SMEs in the G7 are falling behind larger firms in adopting AI, constrained by connectivity, skills, data inputs and financing. The OECD recommends targeted policies and a taxonomy to speed SME AI uptake and ensure productivity gains benefit all.

Main Finding

AI adoption by small and medium-sized enterprises (SMEs) in the G7 and OECD is rising rapidly but remains substantially lower than in large firms and behind other digital technologies. Closing this adoption gap—and promoting complementary investments in connectivity, data/compute inputs, skills and finance—will be key to realising AI’s productivity potential across economies.

Key Points

• Adoption trends
  • OECD average share of firms (≥10 employees) using AI rose from 5.6% in 2020 to 14% in 2024.
  • Adoption in core business functions remains low in the G7 (2024): range ~1.9% (Japan) to 6.1% (US).
  • Consumer and worker uptake of generative AI has been much faster (e.g., ~40% of US adults 18–64 used generative AI by end‑2024; ChatGPT had ~700 million weekly active users in July 2025, with ≈27% of messages work‑related).
• Size and sector gaps
  • Large firms (≥250 employees) use AI far more: OECD 2024 figures — 40% (large) vs 20.4% (50–249 employees) vs 11.9% (10–49 employees).
  • In many metrics large firms adopt AI about three times more than small firms.
  • Sectoral heterogeneity: ICT and professional/freelance services lead; manufacturing and other sectors lag.
• Productivity potential and constraints
  • Growing evidence of a strong positive association between AI use and firm productivity.
  • OECD macro estimate: AI could raise annual labour productivity growth in G7 economies by ~0.2–1.3 percentage points over the next decade.
  • Realising gains requires complementary investments (skills, data, compute, organisational change), not AI tools alone.
• Key enablers (four success factors)
• Net connectivity (quality and geographic parity of internet access).
• AI inputs (access to high‑quality datasets, compute/cloud resources).
• Skills and human capital (technical and adoption/management capabilities).
• Finance (access to funding for long‑term digital transformation).
• SME typology for targeted policy
  • Proposed taxonomy (based on digital maturity, complexity and scope of AI use): KI‑Neulinge (AI novices), KI‑Optimierer (optimisers integrating multiple tools), KI‑Entdecker (developers of tailored solutions), KI‑Champions (AI integrated in strategy).
• Risks and barriers
  • Accuracy, harmful or biased outputs, and legal/regulatory uncertainty are recurring concerns for SMEs.
  • Skill shortages and financing constraints are repeatedly cited as major obstacles.
• Policy responses
  • G7 countries deploy multi‑dimensional strategies: infrastructure investment, skills training, data access, financial instruments, regulatory guidance.
  • International cooperation (aligned with OECD AI Principles) can reduce frictions and support responsible diffusion.

Data & Methods

• Data sources
  • OECD ICT Access and Usage by Business Database (primary source for firm‑level AI adoption trends, 2020–2024).
  • Eurostat enterprise surveys (cross‑country application breakdowns, 2024).
  • Country‑level studies and case studies across G7 members; platform usage reports (OpenAI, Anthropic) for generative AI metrics.
• Coverage and units
  • Analysis focuses on enterprises with ≥10 employees (standard OECD firm threshold in the dataset).
  • Statistics reported as unweighted averages across OECD members unless otherwise noted; G7 country comparisons use available national data (see annex tables referenced in the paper).
• Methods
  • Descriptive statistics and cross‑sectional comparisons by firm size, sector and country.
  • Development of a pragmatic taxonomy based on three dimensions: digital maturity, complexity of AI use, and breadth/scope of application.
  • Illustrative firm‑level case studies to show adoption paths and highlight success factors and challenges.
• Limitations noted by authors
  • Cross‑country comparability issues due to differences in survey definitions and timing.
  • Rapid evolution of tools (especially generative AI) means measures can lag current usage patterns.
  • Adoption metrics do not always distinguish between embedded third‑party tools and bespoke AI system development.

Implications for AI Economics

• Distributional effects and productivity
  • Persistent adoption gaps suggest uneven distribution of AI’s productivity gains across firms and regions; without policy correction, AI could widen firm‑level and regional inequality.
  • Estimations of economy‑wide productivity gains depend crucially on SME diffusion and on complementary investments; economic models and forecasts should incorporate adoption frictions and complementarity effects.
• Role of complementary capital
  • AI is a general‑purpose technology whose returns hinge on investments in data, compute, skills and organisational change—models of AI’s economic impact must endogenise these complementarities.
• Policy targeting and measurement
  • The proposed SME taxonomy provides a framework for targeted interventions and for more granular evaluation of policy effects; empirical work should disaggregate SMEs by adoption stage rather than treating them as homogeneous.
  • Improved, harmonised firm‑level metrics (distinguishing core business integration, type of AI, and degree of customisation) are needed to refine impact estimates.
• Speed of diffusion and structural change
  • Generative AI’s rapid household and worker uptake implies faster diffusion channels (via employees, off‑the‑shelf tools) than prior technologies—short‑run labour reallocation and task‑level substitution effects may be larger and faster than earlier forecasts assumed.
• Risk, governance and innovation incentives
  • Economic analysis must incorporate regulatory uncertainty, liability risks and trust considerations, since these affect firm incentives to adopt and to invest in higher‑risk, higher‑return AI projects.
  • International coordination on standards and data access can reduce cross‑border frictions and increase returns to AI investments, affecting trade, competition and global value chains.
• Research priorities
  • Micro‑level causal studies on AI adoption and productivity in SMEs (randomised trials, matched firm panels).
  • Quantification of the role of each success factor (connectivity, inputs, skills, finance) in enabling adoption and productivity gains.
  • Modelling distributional outcomes and optimal policy bundles for inclusive AI diffusion.

If you want, I can produce a one‑page executive summary in German tailored to policymakers, or extract the specific policy recommendations and country examples listed in the paper.