BCG says firms are being pushed to increase AI token usage, but leaders should measure impact, not just consumption.
AI is no longer just an experiment inside companies, and BCG says the next pressure point is how employees “consume” AI—starting with token usage.
Sylvain Duranton. global leader of BCG X. Boston Consulting Group’s tech build-and-design unit. warned that firms need to be ready for what comes next in AI deployment.. His message: companies should begin actively increasing their usage of AI “tokens” early on. rather than tightly limiting spending from day one.
In this context. “tokens” are the basic units that large language models process to generate and interpret text. and they function as a practical proxy for how much computing effort AI systems consume.. As AI agents and copilots spread. the internal conversation has shifted from whether to use AI to how intensively teams use it.
That shift is creating friction at the top.. Duranton described a strain between CFOs focused on costs and CIOs focused on execution. where technology leaders argue that teams need real usage to deliver measurable results.. He also pointed to a workplace dynamic where some engineers aggressively adopt AI tools. while others hold back—and suggested that the latter could end up falling behind.
An important detail in Misryoum reporting from this discussion: measuring token consumption is necessary, but it’s not sufficient. If a company only tracks raw usage—such as how many tokens someone burns per day—it can miss whether AI is actually improving performance.
Duranton emphasized that the goal should be efficiency with outcomes.. Tokens used to support low-value activity, like generating non-business content, may not move the needle.. By contrast, using AI in ways that raise productivity or improve sales conversion in customer-facing work can matter far more.
Misryoum sees the underlying trend clearly: this is shaping how companies will manage AI internally. similar to how they have historically managed scarce resources like hiring and capital allocation.. Tokens. he argued. are likely to become the next resource that leaders compete over—but the competition will increasingly favor teams that can link usage to business impact.
For companies trying to find their footing. the practical starting point is straightforward: begin measuring token consumption if they aren’t already doing so. then refine policies to encourage productive use rather than sheer volume.. The strategic question isn’t whether employees should use AI. but how to ensure that usage translates into results the business can justify.