Key facts
- BNP Paribas CIB's Chief AI Officer Charles Holive considers 'tokenmaxxing' a 'vanity metric'.
- Holive measures AI success by revenue and productivity gains, not token consumption.
- Some tech firms like Amazon and Uber are questioning the ROI of AI spending.
- Holive's team sets KPIs based on expected revenue or productivity and monitors progress.
- Token consumption is still tracked at BNP Paribas CIB to control costs.
Charles Holive, the Chief AI Officer at BNP Paribas CIB, has stated that the concept of 'tokenmaxxing'—maximizing AI usage and token consumption—is a 'vanity metric.' Instead, Holive emphasizes measuring the return on investment (ROI) of AI initiatives by focusing on tangible outcomes such as increased revenue and productivity. His comments come at a time when several technology firms, including Amazon and Uber, are reportedly scrutinizing their AI spending and its effectiveness. Holive explained that his team at BNP Paribas CIB sets specific key performance indicators (KPIs) for AI projects based on anticipated revenue or productivity gains and monitors progress monthly or quarterly against these goals. While token consumption is still tracked to manage costs, it is not the primary measure of success. This approach is shared by other executives, such as Amit Kapur of Tata Consultancy Services, who advocates for focusing on business outcomes, and Antoine Pichot of La Banque Postale, who evaluates AI projects based on efficiency, customer satisfaction, employee satisfaction, and financial impact. Holive acknowledged that token usage can indicate adoption, particularly in software engineering, but stressed that his bank's strategy differs from those heavily focused on usage metrics.