SAP CFO says AI must move beyond chatbot 'low-hanging fruit' before seeing returns

Dominik Asam, the Chief Financial Officer of SAP, believes that the true value of artificial intelligence in enterprise software lies beyond the current focus on chatbots and coding assistants. He asserts that companies need to move towards integrating AI into more complex business processes where data quality, reliability, and cost control are paramount.

“

The lion's share of AI token consumption today was spent in 'low-hanging fruits' coding assistant and chatbots, where AI's hallucinations matter less because the output carries limited risk if it fails.

Asam noted that while companies have invested heavily in generative AI, widespread productivity gains are still elusive. He suggests that the most substantial returns will not come from general-purpose AI models but from governed systems embedded within specific business operations. The CFO pointed out that a significant portion of current AI spending is directed towards what he termed "low-hanging fruit" – applications like coding assistants and chatbots.

In these applications, the risk associated with AI hallucinations is relatively low because the output has limited consequences if it fails. However, applying AI to critical business areas such as finance, supply chain management, or other core processes presents greater challenges. Errors in these domains can compound across multiple steps, increasing the risk of non-compliance with established standards.

“

If you have some hallucinations in the process, the errors will actually compound statistically over many steps. It requires much more excruciating assurance levels.

Asam elaborated that the "high-hanging fruit" of AI involves building systems tailored to specific business needs, rather than relying on a universal, plug-and-play large language model. This approach necessitates that companies ensure their own data is usable and well-governed, enabling AI to operate with accurate, company-specific knowledge. He cautioned against the notion that AI can solve all problems when faced with messy, legacy data silos, as such an approach can incur extremely high costs.

“

The idea that AI will solve all these problems if they are messy, legacy data silos is not true.