AI adoption doubled in a year. Nearly half of adopters spent nothing on it.

The UK government published the second wave of its AI adoption research last week: a telephone survey of 3,503 UK businesses, fielded between January and April 2026. The headline is real - adoption doubled, from 16% of businesses using at least one AI technology in 2025 to 32% in 2026. The more interesting numbers are underneath it. Of adopters able to estimate their AI expenditure, 47% reported spending nothing at all in 2025. Only 14% of all businesses consider AI tools of high or critical importance to their operations.

That combination - doubling usage, near-zero spend, low strategic weight - is the actual state of enterprise AI right now. Not transformation. Diffusion.

What adoption looks like up close

The growth is concentrated where you'd expect. Large firms are at 52% adoption. Information and communication leads sectors at 61%, followed by real estate (45%) and business services (39%). The tools are mostly off-the-shelf and low-complexity: among businesses using or planning to use AI, 77% rely on LLM-based text generation, applied to administration (81%), marketing (71%) and IT functions (66%). The top driver is efficiency and productivity (71%) - not new products, not new business models.

Usage is frequent but shallow. A majority of adopters (57%) say they use AI constantly, yet most describe themselves as early-stage or exploratory. Half of adopters report an overall productivity increase - described as modest. Nearly two-thirds (64%) have seen no immediate effect on operating costs. The most common benefits are task automation and workflow optimisation, the 2023 definition of AI value.

The pattern isn't British. The OECD's 2026 D4SME survey - over 2,000 small and medium-sized enterprises across 12 countries, a non-representative sample the OECD itself flags - finds the same shape: rapid take-up of off-the-shelf AI tools, uneven strategic integration, and persistent drag from time constraints, maintenance costs and skills gaps.

The real constraint

The barriers section of the UK report is more informative than the adoption numbers. Limited skills, expertise or knowledge affects 67% of businesses that have adopted AI and 78% of those planning to. Among non-adopters, 65% say they lack a clear business case or proven return on investment. Cost matters, but mainly as the time, effort and resource required to embed the tools - not the price of the tools themselves. Meanwhile 65% of businesses expect their AI budgets to grow.

Trust remains conditional: 81% of businesses apply human oversight to AI outputs, though that figure is slipping. Agentic AI is the least trusted category - more scrutiny, more caution.

My read

Three implications for leaders. First, recalibrate your competitive anxiety: if a third of businesses "use AI" and nearly half of those spend nothing on it, the adoption race you think you're losing barely exists. The bar is depth, not adoption. Second, the 14% figure is the opportunity. Value is concentrating in the minority that treats AI as critical infrastructure - embedded in core workflows, measured, resourced. Everyone else is running free-tier experiments. Third, spend on skills before licenses. The data says the binding constraint is people who can turn a general tool into a specific workflow, not the tool. The businesses that close that gap in the next two years - while their peers double adoption at zero cost - will own the productivity gap that follows.

Sources

• UK Department for Science, Innovation and Technology, AI adoption survey report 2026: executive summary (published 29 October 2026): gov.uk
• OECD, Empowering SMEs in the age of AI (2026 D4SME survey, published 13 April 2026): oecd.org

Full disclosure: Ganesh uses Instinct (instinct.ai), an AI assistant, to research, write and publish this blog series.