Europe does not have to choose between speed and trust. The real challenge is to apply AI governance proportionately, so that low-risk applications can be deployed quickly while higher-impact applications are subject to appropriate oversight.
Artificial intelligence is fundamentally changing how organisations operate. AI can summarise documents, prepare analyses, automate repetitive tasks and give employees faster access to relevant information. The potential efficiency gains are clear: less time spent searching for and processing information, and more time available for interpretation, advice and decision-making.
Rapid growth in business AI adoption
The use of AI is growing rapidly. In 2024, 78% of organisations surveyed worldwide reported using AI, compared with 55% in 2023. In the same survey, the use of generative AI in at least one business function increased from 33% to 71%.[1]
However, the financial impact often remains limited. Within business functions where organisations report cost savings from AI, these savings are generally below 10%. For revenue increases, the most commonly reported level is less than 5%.[2]
Europe opts for trustworthy AI: obstacle or advantage?
Europe explicitly links the use of AI to safety, fundamental rights, transparency and human oversight. The European AI Act takes a risk-based approach, treating prohibited practices, high-risk systems, applications subject to specific transparency obligations and applications presenting minimal or no risk differently.[3]
In addition, the General Data Protection Regulation (GDPR) continues to apply when an AI system processes personal data. Organisations must therefore take into account requirements including lawfulness, transparency, purpose limitation, data minimisation, security and accountability.[4]
Protection is crucial, but it can also slow things down
These safeguards are essential, but they can slow down adoption when organisations lack sufficient clarity about risks, responsibilities and permitted data. Before implementing an AI solution, organisations need to establish which data it will process, where that data is located, for what purpose it will be used and which security and oversight measures are required.
According to an IMF modelling study covering 31 European countries, AI could cumulatively increase total factor productivity in Europe by approximately 1% over five years. The authors emphasise that the outcome varies significantly depending on the scenario and the country.[5]
In one of the scenarios examined, the potential productivity gain falls by more than 30% when national and European rules relating to professional requirements, AI safety and data protection result in AI exposure being 50% lower across the affected tasks, occupations and sectors. This is a conditional model estimate, not a prediction that regulation will automatically reduce productivity gains by 30%.[5]
While Europe weighs its options, others are investing faster
The United States has a significant lead in investment. In 2024, private AI investment in the US reached $109.1 billion. That year, US-based institutions developed 40 notable AI models, compared with 15 in China and 3 in Europe.[1]
Meanwhile, China is narrowing the performance gap on key AI benchmarks and remains very strong in industrial automation. In 2023, China installed 276,300 industrial robots, accounting for 51.1% of global installations.[1][2]
This lead is not solely the result of differences in regulation. Capital, digital infrastructure, computing power, talent and a strong technology ecosystem also influence how quickly organisations can experiment and scale successful applications. These figures mainly demonstrate that regulation is just one element of a broader competitive position.
Speed and trust
Europe does not have to choose between innovation and protection. The challenge is to apply obligations proportionately, based on the purpose, context and risk of the application. An internal application that structures text or makes knowledge more accessible will generally require a different assessment from a system that supports or influences decisions about individuals. The exact classification will continue to depend on the intended use and the applicable regulatory framework.[3][4]
Organisations can accelerate adoption themselves by establishing clear ground rules in advance: which tools are approved, which data may be entered, when a risk or privacy assessment is required, who remains responsible for the output, and how incidents or errors should be reported.
This creates room for safe experimentation while ensuring that higher-impact applications are subject to stricter controls in a timely manner.
What does this mean in practice for organisations?
- Maintain an overview of approved AI tools and permitted data categories.
- Apply more rigorous assessments as the potential impact on individuals, processes or rights increases.
- Maintain human oversight and clear accountability for business decisions.
- Measure not only usage, but also quality, time savings, costs, risks and value actually delivered.
Sources
[1] Stanford Institute for Human-Centered Artificial Intelligence, The 2025 AI Index Report, 2025. Used for figures on AI adoption, private investment, notable AI models and China's position.
[2] Stanford Institute for Human-Centered Artificial Intelligence, The 2025 AI Index Report, Chapter 4: Economy, 2025. Used for the financial impact by business function, regional adoption and industrial robot installations.
[3] European Commission, AI Act: Regulatory Framework for Artificial Intelligence. Used for the description of the risk-based approach.
[4] Belgian Data Protection Authority, Artificial Intelligence Systems and the GDPR: A Data Protection Perspective, December 2024. Used for the interaction between the GDPR and the AI Act.
[5] Florian Misch, Ben Park, Carlo Pizzinelli and Galen Sher, Artificial Intelligence and Productivity in Europe, IMF Working Paper No. 2025/067, 3 April 2025.