Tokenmaxxing
Companies like Meta set up leaderboards to push for rapid AI adoption (before eventually figuring out who to lay off!). In some places, we even saw
Token budgets
However, now we are seeing the second phase of the AI adoption curve in tech. Token spend budgets have become commonplace. Uber ran through its 2026 token budget within 4 months.
CFOs continue to predict AI coding tools as a line item; they can forecast like a SaaS seat license, but we all know that is not feasible anymore. A 10x engineer can end up spending vast amounts of tokens more than the enxt best engineer, while a sloppy new grad coulds also waste millions of tokens along a meaningless rabbit hole.
Tokens would become more commonly used over time. Output usage tokens would also vary across different AI modalities (text, voice, image, video), making ROI measurement even more challenging.
CFOs need to build meaningful metrics in collaboration with product and engineering around determining how to measure the ROI on token spend. Tools like Claude efficiency analyzer can help you analyze and understand how token budgets are spent.
Reading material