AI Productivity Statistics (Individual and Company-Wide)

AI productivity numbers are everywhere, and most of them point in the same direction. Workers move faster, clear more tasks, and save a few hours each week. Many controlled studies back that up. But team and company-level productivity measures are often overlooked by these reports. Time one person saves can reappear as review work for a colleague.

This page pulls together the most reliable data on both. It covers what AI does for an individual worker, what happens when that work flows through a team, and where the company-wide returns land.

Why AI Adoption Numbers Range From 4% To 88%

When you look at AI adoption statistics, it’s essential to understand that current and credible adoption figures range from about 4% to 88%. They disagree because they measure different things.

• 88% of organizations reported using AI in 2025 in Stanford’s AI Index. That counts any employee using any AI tool for anything.
• 89% of workers reported using AI in the Founder Reports 2026 AI in the Workplace Report, which also counted any use of an AI tool.
• About 20% of EU enterprises with 10 or more employees used AI in 2025, per Eurostat, which measures firm-level deployment of AI technologies.
• 7% of US firms and 4% of EU firms used AI in production in 2025, per a 2026 St. Louis Fed working paper, which only counts AI that’s used for making products and services.
• About 4% of firms across OECD countries have high-intensity AI adoption in core business functions, ranging from 0.2% in Mexico to 7% in Ireland.

(Sources: Stanford, Founder Reports, Eurostat, Federal Reserve Bank of St. Louis, and SUERF)

AI And Individual Worker Productivity

The clearest evidence for AI productivity gains comes from randomized studies where researchers give one group access to a tool, withhold it from another, and measure the difference. When the task is well-defined, the gains are real and often large.

A Microsoft Research and Harvard Business School field experiment followed 7,137 knowledge workers across 66 firms over six months. The 80% of workers who used the generative AI tool spent two fewer hours on email each week and reduced time working outside regular hours. The tool was built into the apps they already used for email and writing, so this reflects everyday office work rather than a lab exercise.

Customer support shows a similar result. In a study summarized in the 2026 AI Index, researchers tracked support agents at a Fortune 500 software company after a generative AI assistant rolled out. Agents resolved 14–15% more issues per hour on average. However, the gains weren’t spread evenly. Novice and lower-skilled agents improved by 30–35%, while the most experienced agents barely moved.

That uneven distribution keeps coming up. In a consulting experiment published in 2025, 758 Boston Consulting Group consultants using GPT-4 completed 12.2% more tasks, worked 25.1% faster, and produced work rated 40% higher in quality, as long as the tasks sat inside the model’s capabilities. Lower-performing consultants saw the biggest jump.

(Sources: National Bureau of Economic Research, AI Index, and Harvard Business School)

AI Productivity By Job Function

Gains vary a lot depending on the work.

Software Development

This is the most-studied function, and the picture is mixed. Studies summarized in the 2026 AI Index found developers using GitHub Copilot completed 26% more pull requests, with junior and less-experienced developers benefiting the most.

A 2025 randomized trial from METR found the opposite for a specific group. Experienced open-source developers working on codebases they knew well took 19% longer with AI tools than without them. What’s just as interesting is that they had no idea it was happening. Before the study, they expected AI to speed them up by 24%. Afterward, having lived through the slowdown, they still believed AI had made them 20% faster. (Per the 2026 AI Index, METR was unable to replicate the result in a later study, partly because developers became reluctant to work without AI. Treat this as one data point, not a verdict.)

Customer Support And Consulting

Covered above. Support agents resolved 14–15% more issues per hour, and consultants completed 12.2% more tasks at 40% higher quality.

Marketing, Content, And Sales

Marketing teams using multimodal AI for ad creation saw a 50% increase in output per worker, per studies summarized in the 2026 AI Index. On the sales side, a set of large randomized field experiments at a cross-border online retailer, described in a 2025 NBER working paper, found that adding generative AI to consumer-facing workflows lifted sales by up to 16.3%, mostly through higher conversion rates rather than bigger orders. Effects ranged from nothing to that 16.3% ceiling depending on how much the AI improved on what the company already did.

Accounting And Finance

Accountants using AI-based accounting tools saw a 55% increase in weekly client-support throughput, per research summarized in the 2026 AI Index. The accounting figure is one of the larger measured gains in any function.

