AI adoption statistics for 2026: what the data says for mid-size businesses

Every figure in this guide was checked against its original source on 23 September 2026. Adoption numbers for AI vary wildly depending on who is asked, because a national statistics agency, a consultancy and a work trends survey each measure a different population and a different definition of "using AI." This guide sources every figure to the original study, states what it measured and when, and translates it for a 20 to 500 person company deciding whether, and how fast, to move.

How many businesses use AI today?

Depending on the survey, current AI use runs from about one in six businesses to well over half of individual adults.

• National AI use among US employer businesses ran 17% to 20% between December 2025 and May 2026, with 20% to 23% expected within six months. US Census Bureau, May 2026.
• The Census Bureau's own survey put US firm level adoption at about 18% by December 2025, while the Survey of Business Uncertainty, a methodology deliberately weighted toward larger employers, put it at around 78% on an employment-weighted basis and about 54% for large language models, in November 2025. Federal Reserve Board, April 2026.
• 20.2% of firms across OECD countries used AI in 2025, up from 14.2% in 2024 and 8.7% in 2023. OECD, January 2026.
• 88% of organizations worldwide report using AI in at least one business function, and 70% report using generative AI specifically. Stanford HAI, 2026 AI Index Report.
• 62% of US adults and 45% of US workers reported using generative AI as of May 2026, up from 45% and 33% respectively in August 2024. Federal Reserve Bank of St. Louis, Real-Time Population Survey.

The number depends entirely on what counts as "using" AI. A firm level count like the Census Bureau's puts adoption under one in five. A survey of individual adults puts personal use over 60%. Both are accurate; they count different things, which is why the next section breaks adoption down by company size instead.

How does AI adoption vary by company size?

Every dataset broken out by employee count shows the same shape: larger companies adopt AI at two to three times the rate of the smallest ones, and a 20 to 500 person company sits across the middle of that curve.

In the UK, a secondary market for Kastling, AI use among businesses with 10 or more employees rose to around 35% in a June 2026 survey wave, up from about 12% in September 2023, on the same widening curve by size. The Census Bureau's own analysis found that AI use increased among US firms with at least 20 employees over the period but did not change significantly below 20 employees. A 20 to 500 person company usually sits across the fastest growing bands, not the slowest.

Illustrative example: An operations leader at a 140 employee distributor reads that 88% of organizations use AI and assumes her company is behind. Her staff already use a general chatbot, so by that broad definition she is not. Checking agent adoption specifically, she finds scaling rates closer to a fifth of organizations her size, and that knowledge gaps, not budget, are the most cited barrier nationally. The number does not tell her whether to proceed. It tells her which question to ask next: not "are we behind," but "which of these gaps applies to us."

What do businesses actually use AI for?

Once adopted, AI use concentrates in a handful of functions, most of them the same document heavy, judgment light work Kastling sees inside operations, finance and customer service teams.

• Information and communication technology shows the highest AI use of any industry across the OECD, at 57.3% in 2025, followed by professional and scientific services at 36.8%. OECD, January 2026.
• Among US workers using generative AI, the most AI-assisted tasks are reading documents for technical information (61.3%), preparing research reports (60.7%) and analyzing data to identify trends (57.5%). Federal Reserve Bank of St. Louis, Real-Time Population Survey.
• 66% of enterprise leaders report productivity gains from AI, 53% report better decision making, and 40% report cost reductions. Deloitte, 3,235 leaders across 24 countries, August to September 2025.
• 56% of organizations use AI in three or more business functions, up from 51% a year earlier, and chatbots remain the most widely scaled AI tool at 47%. McKinsey, The state of AI in 2026.

The heaviest use is reading technical material, preparing reports and a first pass at analysis, the same category which business processes to automate first recommends starting with. A general chatbot, not a specialized agent, still does most of that work.

How fast are companies adopting AI agents?

An AI agent is software that can plan and carry out a multi-step task using tools and data, rather than only answering a single question (see what is an AI agent). Adoption is growing fast from a small base and stays concentrated in large enterprises.

• Agent deployment "remained in the single digits across nearly all business functions" in 2025, despite 88% overall AI use. Stanford HAI, 2026 AI Index Report.
• 40% of organizations with more than $1 billion in annual revenue report scaling AI agents in at least one function, up from 27% a year earlier; smaller organizations held flat at 22%. McKinsey, 2026.
• Gartner forecasts 40% of enterprise applications will carry task-specific AI agents by the end of 2026, up from under 5% in 2025, and that agentic AI could reach roughly 30% of enterprise application software revenue by 2035, up from 2% in 2025 in its best case scenario. Gartner, August 2025.
• Gartner also forecasts over 40% of agentic AI projects will be canceled by the end of 2027 over rising costs, unclear value or weak risk controls, and estimates only about 130 of the thousands of vendors marketing agentic AI offer genuine agentic capability. Gartner, June 2025.
• Active AI agents inside Microsoft 365 grew 15 times year over year, and 18 times among large enterprises. Microsoft WorkLab, 2026 Work Trend Index.

These AI agents adoption statistics describe an early technology, not a missed wave. A 20 to 500 person company weighing agents is deciding whether to be an early adopter of something with a documented cancellation rate, which argues for one narrow, well defined task rather than an agent for everything.

What results do companies report, and how many see value at scale?

Individual gains are real and widespread. Enterprise-wide financial impact is a smaller number, and it has barely moved in a year.

