Europe’s AI Divide Is Already an Enlargement Problem

Montenegro’s business register counted 63,823 active business entities in 2025. Of these, 61,195, or 95.9 per cent, were micro enterprises. The figures exclude entrepreneurs, so the weight of very small businesses is greater still.

This structure matters for Europe’s artificial intelligence (AI) debate. If the firms that dominate candidate economies cannot absorb new technologies, the distribution of AI productivity gains is not only an innovation issue. It is already an enlargement issue.

The first problem is measurement. Montenegro’s latest published survey of information and communication technology (ICT) use in enterprises covers companies with 10 or more employees in selected sectors. It reports universal internet access among the surveyed firms, and a website for 87.3 per cent of them. But it cannot show how firms below that threshold use AI, whether they reorganise work around it, or whether adoption improves business performance.

Eurostat’s headline figure that 20 per cent of EU enterprises used AI in 2025 is based on the same minimum-size threshold of 10 persons employed. The threshold makes statistics comparable, but it leaves much of the dominant firm type in candidate economies outside the evidence base.​

The productivity case is real, but conditional. A recent European Investment Bank (EIB) study of more than 12,000 non-financial firms in the European Union and the United States estimates that AI adoption raises labour productivity by about four per cent. In the short run, the result reflects capital deepening rather than job losses, although the longer-term employment effects remain uncertain. The gains are concentrated in medium-sized and large firms, and they are stronger where companies also invest in software, data, and training. These complementary investments are precisely what micro firms find hardest to finance and organise.

Support Schemes Exist, But Few Small Firms Use Them

Montenegro’s recent experience shows why the delivery mechanism matters. The 2024 assessment by the Organisation for Economic Co-operation and Development (OECD) reports that digitalisation-focused competitiveness programmes supported 353 small and medium-sized enterprises (SMEs) in 2022, with €2.04 million allocated: 324 companies received financial support and 29 used mentoring services. Yet little more than one third of surveyed companies knew that the programmes existed. The OECD described the number of beneficiaries as modest and awareness as insufficient. A scheme may be well designed and still fail to diffuse technology widely if the smallest firms do not know about it, cannot navigate the application process, or lack the capacity to turn a grant into organisational change.

Montenegro now has a delivery channel designed for this problem. MontEDIH, the country’s European Digital Innovation Hub, began in January 2025 as a four-year project with a budget of about €2 million. It supports SMEs and public bodies through pre-investment assessments, testing and demonstrations, knowledge and technology transfer, and financial and business advice. These services address costs that a small firm cannot easily absorb alone. The hub should therefore publish who uses them and what changes afterwards: firm size, sector, and municipality; whether a tool reaches a core business process; whether workers are trained; and whether use continues after public support ends.

Pre-accession support faces the same delivery problem. An EU-funded project under the Instrument for Pre-accession Assistance (IPA) has supported stronger business-development services and preparations for a national agency for micro, small, and medium-sized enterprises (MSMEs). At the project’s final conference in July 2025, Johann Sattler, the EU ambassador to Montenegro, said that only half the work had been completed and that the essential outcome was the agency’s full operational readiness. Montenegro’s Reform Agenda under the Growth Plan covers the green and digital transitions, private-sector development, and the business environment, and the Reform and Growth Facility behind it is performance-based. This allows small-firm diffusion to be treated as a measurable reform result rather than an assumed consequence of institution-building.

Accession is meant to support economic convergence. It also exposes firms to stronger competitive pressure within the single market. If enterprises enter that market with formally aligned rules but without the capabilities to use productivity-enhancing technologies, integration may widen the gap between a small group of prepared companies and the rest of the economy. Larger firms can spread the fixed costs of data preparation, cybersecurity, software integration, and staff training across more output. A micro enterprise must absorb many of the same costs with fewer workers, less specialised management, and weaker bargaining power with technology suppliers.

The distributional effects extend beyond firms. Higher productivity does not automatically translate into higher pay, and the EIB study does not quantify any such pass-through. But the evidence does raise a practical policy question: who receives the capital, advice, and training required to capture the productivity gain? If support is concentrated in already capable firms and in capital cities, workers and businesses elsewhere may face technological change without sharing proportionately in its benefits. For a candidate country, that is a regional-development and labour-market issue as well as a digital one.

Enlargement Funding Must Become Diffusion Policy

The first step is to measure participation by micro firms and entrepreneurs. Business digitalisation dashboards need a proportionate module for firms with fewer than 10 employees. Without that baseline, a programme can report hundreds of beneficiaries without showing whether it reaches the businesses that dominate the economy. Reporting should distinguish between attendance at an event, completion of a digital-readiness assessment, implementation in a core business process, and continued use after support ends.

Support must extend beyond initial interest and demonstration. Hubs and national programmes should finance diagnosis, data preparation, cybersecurity, software integration, independent advice, and worker training, not only licences. Small firms need one accountable contact who can connect technical support, finance, and skills. Sectoral groups in tourism, retail, agriculture, and manufacturing can reduce fixed costs by developing shared demonstrations, model procurement clauses, and vetted advisory pools for businesses with similar needs.

The Growth Plan provides leverage. Relevant milestones should track the share of supported firms that are micro or small, participation outside capitals and major business centres, implementation in a core business process, measurable improvements in time, quality, or error rates, and continued use after 12 months. Worker participation should also be recorded. Employees often know where information is duplicated and which tasks consume time without adding value; involving them early can improve both the investment decision and the training.

For candidate countries, these indicators should be built into monitoring before accession, while administrative capacity is being tested. The Commission and national authorities could use them in annual reform reporting and programme evaluations without imposing unrealistic adoption quotas. The objective is verified capability, not technology uptake for its own sake. Regional cooperation among Western Balkan hubs could pool scarce expertise and give small firms the specialised support that no national system can efficiently duplicate.

Europe does not need every micro enterprise to become an AI company. It needs ordinary firms in candidate economies to use the technology productively. Montenegro already has EU funding, a digital innovation hub, national reform commitments, and institutions being prepared for membership. The decisive question is whether these resources reach the shop, the hotel, the manufacturer, or the professional service firm without a digital department. Measuring outcomes rather than activity would let AI support convergence instead of embedding a new productivity divide before accession.