At the Mistral AI summit in Paris, several corporate executives stated that businesses are shifting their focus when measuring AI ROI from token consumption to more direct business outcomes, including improved efficiency, enhanced customer experience, and better cost-to-output performance.
Executives place greater emphasis on efficiency and results.
Charles Holive, Chief AI Officer at BNP Paribas Corporate and Institutional Banking, said the team strives to avoid "vanity metrics." In his view, the number of tokens consumed daily does not directly indicate whether AI is generating real returns.
He is more focused on what tasks employees have accomplished with AI that were previously impossible, and how much faster processes have become. Antoine Pichot, a senior executive at La Banque Postale in France, also noted that internal evaluations primarily assess efficiency, customer service, and whether the return on investment aligns with the inputs.
The sentiment that "the more tokens, the better" is cooling down.
As these statements emerge, some U.S. companies are beginning to reconsider the practice of "tokenmaxxing"—equating higher AI usage directly with greater productivity.
Last month, Amazon shut down an internal AI usage leaderboard after some employees began prioritizing tasks that boosted their rankings. Uber’s Chief Operating Officer, Andrew Macdonald, has also publicly questioned whether higher AI spending truly leads to more useful products.
The token is still being tracked but is no longer centered.
Several executives did not deny the role of token data. Holive stated that the team will still monitor token consumption to control costs and track adoption, but such data should not be the primary metric for measuring returns.
Meanwhile, OpenAI, Anthropic, and GitHub are all promoting pay-as-you-go models for enterprise customers, making it increasingly necessary for businesses to demonstrate whether increased AI usage is truly translating into measurable business benefits.