s companies rush to adopt agentic AI, they are learning that the cost dynamics are both complex and, at times, unexpectedly expensive. While the cost of individual tokens continues to fall, enterprise spending on AI is spiking. At the same time, the pricing structures continue to evolve and shift, injecting a high degree of uncertainty into “tokenomics”—or AI economics, the discipline of managing agentic AI spend at scale. As of early July 2026, using a frontier model such as OpenAI’s GPT-5.5 costs about $5 per one million input tokens and $30 per one million output tokens, compared with just $0.20 and $1.25, respectively, for an earlier lightweight model. Clearly, these models offer different capability levels, but not all tasks require the most expensive model option. To date, however, many leaders don’t operate based on that understanding, leading to significant cost implications.

Full commentary : Early experience implementing agentic workflows highlights emerging trade-offs that frontline leaders need to understand.