Key facts
- SAP CFO Dominik Asam believes AI must move beyond chatbots and coding tools to deliver substantial business returns.
- He stated that current AI applications are largely 'low-hanging fruit' with limited risk from errors.
- Applying AI to core business processes like finance and supply chain is more complex due to higher risk and compliance requirements.
- Asam emphasized the importance of clean, governed data for AI to function effectively in specific business functions.
- He advised that companies should use the most cost-effective and reliable AI tool for a given task.
SAP's Chief Financial Officer, Dominik Asam, has stated that the widespread adoption of artificial intelligence in enterprise software needs to progress beyond current applications like chatbots and coding assistants to achieve meaningful business returns. Speaking after SAP's second-quarter results, Asam explained that while companies are investing heavily in generative AI, the evidence of broad productivity gains is still emerging.
Asam characterized the current use of AI, particularly in chatbots and coding tools, as 'low-hanging fruit' where the risk associated with AI 'hallucinations' is minimal due to limited consequences if errors occur. However, he stressed that applying AI to more critical business functions, such as finance or supply chain management, presents greater challenges. In these areas, errors can compound across multiple steps, increasing risks and potentially violating compliance standards.
He further elaborated that the true value of AI, or the 'high-hanging fruit,' lies not in generic, plug-and-play large language models, but in building systems tailored to specific business needs. This approach requires companies to ensure their data is usable and well-governed, enabling AI to operate with company-specific knowledge. Asam cautioned against the notion that AI can solve problems stemming from messy, legacy data silos, noting that such implementations can incur 'extremely high token costs.'
Ultimately, Asam suggested that the most advanced AI model is not always the most appropriate. In practice, he believes companies will opt for the most reliable and cost-effective tool that can safely achieve the desired outcome, whether that involves simple software, an open-source model, or a cutting-edge frontier model.