Poor data quality is the biggest obstacle for generative AI

Generative artificial intelligence (AI) offers companies the opportunity to speed up their work processes, improve their efficiency and increase their competitiveness. However, companies often fail to integrate AI applications into their business processes in a sustainable and profitable way. German companies in particular have some catching up to do in this respect. This is the conclusion of the global study “CDO Insights 2025” by Informatica, a provider of data integration software. According to the study, around a third of German companies have integrated artificial intelligence into their business practices so far. In an international comparison, this puts Germany well behind the UK (63%), the Asia-Pacific region (51%) and the USA (48%).

SAP Business Data Cloud creates the prerequisites for your AI scenarios

Increasing investment in AI applications

German companies are also reluctant to invest in generative AI on a global scale. While 72% of companies in Germany that have already introduced or plan to introduce AI applications expect to increase their investments in artificial intelligence in 2025, the figures are higher for the USA (90%), the Asia-Pacific region (86%) and France (80%). Training employees in the use of generative AI is high on the agenda of companies in Germany.

The step from testing AI applications to putting them into practice is proving difficult. Expressed in figures: For almost three quarters of those surveyed, less than half of the AI pilot projects could be transferred to the productive phase. The most common reason for failure is the lack of reliability of the available database (94%). In addition, many data challenges are related to privacy and data protection (50%), the ethical use of AI (50%) and data quality (41%). A lack of technical maturity and interoperability (48%) and deficits in the quality, completeness and readiness of data (42%) also play a role in unsuccessful AI projects.

Data management as the key to high-quality data

In the search for ways in which companies can provide high data quality as the basis for AI applications, the focus is increasingly shifting to data management. Increasingly complex system landscapes pose a considerable challenge for the establishment of functioning data management. As many as 66% of data managers in German companies expect to invest more in data management in 2025 than in the previous year. This is a significantly lower proportion than in the USA (93%) and the Asia-Pacific region (86%). In addition, companies see a need for investment in areas such as data literacy (56%), improving data protection and security (37%) and meeting legal requirements (30%).

Expectations exceed business added value

96% of respondents consider it a challenge to prove the concrete business added value of their AI initiatives. Reasons include limitations in terms of cyber security and data protection (56%), lack of confidence in data quality (48%), lack of key figures in the pilot project (37%) and the reliability of results (35%). 82% of survey participants complain about unrealistic expectations in their own company when it comes to the return on investment (ROI) of projects with generative AI.

Conclusion: AI applications need the right database

The potential of generative AI to improve collaboration, increase efficiency or optimize the user experience is enormous. However, certain requirements must be met in order for this potential to be exploited. These include, in particular, functioning data management, which forms the basis for ensuring that the required data is harmonized and available in high quality.

Only with high-quality data can AI applications be meaningfully integrated into business processes and offer companies competitive advantages. In this context, recently launched SAP Business Data Cloud represents a promising solution. The data platform creates a uniform semantic layer for data from SAP systems and third-party providers and prepares the data for AI use so that companies can gain better insights from their data pools.