The AI Scale Divide: When Your Largest Suppliers Learn Faster Than the Rest
By Daniel Frost, Head of Supply-Chain Intelligence
On 5 November the Organisation for Economic Co-operation and Development (OECD) published its Digital Economy and Society Outlook 2026. Its press release gives the headline estimate: in 2025, 20 per cent of enterprises with 10 or more employees across the OECD used artificial intelligence (AI). Among large enterprises, those with 250 or more employees, the share was 52 per cent. Large firms, the release notes, "are more than twice as likely to use AI as the average firm".
In the World Research Institute's reading, AI adoption is becoming a matter of supplier segmentation: sorting suppliers into groups that are managed differently. If the gap persists, buyers may gain efficiency from their largest suppliers. Smaller and regional suppliers may fall behind on forecasting, quality control, documentation and customer response. More capability would then sit in fewer places inside each supply network.
The Gap Is Wider Than the Average Suggests
The report's statistical annex breaks the 20 per cent average down by firm size. Of small firms, with 10 to 49 employees, 17.5 per cent used AI. Of medium-sized firms, with 50 to 249, 29.5 per cent did. Large firms stood at 52.1 per cent, nearly three times the small-firm rate.
In relative terms, the gap is steeper still for firms that combine AI with two related tools. Cloud computing is computing power and software rented over the internet. Big data analytics is the analysis of very large datasets; for most European countries, the data cover data analytics in general. Large firms are more than four times as likely as small firms to use all three.
| Firm size | Used AI | Used AI, cloud computing and big data analytics together |
|---|---|---|
| Small (10 to 49 employees) | 17.5 | 9.5 |
| Medium (50 to 249 employees) | 29.5 | 21.0 |
| Large (250 or more employees) | 52.1 | 44.0 |
| All firms (10 or more employees) | 20.3 | 12.2 |
Per cent of enterprises across the OECD, 2025 or most recent year. Source: OECD, Digital Economy and Society Outlook 2026, Statistical Annex, Tables 1 and 3.
Looking ahead, smaller firms are expected to narrow the gap in cloud computing by 2030, the press release says. Not so in AI: "the divide in AI adoption is projected to persist without targeted support". By 2030, it adds, AI uptake among large firms could be nearly 80 per cent higher than the average across all firms. In the data behind the OECD's chart (Excel file), the 2030 projections are about 81 per cent for large firms and 46 per cent for all firms. That is a gap of some 35 percentage points, against about 32 in 2025. In relative terms, though, it narrows: the large-firm rate goes from about 2.6 times the average to about 1.8 times.
These are projections, not measurements. The OECD extends past adoption trends, and its methodology note says the paths "should be interpreted as informed estimates rather than precise predictions". The 2025 figures are estimates as well.
Cloud Access Is Not AI Capability
Cloud services have spread far wider than AI. More than half of all firms, and over 80 per cent of large firms, used cloud computing in 2025, the report's executive summary says.
Investment in AI computing is "booming", says the report's spotlight on the subject. One gauge is venture capital, the equity funding of young firms. On the OECD's broad, indicative measure of AI infrastructure firms, it rose from US$54.5 billion in 2024 to US$123.9 billion in 2025.
Yet access does not turn into capability by itself. About 18 per cent of firms use cloud services without big data analytics or AI, the report's chapter on technology diffusion finds. About 12 per cent use all three. Combining them, the OECD writes, requires "not only access to digital infrastructure, but also data readiness, specialised skills and complementary investments in organisational change".
Larger firms lead, the chapter says, "reflecting greater resources, technical capacity and access to skilled personnel". Smaller firms are held back by "limited access to finance, digital infrastructure and specialised skills", the press release says. It adds that firms also face "implementation costs, uncertainty about legal and data protection requirements, and difficulties integrating AI into existing systems". In the chapter, these are obstacles cited in a European Union (EU) survey by enterprises that did not use AI but had considered it. As use spreads, the chapter notes, "the key distinction becomes how well firms use AI rather than whether they use it".
