AI adoption

AI Adoption Statistics 2026: The Numbers Behind the Adoption Gap

AI adoption statistics for 2026, organised around the gap that matters: the distance between how many organisations bought AI and how many people use it.

By Dr. Jacqueline Kerr · · 7 min read

AI Adoption Statistics 2026: The Numbers Behind the Adoption Gap

The licence was issued in February. The board saw a strategy. Finance saw a cost. The analyst opened it twice.

All four of those are true at once, and only the last one tells you anything about whether the work changed.

That is the state of AI adoption statistics in 2026. There is no shortage of numbers: procurement figures, licence counts, board-level intent surveys, national adoption rate comparisons. What is scarce is the number a leader could actually use. Of the people handed access, how many changed how they work, and kept it changed after week three.

The space between what an organisation bought and what became normal is where the reported adoption rate stops meaning very much.

This page collects what is measurable, with the collection dates attached, and is explicit about what is not.

The headline numbers, and why they disagree

88% of organisations now use AI in at least one business function, up from 78% a year earlier (McKinsey Global Survey on the state of AI, published November 2025). Stanford's 2026 AI Index reports the same 88% organisational adoption figure for 2025, with generative AI in use at 70% of organisations.

Meanwhile the US Census Bureau, surveying actual businesses of every size rather than survey panels of larger firms, finds that 17% to 20% of US businesses used AI between December 2025 and May 2026.

Both numbers are real. One counts mostly larger organisations answering "do we use AI anywhere"; the other counts every business in the economy, most of which are small. The difference between 88% and 19.8% is not a contradiction. It is the first lesson in reading this page: check who was asked, and what counted as adoption, before you quote anything in a steering meeting.

Adoption rates by country

  • EU: 20.0% of enterprises with 10 or more employees used AI technologies in 2025. Among large EU enterprises, 55.0% (Eurostat, published December 2025).
  • US: 17–20% of businesses used AI between December 2025 and May 2026, with 20–23% expecting to use it within six months (US Census Bureau, Business Trends and Outlook Survey).
  • Consumer generative AI, by country: generative AI reached 53% population adoption within three years of launch, faster than the personal computer or the internet. Adoption correlates strongly with GDP per capita, but not perfectly: Singapore is at 61% and the UAE at 64%, while the United States, despite leading investment and model development, ranks 24th at 28.3% (Stanford AI Index 2026).

Note what just happened across those three bullets: the US is simultaneously 19.8% (share of firms) and 28.3% (share of people, generative tools). National comparisons mix definitions like this constantly. Read the methodology before repeating any of them.

Adoption by industry

Sector comparisons are the ones most likely to be quoted at you in a steering meeting, usually to argue that you are behind. Handle them carefully. A sector figure tells you what peers have bought, not what their people do on a Wednesday.

US Census data as of May 2026: Information sector 39.7% and Finance and Insurance 33.9%, both roughly double the national rate of 19.8%. Retail Trade sits below it at around 14%. None of these sectors moved significantly in the preceding six months, which is its own finding: the easy adoption has happened, and the curve has flattened into the hard part.

Adoption by company size

Size shapes adoption in two directions at once. Large organisations have the budget, the platform, and the governance to deploy at scale, and the layers of approval that make any individual experiment expensive. Small organisations can decide on a Tuesday and be using it on Wednesday, with nobody to help when it goes wrong.

The data says scale is winning, at least on the buying:

  • 37% of US firms with 250+ employees used AI as of May 2026, against under 20% of firms with four or fewer employees (US Census Bureau).
  • In the EU, 55% of large enterprises used AI in 2025, against 20% of enterprises overall (Eurostat).
  • Nearly half of companies with more than $5 billion in revenue have reached the scaling phase of their AI programmes, compared with 29% of companies under $100 million (McKinsey, November 2025).

The usage versus value gap

This is the section that matters most, and the one where the public numbers get honest.

  • 88% of organisations use AI somewhere, but nearly two-thirds have not begun scaling it across the enterprise, and just 39% report any EBIT impact at the enterprise level (McKinsey, November 2025).
  • AI agents, the current wave of spending, are earlier still: no more than 10% of organisations are scaling agents in any single business function (McKinsey, November 2025; Stanford's AI Index makes the same single-digits observation).
  • On the individual side, usage runs ahead of the organisation, not behind it: 75% of global knowledge workers reported using generative AI at work as far back as mid-2024, nearly half of them having started within the previous six months (Microsoft and LinkedIn Work Trend Index, May 2024).
  • And 57% of employees say they hide their AI use and present AI-generated work as their own (KPMG and University of Melbourne global study of 48,000+ people across 47 countries, fieldwork November 2024 to January 2025).

Put those four numbers in one room and the shape of the problem appears. Organisations bought AI and cannot convert it to enterprise value. Individuals adopted AI and will not tell their employer. The gap is not between the technology and the market. It is between the official change programme and what people are actually doing, quietly, without support.

One number that would complete this picture is seat utilisation: of licences assigned, how many see active weekly use months later. Vendors hold that telemetry and do not publish it in comparable form. Treat any confident public claim about it with suspicion.

Employee sentiment

  • 52% of US workers say they are worried about the future impact of AI in the workplace; 32% expect it to mean fewer opportunities for them (Pew Research Center, February 2025).
  • Across 25 countries, a median of 34% of adults are more concerned than excited about AI in daily life (Pew Research Center, October 2025).
  • Only 47% of employees say they have received any AI training, and only 40% say their workplace has a policy or guidance on generative AI use (KPMG and University of Melbourne, 2025).

Read the training number next to the hiding number. A majority of employees use AI and conceal it; a minority have been trained or given guidance. People are not waiting for the change programme. They are working around it, and carrying the risk personally.

The one number that already explains most of it

Gartner found that only 32% of leaders get employees to adopt change in a healthy way. That figure is not about AI specifically, which is exactly why it matters here. The bottleneck predates the technology. We have simply handed the same 68% a faster tool and asked them to try again.

The adoption gap is a people gap

Every number above is a proxy. Purchase is a proxy for intent. Licence assignment is a proxy for access. Even weekly active use is a proxy for the thing that actually matters, which is whether someone changed how they do their job and kept it changed when it got inconvenient.

Statistics tell you the size of the gap. They tell you nothing about what it will take to close it, because what it takes is different for the manager protecting her standing, the analyst at capacity, and the site lead who has never been asked what he thinks.

None of this is a technology curve. It is a room full of people, each at a different stage, each protecting something specific.

One question to sit with

Think about the last adoption figure you reported upward.

What would that number have been if it counted only the people whose Tuesday actually changed?

Sources

Keep reading