AI beyond the hype – 2025, the year AI becomes core infrastructure

AI governance and the EU AI Act: a regulatory roadblock? 

Paloma Laye, AI Quality Advocacy Manager, TÜV SÜD addressed a major regulatory milestone – the EU AI Act, the world’s first comprehensive AI legal framework. This legislation classifies AI systems into four risk levels, ranging from prohibited and high risk to low and minimal risk. 

Highly regulated industries, such as healthcare and law enforcement, must meet strict transparency and compliance requirements. While these measures uphold ethical standards and accountability, the ongoing debate around their implementation feeds concerns about potential challenges for innovation – particularly for start-ups navigating compliance complexities. 

The balance between self-regulation and legal oversight remains intricate. Large companies can and do develop in-house AI governance policies, while smaller firms often lack the resources to do so. However, companies that proactively integrate AI into their strategy can gain a competitive advantage rather than merely reacting to legal constraints.

Investments and trends: where is capital flowing? 

Jessica Lieber, Principal, 10x Founders provided compelling insights into the AI investment market. In Q4 2024, half of global venture capital was invested in AI, with a significant portion flowing into foundation models with general-purpose applications. In her trend forecast for 2025, she anticipates a rapid surge in funding rounds for industry-specific AI solutions – spanning both horizontal and vertical models. At the same time, embodied AI is on the rise, as AI advancements increasingly extend into the physical world. 

The reality of AI in enterprises 

Patrick Simon, Program Manager, New Digital Tech/GenAI, ZEISS highlighted the five biggest challenges companies face when implementing new digital technologies such as generative AI. Central to this is the balance between technological innovation and corporate responsibility. While companies seek to harness the potential of AI, they must simultaneously uphold high standards in data security and compliance. 

Moreover, successfully implementing AI systems requires more than just technical expertise. What truly matters are measurable business value, well-thought-out processes, and the systematic integration of employees. He particularly emphasised the importance of a well-balanced strategy: On the one hand, companies must be able to respond quickly to new technological developments; on the other hand, it takes time to embed lasting change within the organisation. 

Another key focus of his analysis was the role of employees in the transformation process. Instead of introducing AI solely from the top down, he advocates for a combined approach: Strategic initiatives from leadership should be complemented by bottom-up innovation, empowering employees across all departments to identify and leverage AI potential. 

Is the ‘iPhone moment’ for AI agents still a long way off? 

The panel discussion made it clear that AI agents are not yet indispensable in daily life – as noted by Jessica Lieber, the much-anticipated ‘iPhone moment’ has yet to arrive. However, as AI agents become capable of independently managing complex tasks like scheduling and decision-making, this will change. Adoption is expected to start at the individual level before spreading into business applications, as people who integrate AI into their daily routines will naturally seek similar support in the workplace.