Investment banks have been cautious about publishing AI ROI figures. Deutsche Bank’s Investment Bank just changed that with specific numbers that the rest of the industry will now have to respond to.
Denis Roux, the bank’s Chief Information Officer for the Investment Bank, disclosed that AI has reduced certain project completion timelines from two years to three months, an 87.5 percent reduction in delivery time. The bank uses simpler, more predictable models for routine tasks rather than deploying frontier AI uniformly, and it manages AI cost discipline by allocating tokens to engineers who must demonstrate measurable value to receive additional allocations.
The use cases are concentrated in financial data extraction and analysis, and in linking external market events to portfolio exposure assessments, both high-frequency, high-stakes tasks where speed improvement translates directly into competitive advantage. The bank remains cautious: “We don’t want to slow people down and want them to keep going,” Roux said, “but we also want to get a return.”
The broader industry data supports the direction if not yet the magnitude. A recent survey cited in the Reuters report found that 85 percent of financial services firms with revenue above $1 billion plan to increase AI budgets over the next 12 months, with 65 percent citing productivity and efficiency gains as the primary justification.
The original insight: Deutsche Bank’s disclosure matters not because of the scale of the results but because of the specificity. Two years to three months is a number that finance CIOs can present to boards and use in their own business cases. As quantified AI ROI examples accumulate in financial services, the organizations still in evaluation mode will face increasing internal pressure to show equivalent results or explain the gap. This is the same pressure driving agentic AI adoption in treasury operations into production from proof-of-concept.
Source: PYMNTS
