AI adoption is highest in information, communications, finance, professional services, and other knowledge-intensive industries. In May 2026, US Census data showed 39.7% adoption in information and 33.9% in finance and insurance, versus about 14% in retail. Industry differences reflect task fit, firm size, digital infrastructure, skills, and the survey definition of AI use.
AI
AI adoption is highest in information, communications, finance, professional services, and other knowledge-intensive industries. In May 2026, US Census data showed 39.7% adoption in information and 33.9% in finance and insurance, versus about 14% in retail. Industry differences reflect task fit, firm size, digital infrastructure, skills, and the survey definition of AI use.
Key findings
• The US information sector had a 39.7% reported AI-use rate as of May 3, 2026, and finance and insurance had 33.9%, both above the 19.8% national rate.
• Retail trade was below the US average at roughly 14%, with about 17% expecting to use AI within six months.
• OECD 2025 data placed AI adoption at 57.3% among ICT firms and 36.8% in professional and scientific services.
• Previously low-adoption sectors were growing quickly from small bases. OECD reported year-over-year adoption growth of 62.5% in accommodation and food services and 59.1% in construction.
• Across EU enterprises, written-language analysis was the most common measured AI technology at 11.8% in 2025.
• Sector adoption is not a direct measure of value. A high adoption rate can consist of many shallow uses, while a low-adoption sector can have a small number of high-value workflows.
Methodology and definitions
This report compares industry results from four official statistical systems: the US Census Bureau, OECD, Eurostat, and the UK Office for National Statistics.
The US Census Bureau’s Business Trends and Outlook Survey is a biweekly, nationally representative survey of US employer businesses. Its AI question was revised in November 2025 to ask about use in any business function rather than use in producing goods or services.
OECD figures aggregate national statistical data from countries where comparable information is available. Eurostat’s enterprise survey covers firms with at least 10 employees in specified economic activities. ONS data cited here also focus on UK businesses with at least 10 employees for many of its published comparisons.
Adoption rate means the share of firms in a sector reporting qualifying AI use. It is not the share of employees, workflows, revenue, or output affected by AI.
Year-over-year growth is the percentage increase in an adoption rate. A sector moving from 5% to 8% has 60% growth but remains at 8% adoption. Growth and level must be shown together.
Industry and business function are different dimensions. Retail is an industry. Marketing is a function that exists in retail, manufacturing, software, and other industries. A sector with limited AI in its core production can still use AI in support functions.
US industry adoption
The Census Bureau’s May 2026 BTOS analysis reviewed data from December 14, 2025, through May 3, 2026. Overall AI use ranged between 17% and 20% over the period. In the final collection window, the national estimate was 19.8%.
Information led at 39.7%, about twice the national rate. Finance and insurance followed at 33.9%. Neither sector’s change from December to May was statistically significant, so the evidence supports a high level, not a claim of rapid recent acceleration.
Expected use was also elevated. About 42% of information businesses and roughly 39% in finance and insurance expected to use AI in the following six months.
Retail trade reported around 14% current use and 17% expected use. That lower level is plausible because a larger share of retail work involves physical operations, store execution, inventory movement, and customer interaction that cannot be changed by a standalone language model. Retail still has applicable workflows in merchandising, support, demand analysis, product content, fraud review, and back-office operations.
The survey asks whether a business used AI, not how much value it obtained. Sector comparisons should guide workflow discovery, not produce a league table of management quality.
OECD industry patterns
The OECD’s January 2026 statistical release found that 20.2% of firms used AI in 2025, up from 14.2% in 2024. ICT had the highest sector rate at 57.3%, followed by professional and scientific services at 36.8%.
In several countries, ICT adoption approached saturation under the survey definition: 87.9% in Sweden, 79.9% in Austria, and 79.8% in Finland. Those country-sector values should not be generalized to every ICT firm worldwide, but they show what adoption can look like in digitally mature markets.
The fastest relative growth appeared in sectors that had lagged. Accommodation and food services grew 62.5% year over year, while construction grew 59.1%. These figures describe growth rates, not adoption levels. They should be framed as diffusion into new sectors rather than evidence that either sector overtook ICT.
