OECD: AI Adoption by Small and Medium-Sized Enterprises

The AI Revolution Is Here — But Most SMEs Are Still on the Sidelines

Artificial intelligence is no longer a futuristic concept reserved for Silicon Valley giants. Between 2020 and 2024, the share of firms using AI across OECD member countries more than doubled — rising from 5.6% to 14%. Yet behind this headline growth lies a stark reality: the vast majority of small and medium-sized enterprises (SMEs) remain locked out of the AI revolution.

A landmark OECD discussion paper, prepared for Canada’s 2025 G7 Presidency, reveals the full picture of AI adoption among SMEs — the barriers they face, the productivity gains available, and the policy levers that could close the gap. For investors and business leaders watching the AI transformation unfold, understanding these dynamics is essential for identifying where value will be created — and where it will be destroyed.

This analysis draws on extensive data from across G7 nations, original OECD surveys, and real-world case studies to map the landscape of SME AI adoption. The findings carry significant implications for anyone with exposure to the small business economy — which, across the OECD, accounts for over 99% of all firms and the majority of employment.

AI Adoption by the Numbers: A Widening Gap Between Large Firms and SMEs

The data paints an unambiguous picture of divergence. In 2024, 40% of firms with 250 or more employees were actively using AI. Among mid-sized firms (50-249 employees), the figure drops to 20.4%. For small firms (10-49 employees), it falls further to just 11.9%.

This means large firms are more than three times as likely to use AI compared to small firms — a gap significantly wider than for other digital technologies like cloud computing or the Internet of Things, where small firms are roughly half as likely as large ones to adopt.

The gap persists even after controlling for sector, firm age, and asset composition. Larger firms simply have more resources, more data, and more capacity to experiment with new technologies. But size alone doesn’t explain everything — younger firms, including startups, show a notably higher tendency to adopt AI regardless of their scale.

Perhaps most telling: among SMEs that do use generative AI, only 29% report using it in core business activities. The majority confine AI to peripheral tasks — a pattern that limits the technology’s transformative potential.

Sectoral Winners and Laggards: Where AI Adoption Is Concentrated

AI adoption is not spread evenly across the economy. The ICT sector leads dramatically, with nearly 45% adoption in 2024 — rates often more than three times higher than manufacturing across G7 countries. Professional, scientific, and technical activities follow at over 25%.

At the other end of the spectrum, construction sits at just 7.2%, accommodation and food services at 7.8%, and transportation and storage at 9.2%. These are precisely the sectors where SMEs are most concentrated, deepening the adoption divide.

Between 2021 and 2024, most sectors doubled their AI adoption rates. The ICT sector surged by 20 percentage points, while professional and scientific services added 13 percentage points. Manufacturing and transportation showed lower relative increases of around 60-70%, suggesting that the sectors most in need of productivity gains are adopting AI most slowly.

The Productivity Promise: What AI Can Actually Deliver for Small Businesses

The potential rewards for SMEs that successfully adopt AI are substantial. OECD estimates suggest AI could add 0.2 to 1.3 percentage points in annual labour productivity growth across G7 economies over the next decade — comparable to the gains seen during the US ICT boom of the mid-1990s (0.5 to 1.5 percentage points).

At the firm level, the evidence is even more compelling. In France, the share of AI adopters in the top productivity decile was 40% higher than in the bottom decile. In Germany, it was 120% higher. In Italy, an astonishing 240% higher. AI users consistently demonstrate productivity premiums over 4%, with some studies finding gains exceeding 15% compared to similar non-adopting firms.

However, the OECD sounds an important caveat: productivity advantages shrink significantly once researchers account for the broader digital capabilities of firms and workers — including ICT skills, cloud computing adoption, and data management practices. AI doesn’t work in isolation; it amplifies existing digital maturity.

Furthermore, productivity gains may follow a J-shaped curve, where performance actually declines temporarily before improving. This pattern, well-documented in past technology transitions, means SMEs need staying power and realistic expectations about the timeline to returns.

Generative AI: A Potential Game-Changer for SMEs

The rise of generative AI tools like ChatGPT, Gemini, and Claude has fundamentally changed the accessibility equation. In the United States, approximately 40% of the population aged 18-64 used generative AI either at work or at home by late 2024, with over 22% of workers using it weekly for work tasks.

As of July 2025, ChatGPT alone had reached 700 million weekly active users, with 27% of 2.6 billion daily messages being work-related. The adoption curve for generative AI is steeper than for personal computers or the internet — suggesting rapid diffusion ahead.

For SMEs specifically, improved employee performance is the number one reported benefit of generative AI, followed by cost savings and the ability to perform entirely new tasks. A particularly striking finding: 39% of SMEs that use generative AI and have recently experienced a skill gap said that generative AI helped compensate for it — effectively serving as a force multiplier for understaffed teams.

In terms of innovation potential, around 70-80% of respondents in Germany, Italy, Japan, the UK, and the US noted that generative AI can boost innovation. Japan was especially optimistic, with over 90% of SME respondents highlighting this potential.

Four Types of SME AI Adopters: From Novices to Champions

The OECD proposes a useful taxonomy that categorizes SMEs along three dimensions: digital maturity, complexity of AI use, and scope of AI application across the business. This framework yields four distinct categories:

AI Novices are just beginning their AI journey. They rely on embedded AI features (“passive AI”) or off-the-shelf solutions like large language models for writing, marketing, and process optimization. A small coffee roaster in San Francisco exemplifies this category — using ChatGPT for product descriptions, SEO, marketing emails, and shipping cost analysis.

