Enterprise AI bills are rapidly expanding due to increasingly prevalent agentic workflows, which cause token consumption to grow by multiples or even dozens of times through repeated context reading and tool invocation. Simple Q&A tasks require hundreds to thousands of tokens, but integrating Excel, CRM, and web search for reports can accumulate hundreds of thousands or millions of tokens.
Agentic AI adoption is driving disproportionate token consumption, increasing enterprise inference costs and creating a budget control crisis for businesses.