Product Development And Cross-Functional Work

One of the more striking recent findings comes from a pre-registered field experiment at Procter & Gamble with 776 professionals working on real product innovation problems. Individuals working with AI spent 16.4% less time on the task than the control group, teams with AI spent 12.7% less, and individuals equipped with AI matched the performance of two-person teams working without it. The AI also broke down silos. Without it, R&D staff leaned toward technical solutions and commercial staff toward commercial ones. With it, both groups produced more balanced proposals.

(Sources: AI Index, METR, NBER, and NBER)

Team And Organizational Productivity

Individual improvements are only half the story. But looking at the impact on teams often shows different results.

The 2024 DORA report from Google Cloud studied AI’s effect on software delivery across thousands of technology professionals. Individual productivity and job satisfaction went up. Delivery performance went down. A 25% increase in AI adoption was associated with a 1.5% drop in delivery throughput and a 7.2% drop in delivery stability. Thirty-nine percent of respondents reported little to no trust in AI-generated code. The report’s explanation is that AI makes it easy to produce more code in bigger batches, and bigger batches carry more risk.

Then there’s the rework problem.

The workslop research solidified what a lot of managers have felt. 40% of workers said they’d received workslop in the previous month. Each instance took an average of one hour and 56 minutes to deal with. Based on salaries and time spent, the researchers estimated a cost of about $186 per employee per month, which works out to more than $9 million a year at a company with 10,000 employees. The damage isn’t only financial. More than half of workers who received workslop viewed the sender as less capable.

Our own research points the same way. In Founder Reports’ AI in the Workplace study, 57% of managers and above said they’d had to fix or redo work from coworkers that leaned too heavily on AI, compared with 38% of individual contributors. Among VPs and C-suite executives, that figure climbed to 63%. And even when the output was acceptable, 77% of workers said they review AI-assisted work from colleagues more carefully than fully human work. The review cost rises as you move up the org chart.

(Sources: Google Cloud, Harvard Business Review, and Founder Reports)

AI Adoption And Usage

Adoption is high and still climbing, though the depth of use isn’t quite what many headlines would lead you to believe.

At the organizational level, 88% of organizations reported using AI in 2025, up from 78% the year before, per Stanford’s 2026 AI Index. Generative AI reached 79% of organizations in at least one business function. Europe and China posted the largest year-over-year increases, up 11 and 13 percentage points respectively (faster growth than North America).

Worker-level usage is more modest once you look at the depth of use. Gallup’s Q4 2025 data found 26% of U.S. employees use AI at work at least a few times a week and 12% use it daily. Nearly half, 49%, said they never use AI in their role. Usage is high in knowledge-based industries, with technology at 77% total use and retail near the bottom.

Our 2026 AI in the Workplace report found that 51% of US workers use AI at least weekly, and 23% use it daily.

A 2026 St. Louis Fed working paper found 43% of US workers used AI for their job in 2026, compared with 32% across surveyed European countries. More telling: US workers spend about 5.2% of their work hours using AI, versus 2.8% in the UK. Share of work hours is a better productivity proxy than share of workers.

In the UK, the Office for National Statistics tracked 38,637 businesses and found self-reported AI use rose from around 12% in late 2023 to around 35% by June 2026. Among the businesses using AI, though, only 10% report using it extensively, and only 15% say more than half their employees use AI daily. The average number of AI technologies per adopting business moved only from about 1.4 to 1.6.

The gap between spending and results is also interesting. In McKinsey’s January 2025 Superagency report, 92% of companies said they plan to increase AI investment over the next three years, while only 1% of leaders called their company “mature” on AI deployment.

(Sources: AI Index, Gallup, Founder Reports, Federal Reserve Bank of St. Louis, Office for National Statistics, McKinsey, and SUERF)

The ROI Reality

Reliable data on company-level return is harder to find than the individual-level speedups.

The most-cited number here comes from MIT’s Project NANDA, whose 2025 report found that 95% of enterprise generative AI pilots delivered no measurable profit-and-loss impact, despite $30 to $40 billion in collective spending, as Fortune reported. The researchers were clear that this reflects how companies deploy AI rather than a failure of the models themselves. Pilots stalled when tools couldn’t retain context or fit into existing workflows.

