AI is intended to increase efficiency – also in finance. The potential is considerable: data-intensive processes such as incoming invoices, account reconciliations and reporting can be significantly accelerated with appropriate tools. However, a current IDC survey commissioned by the software provider Sage shows that the productivity gains achieved are often lost in practice – because employees have to check the generated results manually.
AI still too often causes additional work
As part of the survey, more than one in four people stated that they would have to directly reinvest the working time saved through AI tools in order to make AI decisions understandable and explained to stakeholders. 29 percent of financial decision-makers in Germany spend between 15 and 29 hours checking AI results. For 18 percent it is even over 30 hours per week. The results come from a study for which a total of 2,275 senior financial decision-makers from North America, Europe, the Middle East and Africa were surveyed, including 205 from Germany.
The traceability of AI decisions plays a central role. The study shows that 68 percent of German respondents would reject an AI tool that promises 99 percent precision but cannot transparently justify its decisions. “In finance, ‘almost right’ has always meant ‘wrong.’ As AI workflows become more complex, the cost of uncertainty continues to rise,” said Aaron Harris, CTO at Sage. “Finance teams cannot afford to act as detectives navigating opaque AI outputs.”
Top Article
${content}
${custom_anzeige-badge}
${custom_tr-badge}
${section}
${title}
Verification reduces AI benefits
Similar problems can also be seen in other industries. Waydev CEO Alex Circei reported to Techcrunch that although AI can produce more code, it needs to be revised more frequently. The initial acceptance rate was 80 to 90 percent, but subsequent corrections reduced it to ten to 30 percent. Faros AI comes to a similar conclusion: The analysis of customer data from two years showed that with intensive AI use, code fluctuation increased by 861 percent.
At this year’s Fortune Brainstorm Tech conference, executives from various companies reported similar experiences, particularly with regard to the traceability of AI agents. “A key question that concerns us is how to design a system that works correctly as often as possible,” said Edwin Olson, founder and CEO of May Mobility. Since errors are inevitable, transparency plays a crucial role: one must understand why an error occurs in order to avoid it in the future.
Human experience weighs more
The IDC survey shows an interesting pattern for Germany: While around the world the main reason given for rejecting AI recommendations was a lack of transparency, in Germany the focus was on the conflict with one’s own expert opinion. 43 percent of German finance managers admitted that they would immediately reject an AI recommendation if it contradicted their own professional judgment. On a global average, 39 percent of the decision-makers surveyed said this. In Germany, AI is therefore seen as a tool that must be measured against human expertise – not the other way around.
