Navigating Compliance: How innovation can thrive under the AI Act

The recent entry into force of the European Artificial Intelligence (AI) Act, enables a conversation on the European strategy for AI, its feasibility and costs for businesses, particularly Small and Medium Enterprises (SMEs) and startups, as well as how the strategy will be implemented to promote innovation without hindering it. Intellera’s consolidated expertise in the field illustrates that a holistic approach that considers long-term sustainability, efficient implementation, and future tech developments is necessary for Europe to stay at the forefront of the digital race and regulatory innovation.

Giovanna Galasso is a Partner at Intellera Consulting. Sara Mancini is a Senior Manager at Intellera Consulting. Pablo Ivankovich is a Senior Associate at Intellera Consulting.

Europe’s vision of AI is twofold, enhance the European ecosystem of AI excellence and foster a trustworthy adoption that safeguards everyone’s fundamental rights. These two directives drive investment and regulatory initiatives such as the AI Act. Which is at the center of Europe’s strategy to differentiate its market from other regions, stimulating the emergence of new approaches to research and industrial development of trustworthy AI solutions, without hindering innovation. The aim is increasing trust in AI-solutions, thus, accelerating their adoption. In parallel, the expectation is to influence R&D through guardrails that facilitate the discovery of innovative and safer solutions.

While a clear regulatory framework is necessary, the economic impacts of complying with it cannot be denied, in particular considering the European SME ecosystem. According to Eurostat, in 2023 the adoption of AI in European SMEs was around 10 percent. This indicates that, while innovation looms large over our lives and everyday activities, there is still ground to cover in terms of AI adoption in the industrial sectors. At the same time, this poses an opportunity to facilitate the introduction of AI solutions that are compliant by design, effectively reducing costs of compliance.

What does it mean to be compliant? In a nutshell, it means

• regularly and centrally monitoring existing AI solutions;
• monitoring their risk-level;
• collaborating and establishing new and improved governance mechanisms with suppliers;
• adjusting and integrating a number of roles and processes, starting from risk management, development of AI solutions, to data governance; and, last but not least;
• adopting specific technology solutions to support the new AI-governance process.

At Intellera, we have been working in this field for years. Already in 2022 we conducted a study building on the “Study to Support an Impact Assessment of Regulatory Requirements for Artificial Intelligence in Europe” (IARR), where we modelled total costs on a generic SME active in the technology sector to estimate compliance costs. We found that such an SME would incur costs between 2.7 percent and 1 percent of its revenue in order to comply with the AI Act. More recently, we drew up a two-year action plan for a medium-sized Italian company in the utilities sector to implement the actions needed to comply with the AI Act requirements in a timely manner. We estimated around 5 Full-Time Equivalents (FTE) per year across the organization, considering also an AI governance framework and best practices. Notably, external and technology development costs, such as acquiring an AI governance tool, were not included in the estimate.

To mitigate the compliance burden, SMEs and innovators alike can access a number of European and national initiatives designed to increase AI adoption and reduce compliance costs. Among the most relevant:

• The European Digital Innovation Hubs (EDIH), over 220 in Europe, of which around 180 specialize in AI. The EDIHs are the result of combined European and national funding schemes, with the aim of helping SMEs, startups and Public Bodies in their digital transformation journey, ensuring 90% of companies have a basic level of digital know-how, and expected to integrate services for regulatory compliance.
• Testing and Experimentation Facilities (TEF), also co-funded by the European Commission and Member States for about EUR 60 million each, offering physical and virtual spaces for testing AI solutions in real-world environments, thus, crash-testing innovation to ensure it is safe and robust for their own specific vertical sectors.
• TEFs will also contribute to the implementation of the AI Act by supporting national regulatory sandboxes on AI. Regulatory sandboxes on AI are to be established within the next 2 years, with the aim of helping SMEs and startups experiment, ensuring compliance with existing regulations and in particular guaranteeing solutions that are compliant with the AI Act.

On top of these, other European Commission initiatives on AI include the AI Pact, the AI Innovation Package, Data Spaces, EDICs (European Digital Infrastructure Consortium), the AI-on-Demand Platform, and AI Factories, as well as encouraging the European standardization process that operationalizes the regulatory framework.

Considering these initiatives, if AI rate of adoption in enterprises remains low, how effective are they really? Our experience enables us to offer some insights:

• European businesses will more often than not be users of AI, rather than developers. Therefore, initiatives should adapt their services to the needs and realities of companies, fostering technology transfer of European solutions into the market.
• To facilitate access, Europe needs to simplify its offering, making differences and potential synergies between initiatives clear, thus, fostering collaboration and asset-sharing, knowledge and best practices. To do so, effective governance and coordination between European and Member State levels is essential.
• Similarly, coordinated and targeted communication campaigns to increase awareness and access to services is fundamental, while establishing mechanisms to monitor the effectiveness and dissemination of services available to SMEs and startups.
• Lastly, there is a need to ensure medium- to long-term sustainability of these initiatives beyond European and national funding, possibly exploring mechanisms for private partnerships or self-sustainment, ensuring continuity of the built knowledge base.

To conclude, it is fundamental to consider the big picture. Today artificial intelligence is at the center of these debates as it is the technology that has seen exponential growth and evolution in recent years. However, we can expect that, in the future, other technologies such as quantum computing or advanced neuroscience techniques will follow a similar path. It is, therefore, of the utmost importance to define an approach that is “future-proof,” that is, flexible and effective for the new and emerging innovations of tomorrow.