from AI investment, we compare firms' reported changes in output per worker due to AI with those implied by their realized and expected changes in sales revenue per worker.
A productivity paradox: Reported productivity growth from AI exceeds the implied gains
In 1987, when computer use was growing rapidly, Robert Solow famously said that we can see computers everywhere but in the productivity statistics because the growth in aggregate productivity lagged growth in computer sales and use. We document a related productivity paradox for AI. As we saw in figure 5, companies reported an increase in productivity (output per worker) from AI usage of 1.8 percent in 2025. Yet when we compute an implied productivity change using the component parts—the AI-attributed change in revenue relative to AI-driven employment changes—the implied gains are much smaller across all major industries, both in 2025 and 2026. This wedge likely reflects delayed output realization and quality improvements not yet captured in measured revenues. More broadly, firms appear to view productivity as improvements in workflows, task efficiency, and organizational capacity whose revenue effects materialize gradually—echoing classic productivity-paradox arguments where transformative technologies are recognized as important before their impact on productivity can be measured.
Despite their differences, both measures suggest that productivity growth is expected to exhibit meaningful sectoral heterogeneity, with high-skilled services (especially finance) exhibiting the largest gains (see figure 7). Interestingly, current revenue-based productivity gains remain substantially smaller than those observed during the late-1990s productivity surge from IT, consistent with the notion that AI deployment is still in its early stages.
AI more likely to substitute routine clerical tasks while complementing high-skilled analytical work
Though the near-term impact on overall employment appears modest, the impact could be more substantial in the future, especially on the composition of the workforce—the types of jobs/tasks that workers will perform. Looking a little farther into the future (into 2028) hints at potential changes in workforce composition and occupations most likely to experience impact. To do so, we gathered information about firms' current and expected workforce composition in 2026 and 2028 across routine/clerical roles, skilled technical roles, and creative/managerial roles. On average, companies expect the proportion of routine clerical workers to decline by 0.76 percent in 2026 and by 2.19 percent in 2028. These declines will be partly offset by increases in skilled technical workers in each of these years. Notably, firms with higher AI investment (and large companies) are significantly more likely to reduce their share of routine workers, and small companies are more likely to expand technical employment. Thus, the reallocation of jobs/tasks is not predominantly within-firm but rather occurs across the economy.
To illustrate which tasks are most affected by AI, we also analyze open-ended survey responses in which firms describe roles and responsibilities they expect AI to enhance or replace. Word clouds constructed from these responses (see figure 8) reveal clear patterns. Tasks most often enhanced by AI include marketing, accounting, finance, management, and analytical activities—in other words, areas where AI appears to complement workers by improving information processing and decision support. In contrast, tasks most often expected to be replaced are more dispersed and include administrative work, data entry, customer service, and other routine operational roles. Overall, these patterns reinforce the earlier findings: AI is most likely to substitute for routine clerical tasks while AI will complement high-skilled analytical and/or decision-oriented work.
Taken together, by leveraging timely survey data from financial decision-makers, our study provides novel firm-level evidence on how AI is affecting productivity and the workforce. AI adoption is already widespread and associated with measurable, revenue-based labor productivity gains. And although AI's impact has not demonstrably resulted in significant job losses, it is expected to shift tasks and types of employment within the labor force.