SAP finance chief says AI must move past chatbots to pay off
TL;DR:
- SAP’s finance chief says enterprise AI must move beyond chatbots and coding tools into complex business processes before it delivers real returns.
- CFO Dominik Asam argues the “lion’s share” of AI spending today goes on “low-hanging fruit” where hallucinations carry little risk.
- The harder, more valuable work depends on clean, governed company data — and the cheapest reliable model, not the most powerful.
Asam’s central point is about compounding error. In coding or chat, a wrong answer is cheap; in finance or supply-chain workflows, mistakes “compound statistically over many steps” and demand “much more excruciating assurance levels”. That is why, he argues, returns will come from AI embedded in specific, governed processes rather than a generic model sprayed across a company. It is a notably grounded message from a vendor whose own results depend on selling AI, and it arrives as investors sour on open-ended AI spending across the sector.
The value is in the data, not the model
His sharpest line punctures a common assumption: “The idea that AI will solve all these problems if they are messy, legacy data silos is not true.” Firms must make their own data usable first — and then choose the cheapest reliable tool, whether simple software, an open-source model or an expensive frontier one. The framing is a useful corrective to the capability arms race, and it lands alongside Barclays telling clients AI is not yet lifting productivity and the ONS finding UK firms adopting AI widely but shallowly.
Looking forward
For UK businesses budgeting AI, the message reframes the spending question. The bottleneck is rarely model access; it is the unglamorous work of data governance that makes core-process automation safe. Firms treating AI as a plug-in risk staying stuck on the low-value tier — while UK finance chiefs report rising optimism that will only pay out if the data groundwork is done first.