A Wall Street Journal survey finds only 26% of firms claim a full view of AI spending. Finance chiefs confront a shifting cost model as vendors move to token‑based pricing, charging per unit of compute rather than flat fees. The change threatens surprise bills for enterprises that have already launched large‑scale language‑model projects.

Traditional budgeting relies on predictable license fees, but token consumption can spike with higher query volumes or more sophisticated models. CFOs now must build real‑time monitoring tools and negotiate usage caps with providers. Without granular tracking, firms risk overruns that could erode margins, especially as AI becomes integral to product pipelines and customer service.

Investors are watching the accounting scramble closely, since AI spend now appears on balance sheets as variable operating expense rather than capitalised software. Companies that publish transparent token usage can signal cost discipline, potentially supporting stock valuations. Conversely, firms blindsided by sudden charges may face pressure from shareholders demanding tighter governance. The emerging pricing regime forces finance teams to treat AI like cloud infrastructure—trackable, billable, and subject to rigorous oversight throughout the year.