Monthly subscription, open chat, ask question: This is how generative AI used to work. Agentic workflows go beyond this model. They consume many times more tokens, run autonomously for hours and make flat rates unaffordable for providers. At the same time, token prices vary according to speed, specialization and the economic value of the result. The third edition of the AI radar categorizes the emerging token economy: How is billing shifting from subscription to consumption? Why does a favorable token price reveal little about the actual costs? And why is token consumption alone the wrong measure of AI value creation?

"The whole conversation shifted from tokenmaxxing and 'go fast' to 'we need guardrails, how do we control this?'"

AI token pricing is replacing flat-rate subscriptions, and IPO season will push bills higher. Here is what a Sri Lankan small team should change now.