Study: AI investments show returns, but data problems slow progress

Companies worldwide are investing heavily in artificial intelligence and achieving measurable returns while simultaneously struggling with fundamental data management problems. This is according to a study by SAP and Oxford Economics, which surveyed a total of 1600 executives in eight countries, including Germany, the USA, China, and Brazil, between July and August 2025.

The companies surveyed are spending an average of 26 million US dollars on AI in the current fiscal year and expect a return of 16 percent, which corresponds to approximately 4.7 million dollars. In two years, the return is expected to increase to 31 percent – equivalent to 12.3 million dollars. In fact, 79 percent report having achieved positive returns on investment within one to three years. According to participants, AI already supports 25 percent of business tasks; this share is expected to rise to 41 percent by 2027.

Only 9 percent rely on a strategic AI approach

Despite these optimistic figures, the study reveals significant weaknesses: According to it, only 9 percent of companies pursue a strategic, holistic approach to AI investments. 44 percent describe their approach as “piecemeal,” 32 percent see it driven by individual departments, and 15 percent call it “ad hoc.” Almost two-thirds of respondents (65 percent) are unsure if AI is already reaching its full potential.

Data maturity emerges as a central challenge. While 71 percent consider data critical for AI investments, 55 percent doubt their ability to responsibly share data across business units. 60 percent have concerns about integration with external partners. The biggest hurdles are incomplete or inconsistent data (75 percent), lack of data quality (69 percent), and data silos (68 percent).

Shadow AI as a Security Risk

Another problem is the uncontrolled use of AI tools: 64 percent of companies report that employees use unapproved shadow AI applications at least occasionally. 77 percent express concerns about the associated risks. More than half report inaccurate results, and many also report data leakage or security vulnerabilities. These findings are consistent with a current study by the industry association Bitkom.

Regarding agent-based AI – meaning AI systems that make decisions and perform tasks independently – 78 percent of respondents are convinced of its transformative potential. However, only 5 percent consider themselves fully prepared for the use of such agents, while 54 percent rate themselves as partially prepared. For investments in agent-based AI, companies expect a return of 10 percent in the next two years, corresponding to 4.3 million dollars.

The evaluated results of the study can be found at SAP for those interested.

(fo)