The firm-level causal evidence lands somewhere in between the extremes. A 2026 European Investment Bank working paper matched EIBIS-ORBIS data on over 12,000 non-financial firms in the EU and US, instrumenting EU adoption by US peer rates to get a causal rather than correlational estimate. AI adoption raised firm labour productivity by 4%. Gains came from capital deepening rather than job cuts, and concentrated in medium and large firms.

Buying AI subscriptions is not the same as getting AI returns. In euro-area firm-level data from a 2025 ECB paper, the association between AI adoption and firm productivity was statistically significant only among firms making significant AI investments. Firms relying solely on externally purchased AI solutions showed no significant productivity premium at all. In France specifically, only firms developing AI in-house showed a significant effect. If your AI strategy consists of giving team members access to ChatGPT and hoping, the research says you shouldn’t expect measurable returns.

(Sources: Fortune, EIB, NBER, and European Central Bank)

Deployment Depth Matters More Than Company Size

If you own a small business and the ROI numbers above feel discouraging, this is one of the most relevant details for you.

In the OECD’s 2026 D4SME Survey, only 2% of firms using AI for isolated tasks reported transformational benefits, compared with 23% of firms deploying AI enterprise-wide (an eleven-fold difference). Among self-entrepreneurs and micro-enterprises, 33% and 31% respectively reported strong or transformational effects, versus 22% of medium-sized firms. (The OECD notes the D4SME sample isn’t fully representative of the broader SME population, so take the exact percentages with that caveat.)

Two things stand out from that. First, what predicts transformational returns is how broadly you’ve wired AI into the business, not how big the business is. Second, the smallest operators have the freedom to redesign every workflow at once (solopreneurs or very small businesses). These businesses report the strongest effects of any size category. That flips the usual narrative that large enterprise budgets are what unlocks AI ROI.

(Source: OECD)

AI and Jobs

Among UK businesses using AI to improve operations, the ONS found 63% reported no change in workforce headcount, 6% reported a decrease, and only 1% reported an increase.

It’s also worth looking at expectations. Stanford’s 2026 AI Index found one-third of organizations expect AI to reduce their workforce over the coming year (35% at organizations with $1 billion or more in revenue versus 30% at smaller firms). 43% expect little or no change. Weighted to the US economy, the CFO survey behind the NBER working paper above implies executives expect employment to decline due to AI by about 502,000 workers in 2026.

Executives also expect the workforce’s composition to shift. They project the share of routine clerical roles to fall by 0.76% in 2026 and 2.19% by 2028, partly offset by increases in skilled technical roles of 0.62% and 1.35% respectively. Office and administrative support is the only occupation group where executives describe AI as replacing work more often than enhancing it.

(Sources: ONS, AI Index, and NBER)

Economy-Wide Context

Zoom out to the whole economy, and the measured effects of AI are real but still relatively modest.

Per Stanford’s 2026 AI Index, US labor productivity growth reached 2.7% in 2025, nearly double the 1.4% annual average of the previous decade. Researchers caution this may reflect a J-curve, where adoption costs show up first and measurable gains follow, so it’s too early to conclude that AI caused this.

The official international statistics are more cautious. The OECD’s 2025 Compendium of Productivity Indicators found labor productivity across OECD countries grew about 0.4% in 2024 and noted that AI’s impact isn’t yet clearly visible in cross-country productivity data, even as adoption climbs. OECD projections for G7 economies summarized in the 2026 AI Index put annual labor productivity gains over the next decade at 0.4 to 1.3 percentage points for the US and UK, versus 0.2 to 0.8 percentage points for Italy and Japan, and attributed to the OECD as a projection rather than fact.

A 2026 survey of 6,000 executives across the US, UK, Germany, and Australia found widespread adoption but minimal realized productivity gains so far, projecting a 1.4% productivity boost, 0.8% output increase, and 0.7% employment reduction over the next three years. The 2026 International AI Safety Report estimates the range of task-level gains at 20 to 60%, with 15 to 30% in most real-world settings and outliers on both ends.

(Sources: AI Index, and OECD)

How to Measure AI Productivity on Your Own Team

The statistics in this article won’t tell you what’s happening inside your own company, and as the CFO survey and METR data above show, perception and reality often don’t always match. The only way to know your return is to measure it.