2
Adoption

The implementation of AI in European businesses continues to evolve, despite challenges along the way.

From challenges to opportunities: How EY and AI work hand in hand

2.

Adoption

Beyond the specifics of where AI can be used, what a lot of decision-makers want to know is the broader success factors for getting the most long-term value from implementation. First and foremost, they want to determine the technological foundations required for AI, with a strong focus on data and the cloud, and their attention is on making investments that deliver value. Stakeholders also note the importance of workforce buy-in and adoption to ensure success at scale, with the primary focus typically on building employees’ confidence to use AI to improve their day-to-day efficiency.

The adoption of AI in European businesses continues to evolve, albeit with its fair share of challenges. While the potential benefits are undeniable, organizations are still grappling with operationalization. Barriers to adoption include internal policies, external regulation, and the complexity of the new technologies.

Although some knowledge workers have been keen to experiment with AI, others are still reticent. In a recent survey, EY teams asked members of the workforce from all over Europe to share their experience with AI. Almost three-quarters of all respondents (73%) already have hands-on experience with AI technologies. Most of them use AI in their private lives (38%), rather than at work (12%). The remaining respondents (23%) have AI experience in both spheres.

From a regional perspective, the share of AI early-adopters is highest in Spain (84%), followed by Switzerland (82%) and Italy (77%). On the other end of the scale, early adopters are less common in the Netherlands (66%) and Germany (67%). Men (75%) have experience using AI applications more often than women (70%). Differences are also evident between ranks, with more than 84% of managers saying they use or have used AI applications, compared with just 67% of respondents among non-executive employees. In the following we look at some of the most common barriers to AI adoption in a work setting.

Security, accuracy, and explainability are viewed as crucial factors on the road to success when implementing AI – more so with respect to business viability than ethics. At the top of the agenda are security and privacy, discussed from a technical as well as a responsibility angle. Secure internal GPT models as well as the need to comply with regulation (current and future) around data and AI are of high importance to stakeholders and decision-makers.

Hurdles to adoption

Organizations face a number of hurdles in the operationalization of AI. For one thing, companies themselves often impose restrictive policies on AI use by employees. In some countries, only a relatively low share of employees in the EY survey report being permitted by their employer to use AI applications, most notably in Germany (42%) and Austria (46%). That contrasts markedly with the situation in Switzerland (73%) and Spain (63%), where most employees are permitted by their employers to use AI applications in their work. The share is likewise relatively high in Portugal (58%), Belgium (57%), Italy (56%), and France (55%), all of which exceed the European average of 52%.

Complexity is another issue organizations are grappling with. Six out of ten respondents (67%) point to the complexity of the implementation process for AI systems in their organization. Adding another layer of complexity subject to increasing regulatory scrutiny, and Europe is no exception (see EU AI Act).

Another important factor is the tone from the top. In Switzerland, most respondents (56%) give their employer a good report card when it comes to the extent to which management has a positive attitude as regards making progress with AI applications. However, only 5% describe their employer as very open when it comes to implementation. Switzerland ranks lowest in the category, together with Germany.

The level of adoption of AI varies across Europe, with some countries and sectors embracing it more readily than others. However, regardless of geographical location or industry, a clear tone from the top is essential for successful integration. Leaders need to champion AI initiatives, fostering a culture that encourages experimentation and innovation while addressing justified concerns about job, displacement, ethical and legal considerations.

Regulatory framework

The European Union’s AI Act unifies how AI is regulated across the single market of the 27 EU member states. It also has important extraterritorial implications, as it covers all AI systems impacting people in the EU, regardless of where these systems are developed or deployed from. The AI Act aims to standardize the use of AI across all its member states. Ratified by the European Parliament on 13 March 2024 and approved by the Council of the European Member States on 14 May 2024, the act is expected to enter into force in June 2024. It introduces a new regulatory framework for AI technology focused on the protection of safety, security and fundamental rights of people in the EU.

Risk and compliance framework

The AI Act adopts a risk-based approach to compliance obligations, categorizing AI systems by application areas and target groups into distinct risk levels. In this tiered compliance framework, most requirements fall upon the developers and deployers of AI systems classified as “high-risk”, and on general-purpose AI models (including foundation models and generative AI) deemed to pose “systemic risks”. For instance, low-risk AI such as chatbots used in customer service will be subject to few requirements beyond notifying users that they are interacting with AI. AI intended for high-risk application areas that may impact health, safety, or fundamental rights of people will have to comply with stricter controls,while some applications areas, such as subliminal manipulation of vulnerable groups, are outright prohibited.

To comply with the AI Act, companies will need to clearly assign within their organizations responsibilities for overseeing AI deployment and compliance. The mandated responsibility extends beyond technical departments to encompass the entire corporate fabric. Non-compliance exposes companies to severe risks, including heavy penalties with maximum fines that even surpass the maximum fines under the EU’s General Data Protection Regulation (GDPR).

Compliance strategy and adaptation

A strategic approach to implementation of AI Act compliance begins with companies identifying gaps in their current practices and outlining a meticulous plan customized to the specific manifestation of AI deployment in their organizations. The approach involves an as-is assessment encompassing current procedures, employee training levels, and a technical understanding of AI solutions, including an exhaustive inventory of the AI solutions deployed in the organization, what they are used for by whom.

Given the phased transition period, with enforcement of the AI Act prohibitions taking effect within six months, obligations for General Purpose AI mode developers starting after one year and most of the obligations for high-risk AI applications coming into force after two years, companies must hasten to adjust their operations and implement the required changes in a phased, monitored process. Not only is initial compliance by the end of the implementation deadline critical, it also needs to be accompanied by a sustained commitment to adapt to ongoing legislative amendments and to provide staff with appropriate training at regular intervals. In this way, companies can align with the EU’s goal of ensuring safe AI use without stifling innovation.