Executive AI Use Outpaces Proof of Firm-Wide Productivity Gains

Leaders sit at the top of the adoption curve

Tips on how executives can use AI to get more done are now a genre of their own. The data behind that genre shows something more specific than a set of tricks. Senior leaders have become the heaviest and most enthusiastic AI users in the workplace. The rest of the organization is moving more slowly, and the company-wide gains that would justify the enthusiasm have mostly not appeared.

Gallup's measurements show how wide the gap at the top has become. As of the second quarter of 2026, 51% of leaders said they use AI frequently, up from 17% in 2023. Among managers the figure rose from 15% to 36%, and among individual contributors from 9% to 26%.2 Leaders also report the strongest results: 21% call the effect on their productivity extremely positive, compared with 13% of individual contributors.2 SHRM's survey of 5,875 U.S. workers found the same pattern by rank. Directors and above said 63% of their work involves AI assistance, compared with 49% for managers and 34% for individual contributors.11

At the C-suite level, AI is increasingly a personal priority for the CEO. BCG reports that nearly three-quarters of CEOs say they are their company's key decision maker on AI. Companies expect to roughly double AI spending in 2026, from 0.8% of revenue to about 1.7%.8 The uses executives describe are mostly about decisions rather than drafting: predictive insights, real-time performance analytics, risk forecasting and scenario modeling.4 IBM's 2026 CEO study goes further. CEOs in its survey say 25% of operational decisions are already made by AI with no human involved.7

How deep executive use really goes

The reports disagree on how intensive executive use actually is. BCG says the 70% of CEOs it calls "pragmatists" spend seven hours a week working with, thinking about or learning about AI.8 A summary of a 2026 survey of nearly 6,000 executives puts typical hands-on use by regular executive users at about 1.5 hours a week.1

The two figures measure different things. BCG's number counts time spent thinking and learning about AI, while the survey figure counts actual use. Taken together, they suggest that for many leaders AI takes up more strategic attention than working time. That matters for how to read the tip-list genre. A leader who uses AI to prepare a board memo or sharpen a forecast may see a real gain on that task. That is a long way from AI being part of how the whole organization works.

The productivity paradox at company level

On this point the evidence is unusually consistent. Researchers surveyed nearly 6,000 senior executives in the U.S., U.K., Germany and Australia. About 69% of their firms reported using AI, but roughly nine in ten executives said it had no measurable effect on employment or productivity over the past three years.1 Looking ahead three years, they expect an average productivity gain of 1.4% and a 0.7% drop in employment.1 Other coverage of the research, which involved the Atlanta Fed, put the result even more bluntly. It said the companies seeing gains are the ones that redesigned workflows and trained staff, not the biggest spenders.6

The Atlanta Fed's own analysis of about 750 U.S. executives gives a more nuanced version. Firms reported that AI raised output per worker by 1.8% in 2025. But when researchers recalculated the gain from AI-attributed revenue and staffing changes, it came out much smaller in every major industry.9 The authors compare this to Robert Solow's 1987 remark that computers showed up everywhere except in the productivity statistics. They attribute the gap mainly to output that has not yet shown up in revenue.9 Where gains are real, they cluster in high-skill services and especially finance.3

Worker-level data shows the same split between individual tasks and the firm as a whole. Gallup found that 65% of employees in organizations that have implemented AI say it improved their productivity. It also found the benefits concentrated in individual tasks rather than wider systems.2 Controlled studies back up large gains on specific jobs. Developers using an AI coding assistant finished a task 55% faster, and professional writers worked about 40% faster.110

Correcting AI output eats into the savings

SHRM's data contains a detail that most productivity checklists leave out. Workers reported saving about six hours a week with AI but spending about four hours correcting what it produced.14 For directors and above, the figures were nine hours saved and six spent fixing output.14 The more senior the user, the more time AI saves, and also the more time goes to cleaning up its mistakes.

