The manufacturing sector in particular has realized that GenAI, beyond its role in internal or service activities, also has positive effects on product development, procurement, and business models. However, the diverse application potential of AI is being increasingly recognized across all industries. Whereas in previous surveys it was mainly internal areas and customer service that were mentioned, AI applications are now either already in use or planned for 2024 in almost every corporate division. 

Previously known mainly as an efficiency booster, AI is now attributed significant influence in the dimensions of decision support, sales potential, and overall performance. 

When asked about what companies are currently working on and their plans for 2024, a substantial package emerges. Three-quarters of companies intend to use the upcoming months to intensively analyze the utilization potential available to them. Both during the analysis phase and the subsequent implementation, the majority are planning on seeking external support. 

Three out of four companies are also aiming to conduct comprehensive leadership training with a focus on AI. Cross-organizational further training measures and targeted hiring of AI experts are planned in the same percentage. And what is reassuring is that an equal number of companies are committed to developing a specific roadmap for enterprise-wide utilization of AI applications and for scaling up. 

Furthermore, there are five additional measures pursued by at least half of the companies, including establishing a foundational GenAI department or investing in AI-specialized start-ups. This multitude of lofty goals is undoubtedly influenced by a sense of urgency exerted by owners, investors, employees, and the public. 

In Germany, a notable characteristic emerges: responsibility for AI transformation within companies lies significantly more frequently with the CEO than with the CIO or other board members responsible for digital matters, or even the head of IT. Specifically, 54 percent of companies in Germany place responsibility for AI with the CEO, compared to only 30 percent in other European countries. A screening of the board structures of DAX companies (including MDAX and SDAX) by Horváth also reveals that there are generally very few board functions with primary responsibility for digitalization. 

What does this mean? It implies, among other things, that decision-makers without experience in digital transformations tend to underestimate the challenges and risks associated with AI—as our study also shows. While only 24 percent of CxOs consider the quality of their own data a significant challenge, the figure at specialist level is 34 percent. A similar pattern emerges regarding risks related to the use of generative AI: Less than half of board members see data protection problems (44 percent)—compared to 64 percent at specialist level. 

Also interesting in this context: At C-level, 85 percent (!) across all functions state that they have at least "advanced" knowledge of generative AI. Simultaneously, boards identify “lack of AI knowledge in overall management” as their most significant challenge. At the same time, board members see a "lack of AI knowledge in overall management" as the biggest challenge. The following recommendations should be allowed in view of these results: 
 

1. Don’t put B before A. 
With an executive team that cannot yet assess what they are endorsing and implementing, the AI transformation is at risk. Results and decisions based on potential analyses must be understood and shared by all stakeholders. 
 

2. Prioritization. 
This applies to all types of transformations. If there are too many loose ends that need to be resolved across organizations, this has a paralyzing and demotivating effect on an organization. It is better to work on the basics and a maximum of three use cases properly than to implement a twenty-point plan with half-baked results. 
 

3. Beware of overconfidence and actionism. 
AI not only presents opportunities but also risks that should not be underestimated. Once the systems have been implemented and produce results of unsatisfactory quality—or even worse, discriminatory decision templates—repairing the damage becomes challenging.