AI Optimisers are integrating off-the-shelf AI across multiple departments and experimenting with more advanced tools. An artisan bakery in Sèvres, France, uses ChatGPT for customer service, content creation, and recipe optimization, while leveraging MidJourney for visuals — all self-taught.

AI Explorers are developing custom AI models and training them on proprietary data. A micro wholesale company in Tokyo created custom AI agents for Q&A, project negotiations, and multi-language translated chat — driving increased revenues and shorter negotiation cycles.

AI Champions lead adoption with enterprise-wide AI deployment supporting both operational and strategic decision-making. A small healthcare company in Calgary uses LLMs, NLP, and computer vision for clinical note transcription and lab report analysis, while a Cambridge biotech has built a knowledge graph integrating 50+ data sources for drug discovery.

The Skills Barrier: Why 50% of SMEs Say Their Teams Aren’t Ready

Skills emerge as perhaps the single greatest barrier to AI adoption among SMEs. A full 50% of SMEs report that their employees lack the skills to use generative AI effectively, according to OECD surveys across four G7 countries.

The picture varies dramatically by country. In Japan, 80% of SME respondents cited lack of knowledge about how to use generative AI as a barrier. In Canada, the figure was 50%, while in the UK and Germany it was approximately 40%.

Despite these acknowledged gaps, training rates remain alarmingly low. Under 30% of SMEs using generative AI report that employees participate in AI-related training. Japan stands out at a mere 11.3%, followed by Germany at 23.2%, the UK at 24%, and Canada at 29.4%.

The skills that SMEs consider more important due to generative AI reveal the technology’s true nature: data analysis and interpretation (46.4%), creativity and innovation (41.9%), programming and coding (39%), and communication and collaboration (35.8%). AI doesn’t replace thinking — it demands more of it.

Infrastructure, Financing, and the Digital Divide Within the Divide

Even before skills enter the equation, basic connectivity infrastructure creates uneven playing fields. As of June 2024, the OECD averaged 36 fixed broadband subscriptions per 100 inhabitants, but the quality varies enormously.

Japan leads G7 nations with fibre representing 79% of total fixed broadband subscriptions, followed by France at 70%. Germany lags significantly at just 12.2% fibre — a critical disadvantage for AI workloads that demand high bandwidth and low latency.

The urban-rural divide adds another layer of inequality. Fixed download speeds in metropolitan areas were 44% higher than in regions far from urban centres by end of 2024. Mobile speed gaps between metropolitan and remote areas grew from 5 Mbps to 45 Mbps between 2019 and 2024 — the divide is widening, not closing.

On the financing side, AI adoption requires investment in tools, training, integration, and organizational restructuring. Yet the cost of SME financing has increased at a record pace in recent years, with sharp declines in SME lending. A troubling pattern has emerged: higher shares of smaller-scale, short-term financing for immediate needs rather than long-term investment.

There is an ironic twist: AI and machine learning are themselves transforming financial technologies, particularly credit scoring. Fintech platforms are enabling lending to businesses with limited credit histories — potentially creating a virtuous cycle where AI helps solve the financing barriers to AI adoption.

Government Policy and SME Concerns About AI

G7 governments are mobilizing substantial resources. Canada has committed CAD 2.4 billion over five years (2024-2029) for AI development, including a CAD 300 million AI Compute Access Fund and CAD 200 million through Regional Development Agencies.

Across G7 nations, the preferred support varies by country. Canada and the UK prioritize training programs. Germany favours information campaigns to raise awareness. Japan calls for direct financial assistance — reflecting the different challenges each country faces.

Even among AI-positive SMEs, serious concerns persist. Harmful content generation is the top worry for over 90% of US respondents and more than 80% in the UK and Canada. Inaccurate information — the hallucination problem — was flagged by over 80% of SMEs in Japan and the UK. Copyright and legal issues were cited by 90% of Canadian respondents.

Notably, 86% of SMEs report either neutral or favorable attitudes toward generative AI — meaning attitude is not the primary barrier. The obstacles are practical, not psychological: skills, cost, infrastructure, and perceived relevance.

Investment Implications and the Path Forward

For investors tracking the AI transformation, the OECD data points to several clear opportunities. First, the massive adoption gap between large firms and SMEs represents a market waiting to be served. Tools that reduce complexity and cost of AI adoption for small businesses have enormous addressable markets.

Second, the skills bottleneck creates opportunity in AI training and education. With under 30% of AI-using SMEs providing training, demand for accessible, business-relevant AI education will only grow.

Third, the infrastructure divide — particularly Germany’s fibre lag and the persistent urban-rural gap — points to continued investment needs in connectivity, especially fixed wireless access technologies growing at 39% year-over-year in the US.

The OECD’s analysis carries an urgent subtext. The productivity gains from AI compound over time, meaning SMEs who delay adoption don’t just fall behind — they fall behind at an accelerating rate. The generative AI wave offers a rare opportunity: unlike previous AI technologies requiring substantial technical expertise, tools like ChatGPT, Gemini, and Claude are accessible at consumer-grade pricing.

As OECD AI policy research consistently shows, the SMEs that experiment now will build the institutional knowledge needed to leverage more powerful AI tools as they emerge. The message is clear: the AI adoption gap is not inevitable. With the right combination of skills development, infrastructure investment, accessible financing, and smart regulation, small businesses can capture their share of the AI productivity dividend.