AI Must Tackle Complex Processes for Real ROI, Says SAP CFO

SAP's Chief Financial Officer, Dominik Asam, said on Thursday that for generative AI to deliver substantial productivity gains, it must evolve beyond simple applications and be embedded into core business processes. Speaking after the company's second-quarter results, he argued that the real value lies in applying AI to complex areas like finance and supply chain, where clean data and reliability are more critical than access to the most powerful models.

The Challenge of 'High-Hanging Fruit'

Asam contrasted the current popular uses of AI with its more demanding, high-value applications. He told reporters that the "lion’s share" of AI usage today is focused on "low-hanging fruits" like coding assistants and chatbots, where potential errors or "hallucinations" carry limited risk.

However, applying AI to core financial or supply chain workflows is significantly harder. Asam warned that in these multi-step processes, errors can compound, increasing risk and threatening compliance standards. "If you have some hallucinations in the process, the errors will actually compound statistically over many steps," he said, adding that this requires "much more excruciating assurance levels."

Data Governance Over Model Power

According to Asam, the idea that a powerful, generic large language model can fix disorganized internal data is a misconception. "The idea that AI will solve all these problems if they are messy, legacy data silos is not true," he stated, noting that this approach comes with "extremely high token costs."

Ad

Instead, he emphasized that companies must first make their own data usable and governed, allowing AI to operate with precise company knowledge. Asam also noted that the most advanced model isn't always the right tool. He said companies will ultimately use the cheapest, most reliable option that can deliver the required outcome safely, whether that is simple software, an open-source model, or a more expensive frontier AI.

Market Implications

Asam's comments reflect a growing sentiment in the corporate world, where massive investment in AI has yet to translate into broad, verifiable productivity gains. His perspective signals a strategic shift from general-purpose AI hype toward practical, governed systems embedded within specific business functions.

For investors, this highlights the long-term value of companies that can provide the data infrastructure and process-specific AI tools necessary for this more mature phase of adoption. It suggests that the future of enterprise AI will be less about the model itself and more about its reliable and cost-effective integration into critical operations.