AI has crossed the billion-user line. Deep adoption is still much rarer

AI adoption figures describe different populations and thresholds. OpenAI reports more than one billion weekly active users. The GenAI Adoption Tracker estimates that 61.8% of US adults aged 18–64 used generative AI in May 2026, including use outside work and use before the preceding week. The ONS reports that 30% of UK businesses used at least one AI technology in September, while OECD firm statistics put use at 20.2% in 2025. None of these measures can substitute for the others.[1][6][3][5]

The market is moving from reach to depth. Trying a chatbot once, using one free tool at work, paying for a seat, embedding AI in a repeatable workflow and delegating a process to an agent are five different levels of adoption. The headline curve is steep; the operational curve remains much shallower.

The 53% adoption figure reproduced in Stanford’s AI Index comes from an earlier point in this US survey series, not a survey of the whole world. The tracker’s May update also reports work use among 45.2% of employed respondents and use during 6.3% of total work hours. Even within one dataset, having tried AI, using it for work and the intensity of that work use are distinct measurements.[7][6]

Breadth is rising faster than intensity

The ONS provides a useful depth check. Among adopting UK businesses with ten or more employees, the average number of AI technologies used rose only from about 1.4 to 1.6 from late 2023 to June 2026. Only 10% of adopters reported extensive use. A company can therefore count as an adopter while using one narrow tool in one department.[4]

The same report finds text generation with large language models was the most common AI technology among UK businesses with ten or more employees, followed by visual content creation. Improving existing operations was the most common purpose. New products, new markets and extensive workforce transformation were less widespread. The adoption curve is real, but much of it still describes augmentation at the edge of current processes.[4]

Large companies lead reach; small companies can use more flexibly

ONS data show adoption increasing with firm size: 29% among businesses with fewer than ten employees and around half among larger businesses in September. Yet smaller businesses are relatively more likely to report using AI to personalise products or support expansion. Large firms have more budgets and internal systems; small firms can sometimes change a workflow faster once a useful case appears.[3][4]

The OECD shows the same size gap internationally: 52% of large firms used AI in 2025 versus 17.4% of small firms. The gap is therefore both an access issue and an implementation issue. Cheap consumer tools reduce entry cost, but data readiness, skill, governance and time to redesign work still separate a trial from a productive system.[5]

The next growth statistic should measure completed work

User and seat counts will continue to climb. The more informative next indicators are the share of workers using AI on a normal week, the percentage of important processes with a defined AI role, accepted-task rates, human correction time, agent actions and the value of outcomes. Those measures reveal whether adoption is becoming deeper or simply broader.

The right conclusion is neither that AI adoption is hype nor that every business has already transformed. AI has reached mass awareness and mass trial. Paid use is expanding quickly. Deep, delegated and measurable use remains concentrated. That gap is now the central commercial opportunity—and the central reason current adoption percentages should be read with their denominators attached.

Report an error or suggest a correction