Lack of skills holds back AI in EU food firms

In 2025, 69% of EU makers of food, drinks and tobacco products that had considered AI but were not using it gave a lack of relevant expertise as a reason, according to Eurostat's survey on ICT use in enterprises. Two years earlier the share was 78%. Over the same period the share citing cost fell from 55% to 41%, while the share citing unclear legal consequences rose from 38% to 44%. A policy brief published on 14 October by the European Commission's Joint Research Centre (JRC) finds the same pattern across the ten sectors it studied and names lack of expertise as the main barrier to further adoption.

Most food makers without AI have never considered using it

Of the food, drink and tobacco makers that did not use AI in 2025, 11.6% had ever considered it, up from 5.4% in 2023. The reasons in this article therefore come from about one non-user in nine. Among all EU enterprises the share was 14.2%, and it rises with size, from 12.7% of small firms to 36.5% of large ones. The survey covers firms with 10 or more employees, so it says nothing about farms or the smallest food businesses.

Cost and data worries have eased since 2023

Among food makers that considered AI, the share citing high costs fell from 55% in 2023 to 41% in 2025, difficulties with the availability or quality of data from 51% to 42%, and incompatibility with existing equipment, software or systems from 50% to 44%. The JRC sees the same movement in almost every strategic sector and reads it as a sign that AI has become more affordable, easier to access and easier to integrate with the systems firms already run. On the JRC's measure, the expertise barrier fell by 8.2 percentage points in agri-food, the largest drop of the ten sectors.

These are shares of firms that considered AI, a group that more than doubled as a proportion of non-users over the period. A falling share means a reason weighs less among those firms, and the JRC warns that it does not necessarily mean fewer firms face it.

Legal and data protection concerns have grown

Lack of clarity about the legal consequences of using AI was cited by 44% of food makers that considered it in 2025, up from 38% in 2023. Concerns about violating data protection and privacy rules rose from 32% to 40%. Both remain below the average for all EU enterprises, at 54% and 53%. Across the strategic sectors, the JRC found that regulatory uncertainty grew more than any other barrier between 2023 and 2025, and its brief calls for pilot projects and real-life experiments that give firms legal clarity.

The EU's AI rules have changed since the 2025 survey. Regulation (EU) 2026/1744, adopted on 8 July 2026, moved the obligations for high-risk AI systems to 2 December 2027 or 2 August 2028, depending on the category, and gave member states until 2 August 2027 to run at least one AI regulatory sandbox, a setting where AI systems can be tested under the supervision of the authorities. Since October 2025 the Commission's AI Act Service Desk has taken questions from firms and offered a checker that helps a firm see whether it is subject to obligations under the Act.

Wholesalers worry more about data protection than food makers do

Eurostat does not publish separate figures for food wholesale, but its figures for wholesale trade as a whole describe the trading end of the chain. In 2025, 23% of wholesalers used AI, against 12% of food, drink and tobacco makers, and 19% of wholesalers without AI had considered it. Among those that had, 73% cited a lack of expertise, 54% data protection concerns and 53% unclear legal consequences. Difficulties with data, at 47%, came ahead of cost, at 39%.

Hand holding a stylus over a tablet above a field of ripe cereal

In a JRC survey of 1,444 farms, high cost was the most common barrier to digital technologies, named by 53% of farmers

Farmers name cost before skills

Farms fall outside the Eurostat survey. The closest evidence is a JRC survey of 1,444 farms in nine member states, among them Greece, Spain and Italy, carried out between June and October 2024 and covering digital technologies in general rather than AI alone. High cost was the most common barrier, named by 53% of farmers, followed by a lack of knowledge about the technologies available (37%), a lack of government support (34%) and a lack of IT skills on the farm (32%). Only 9% named data privacy as a barrier to adoption, yet among farmers who share data electronically, 66% worry about privacy and security, 47% about how the data will be used and 37% about who owns them. The authors point out that the sample does not represent the EU as a whole, and that because AI relies on large-scale use of data, concerns about data security, ownership and control will need to be addressed if farms are to adopt it.

Since 12 September 2025 the EU Data Act has given users of connected products and machines, farmers among them, the right to access the data their equipment generates free of charge and to share it with third parties of their choice. One of the five checks worth making before investing in precision farming tools asks whether enough of the team understands the tool, the farm-level form of the expertise gap.

Each barrier in the survey can be checked before a firm adopts AI

Expertise: someone in the firm has to use the tool, check what it produces and correct it. Since the July 2026 amendment, the AI Act asks firms that use AI systems to take measures that support the AI literacy of their staff, without requiring any specific level, and asks the Commission and member states to help smaller firms do so. European Digital Innovation Hubs, half funded by the EU, offer training and the chance to test digital technologies before investing in them. The gap between people who know food and farming and people who know digital tools runs in both directions.
Legal consequences: a firm needs to know who answers for a price an AI tool quotes or a booking it makes, and who approves it before it reaches a customer or supplier. The answers belong in the supplier contract and in the firm's own approval rules.
Data protection: the question is which data leave the firm when staff use the tool, customer and supplier records included, and where they are processed.
Data and compatibility: a tool that cannot read the firm's existing records or connect to its systems adds work instead of saving it. Poor data quality carries costs of its own, with or without AI.
Cost: for food makers this is the barrier that has fallen most since 2023, which the JRC links to AI becoming more affordable.