The quality problem is widely recognized. Among workers who use AI, 44% describe their own AI-assisted output as "AI slop."11 In the KPMG and University of Melbourne study of 48,000 people, 66% said they rely on AI output without checking it, and 56% said they had made mistakes at work because of AI.17 This is likely part of why reported gains outrun measured ones. A task can feel faster while the time spent checking and fixing it gets pushed onto someone else.

The main lever is organizational

The most useful finding for executives is that individual skill with AI matters less than how the organization is set up. Microsoft's 2026 Work Trend Index surveyed 20,000 knowledge workers and tested 29 variables. Organizational factors explained about 67% of the AI impact employees reported, and individual factors about 32%.17 Reported AI value rose 17 points when managers used AI openly. Yet only 26% of AI users said their leadership was clearly aligned on an AI strategy.10

Gallup reaches a similar conclusion through managers. Employees who strongly agree that their manager supports AI use are 9.3 times as likely to say AI has transformed how work gets done. But only 36% of employees in organizations adopting AI give their manager that rating.2 Breadth of use matters too. Among workers who use AI for one or two purposes, 45% report a productivity gain. Among those using it for seven or more, 90% do.10

The implication is that the most important thing an executive can do with AI is not personal. Redesigning workflows, training managers and setting clear policy will likely do more for company output than any personal prompt routine. The cost of ignoring this is already visible. In SHRM's survey, 30% of workers admitted breaking their organization's AI rules.14 IBM data shows unsanctioned "shadow AI" involved in 43% of security incidents.10

Labor market effects: modest overall, sharp for young workers

The employment evidence is divided. Executive surveys mostly find no damage so far. The Atlanta Fed found little evidence of near-term job losses overall, though larger companies expect AI-driven workforce reductions and smaller firms expect modest gains.3 S&P Global's purchasing managers' survey shows a shift. AI's net employment effect turned negative over the past year, at -5 points globally. Large companies forecast -13 points for the coming year, while medium-sized firms expect +2.22

The pressure is concentrated by industry. Goldman Sachs found call-center employment 39% below trend in the U.S., 33% below in Canada and 27% below in Germany. Entry-level workers faced the strongest AI-related headwinds. Stanford's Digital Economy Lab, using ADP payroll data, reports that employment of 22- to 25-year-olds in AI-exposed jobs now trails other jobs by 19%, up from 13% a year earlier.28 U.S. Census data shows hires of 22- to 24-year-olds in the most exposed industries fell about 9% after ChatGPT launched. Graduates from the most exposed college majors saw starting earnings drop about 13%.25

The evidence is not uniform, though. Data covering entire populations in Norway and Finland shows little or no AI effect on young workers.25 Several studies find that AI-exposed jobs started deteriorating before ChatGPT was released.2529 Researchers at the Peterson Institute for International Economics (PIIE) call claims of harm to specific groups premature.29 Other data points the opposite way. One study of 21,000 U.S. firms found that companies investing heavily in AI grew employment by 10%, and entry-level employment by 12%.25 PwC finds that the companies best able to use AI are growing headcount faster than the least exposed, by 52% versus 36%.27

What it adds up to

The best reading is that AI is changing jobs more than it is eliminating them, and the costs fall on specific groups rather than the whole workforce. Job postings for routine roles fell 13% after ChatGPT, while demand for analytical and creative roles grew 20%.21 BCG estimates that 50% to 55% of U.S. jobs will be reshaped over the next two to three years, and only about 12% fall into its "substituted" category.26 PwC finds that AI-exposed entry-level jobs are seven times more likely than other entry-level jobs to require senior-level skills such as judgment and leadership.27

This is the main problem with the executive-productivity story. Leaders are the group best placed to benefit from AI: they use it most, report the best results, and do the kind of high-skill decision work that AI tends to complement.29 The groups with the most to lose are early-career workers and clerical staff, who make up a smaller share of frequent users.2928 Executives who treat AI as a personal efficiency tool risk widening that gap while still not getting the company-wide gains their spending assumes. The evidence suggests the bigger payoff comes from redesigning workflows across the organization, starting with the managers who determine whether everyone else adopts AI.172