Where the Gap Shows Up for Buyers
The OECD measures who uses AI, not how suppliers perform. What follows is our reading of where buyers may notice the difference. In the OECD's data on uses, firms most often apply AI to marketing and sales, business administration, finance and ICT (information and communication technology) security. Large firms report higher uptake for every purpose. Administrative uses include "document drafting, forecasting, human resources management and planning", the chapter notes. Sales uses include chatbots for customer support.
- Forecasting. A supplier that plans with AI tools may revise forecasts and confirm capacity faster. One that plans by spreadsheet may lag when demand shifts.
- Quality control. Automated inspection and data analysis may flag faults earlier. The OECD's data on uses do not single this out, so treat supplier claims here as unproven until tested.
- Documentation. Test reports and compliance files involve heavy document work. A supplier that assembles them with AI may deliver them faster, provided the underlying evidence is checked.
- Customer response. Order queries and claims are where chatbots and workflow tools sit. Response times may drift apart.
Sector and location add to the divide. AI remains concentrated in knowledge-intensive sectors, the chapter finds. Construction, transportation and storage, and accommodation and food services show comparatively low uptake. In the annex, 13.0 per cent of transportation and storage firms used AI in 2025, against 57.5 per cent in information and communication. In several Nordic countries, the chapter notes, more than one in three firms report using AI. In others, uptake remains below 10 per cent.
The Concentration Risk in Supplier Portfolios
In the World Research Institute's reading, the risk runs both ways. The first is over-concentration. Spend may drift to the few digitally advanced incumbents that seem to forecast, document and respond better. That can thin out alternatives, weaken bargaining power and tie the buyer to one supplier's systems and cyber defences. At country level, the OECD's press release notes that ICT value creation and sourcing are highly concentrated in a few countries, "reflecting both efficiency gains and cross-border interdependencies". The parallel with supplier portfolios is ours.
The second is drift in capability. A buyer keeps a diverse supplier base for resilience, but leaves the capability gap inside it unmanaged. The network may then plan, report and respond at the pace of its slowest members.
The OECD's figures show who uses AI and project how adoption may spread. They do not show that AI adoption alone raised any firm's productivity, or that it is consolidating supply bases. Those effects are for buyers to test, supplier by supplier.
A Supplier AI-Capability Screen Without the Hype
Our reading is that procurement teams should treat supplier AI capability as an operating attribute, like capacity or certification, not as a generic claim to innovate. Five questions separate the two:
- Use cases. Which named processes use AI today, and since when? A general claim to use AI is not an answer.
- Data access. What data does the tool run on, and can the buyer's data move in and out cleanly?
- Process integration. Is AI built into planning, quality and order systems, or is it a stand-alone tool?
- Human oversight. Who reviews outputs before they reach the buyer, and how are errors recorded?
- Pass-through. Can the supplier pass efficiency gains on without weakening evidence quality or cybersecurity?
The fifth question carries the most weight. A forecast produced faster is worth less if no one can trace how it was made. Where a supplier's AI output reaches people or published content in the EU, our analysis of the AI evidence chain explains why transparency evidence must travel with it.
Three Buyer Responses: Support, Substitute or Consolidate
In our reading, three responses cover most cases, and one portfolio may need all three.
- Support smaller suppliers that are hard to replace. Shared data formats, joint planning tools and training can narrow the gap. OECD Secretary-General Mathias Cormann named similar levers for policymakers: "wider access to cloud services, targeted support for AI and big data analytics, and stronger interoperability and open standards". Interoperability means that systems can exchange data.
- Substitute where a capability gap harms critical performance and support is not practical. Plan the transition and record the reasons.
- Consolidate spend with advanced suppliers only where the gain is measured, not promised. Keep qualified alternatives and exit terms.
Cormann said that "we need to make sure adoption spreads to firms of every size". Until it does, buyers need to know which of their suppliers it has reached.
Evidence and data cut-off: 6 November 2026. Third-party figures are as published by the cited sources. See our Research Standards.
The World Research Institute provides supply-chain intelligence, vendor due-diligence research and competitive intelligence programmes. Contact our team to commission tailored research.