For SMBs, the important inference is that AI relevance is expanding beyond software. A restaurant may use it for menu translation, scheduling support, guest-message drafting, or review analysis. A construction company may use it for document classification, bid summaries, change-order intake, or project reporting. The production work remains physical, but supporting information workflows can change.
European technology and country differences
Eurostat reported that 20.0% of EU enterprises with at least 10 employees used AI in 2025, up from 13.5% in 2024. Country rates ranged from 42.0% in Denmark, 37.8% in Finland, and 35.0% in Sweden to 5.2% in Romania, 8.4% in Poland, and 8.5% in Bulgaria.
The most common measured technology was AI for analyzing written language, used by 11.8% of EU enterprises. Eurostat’s broader digitalization statistics also reported 9.55% using AI to generate images, video, or audio and 8.76% using AI to generate written or spoken language or programming code.
These technology categories help explain industry patterns. Sectors with large document, communication, analysis, and code workloads have more immediately addressable tasks. That does not establish that those tasks should be automated without review.
UK evidence on depth
The ONS July 2026 business analysis reported that nearly three-fifths of information and communication businesses used AI, compared with about 35% across businesses with at least 10 employees.
Across all included businesses, large language models were the most widely used category at 18%, followed by visual content creation at 16%, machine-learning data processing at 12%, and machine-learning image processing at 6%.
Depth remained limited. Only 10% of businesses that used at least one AI technology said they used AI extensively. The finding prevents a common mistake: taking a high sector adoption rate as evidence that AI has been integrated throughout the operating model.
Practical SMB implications by sector
An SMB should begin with the information structure of its work, not its industry label. Map four elements:
1. High-volume text: emails, tickets, notes, forms, proposals, policies, or contracts.
2. Repeated decisions: classification, routing, prioritization, matching, or exception detection.
3. System actions: CRM updates, task creation, notifications, or document movement.
4. Human boundaries: pricing, legal judgment, payments, safety, employment, or sensitive customer decisions.
Then select a workflow whose output can be checked. Industry benchmarks can indicate where peers are experimenting, but department-level AI agent use cases are a more practical starting point because sales, support, finance, HR, and operations recur across sectors.
For an information-services firm, the first workflow might be research intake or knowledge retrieval. For finance, it might be document triage with strict approvals. For retail, product-data cleanup or customer inquiry routing may be safer than autonomous pricing. For construction, begin with document and reporting flow rather than site-safety decisions.
Use sector data to set expectations, not to justify a project. Approval should still depend on workflow volume, baseline cost, output quality, data access, human review, and measurable value.
Limitations
Industry classifications and covered firms differ across countries. Census includes US employer businesses; Eurostat and many ONS tables apply minimum employee thresholds. OECD aggregates national data with available comparability.
The Census wording changed in November 2025. Growth rates can look large when starting levels are small. Survey responses do not reveal whether use is approved, frequent, integrated, or valuable. Country differences can reflect economic structure as well as adoption behavior.
What would change our view
We would conclude that industry gaps were narrowing if low-adoption sectors showed sustained increases in both level and depth, especially recurring core workflows and employee coverage. We would conclude that adoption was stalling if reported access rose while extensive use, workflow breadth, and production outcomes remained unchanged.
FAQs
Which industry has the highest AI adoption in 2026?
Information and communications leads most official datasets. US Census reported 39.7% in information, while OECD reported 57.3% among ICT firms across countries with available 2025 data.
Why is AI adoption high in finance?
Finance has large volumes of digital documents, structured data, analysis, service interactions, and compliance work. It also has substantial technology budgets. High risk means many workflows still require strong human review.
Is retail behind in AI adoption?
US Census reported retail use around 14% in May 2026, below the 19.8% national rate. Retail has relevant AI workflows, but much of its work is physical and its firm population includes many smaller businesses.
Does a high industry adoption rate mean high ROI?
No. Adoption records qualifying use, not value. A sector can have broad shallow use, and a low-adoption sector can have a few high-return workflows.
How should an SMB use industry benchmarks?
Use them to identify likely workflows and realistic adoption context. Make investment decisions from your own baseline volume, cost, quality, risk, data, and review requirements.
Get a 20-Minute AI Workflow Audit
AI Operator can compare your workflow with relevant industry and department patterns, then define the smallest production use case worth testing.