• 37% of respondents say AI has contributed to their organization's EBIT, essentially unchanged from 2025, and only about 6% qualify as "AI high performers," attributing at least 5% of EBIT to AI with a self described significant impact. McKinsey, 2026.
• 80% say AI improved their individual productivity and 50% say it helps them make better decisions, fairly consistently across junior and senior levels. McKinsey, 2026.
• High performers redesign work far more often: nearly three quarters report fundamentally redesigning workflows because of AI, up from 55% a year earlier, against about one quarter of everyone else. McKinsey, 2026.
• 20% of enterprise leaders report increased revenue from AI, well short of the 74% who hope to, and only 34% call their AI use a deep transformation rather than a surface level add on. Deloitte, August to September 2025.

Employees feel the benefit well before it becomes a company wide financial result, and the organizations that do see EBIT impact got there mainly by redesigning how work is done, not by adding a chatbot to an unchanged process.

What holds companies back from adopting AI?

The barriers companies report are consistent across surveys: people, budget, data and rules, roughly in that order.

• Knowledge gaps are the most cited obstacle to responsible AI adoption, named by 59% of organizations, ahead of budget constraints (48%) and regulatory uncertainty (41%). Stanford HAI, 2026 AI Index Report.
• About 20% of organizations say AI-related operating costs, including the cost of model usage, already constrain how much they use AI, fairly evenly across company sizes and industries. McKinsey, 2026.
• To close the skills gap, 53% of enterprise leaders are training their existing workforce, 48% run formal upskilling or reskilling programs, and 36% are hiring specialized AI talent. Deloitte, August to September 2025.
• A scarcity of AI skills, especially specialized talent, hinders uptake "even in many large firms," not only smaller ones, per a policy oriented survey of 840 enterprises across the G7 plus 167 in Brazil. OECD, May 2025.

None of these barriers is unique to AI. They are the same skills, budget and data problems that slow any new system, and they respond to the same fix: scope one workflow, agree what data it needs, and train the people who will use it. What happens in an AI readiness assessment walks through checking for these gaps before committing budget, and Kastling's own AI readiness self-assessment scores them for one workflow in about 12 questions.

How do employees use AI at work, including tools nobody approved?

Employees adopted generative AI faster than most employers wrote a policy for it, and a meaningful share of that use still happens on personal accounts the company cannot see.

• 75% of knowledge workers reported using generative AI at work, and 78% of those users said they bring their own AI tools rather than an employer issued one, a pattern Microsoft calls BYOAI, rising to 80% at small and medium sized companies. Microsoft WorkLab, 2024 Work Trend Index, 31,000 knowledge workers across 31 countries.
• 52% were reluctant to disclose their AI use for important tasks, and only 39% globally had received any employer AI training. Microsoft WorkLab, 2024 Work Trend Index.
• By 2026, 49% of conversations inside Microsoft 365 Copilot supported cognitive work such as analysis and problem solving, and 66% of surveyed AI users said it freed up time for higher value work. Microsoft WorkLab, 2026 Work Trend Index, 20,000 workers across 10 countries.
• 41.1% of employed adults across the OECD reported using generative AI tools in 2025, against 36.7% of unemployed adults and just 12.5% of retired or otherwise inactive adults. OECD, January 2026.

A meaningful share of the workforce is already pasting company information into personal AI accounts, policy or not. That is a data governance question before it is a productivity opportunity, which the next section addresses directly.

How much oversight do companies have over their AI use?

Formal governance is catching up to adoption, but a meaningful minority of companies still runs AI, especially agents, without a policy stating what it may do on its own.

• The share of organizations with no responsible AI policy in place fell from 24% to 11% in a single year, and roles dedicated to AI governance grew 17% in 2025. Stanford HAI, 2026 AI Index Report.
• Among organizations citing a specific governance framework, 36% point to the ISO/IEC 42001 AI management standard and 33% to the NIST AI Risk Management Framework, both newly tracked in 2025. Stanford HAI, 2026 AI Index Report.
• The independent AI Incident Database recorded 362 AI-related incidents in 2025, up from 233 in 2024. Stanford HAI, 2026 AI Index Report.
• Only one in five companies has a mature governance model for autonomous AI agents specifically, well behind governance of AI in general. Deloitte, August to September 2025.

Agent governance is the clearest gap in this dataset. Most companies have not yet decided what an autonomous agent may do without a human in the loop, which is exactly the question data security and governance for AI is built to answer before an agent touches a live system.

How to read these adoption numbers

Reading AI in business statistics side by side explains why the headline number keeps changing. A Federal Reserve Board analysis makes the reason clear by comparing surveys directly.

Comparison drawn from Federal Reserve Board, April 2026, with the McKinsey and OECD rows added from the sources above.

Two more distinctions matter. "Use" can mean one chatbot session or a company running AI inside a regulated production workflow, and most survey questions do not force a respondent to say which. And nearly every figure here, including the ones from government agencies, is self reported: a firm or worker describing its own use, not an audit of production systems. Prefer the survey whose population and definition match your own business over whichever headline number is largest.

How Kastling approaches AI adoption

Kastling starts an AI engagement by finding out where a company already sits against numbers like these, not by assuming a starting point. A free discovery call covers what the business already uses and which workflows are candidates. For AI and operations work, a separately scoped paid audit is often the next step: reviewing the workflows, systems, data and current AI use, and matching the barriers above (skills, data, cost, governance) to what actually exists rather than to a national average. The AI Integration & Automation service builds from there, connecting existing systems, keeping a named person in control of consequential decisions, and agreeing upfront how success will be measured.