Tokenomics: Why making AI pay is tricky

What the report says

BBC World News reports that companies buying and selling AI services are finding it difficult to price the technology in a way that is predictable for budgets and sustainable for providers. The issue is especially acute as more firms build products around AI agents, which can trigger large and variable usage of “tokens” — the processing units used by large language models such as ChatGPT, Claude and Gemini.

The article says the economics are changing quickly: token costs have fallen, but total consumption is rising as businesses use AI for coding, automation and other tasks. That makes it hard for customers to forecast bills and for vendors to decide whether to charge by usage, by results or with flat fees. People quoted in the report, including executives and academics, say experimentation inside companies can lead to unexpectedly high costs, particularly when AI is rolled out across many users or used in multi-agent systems.

The BBC notes that some organizations are trying to limit exposure by using simpler pricing plans, being more specific in prompts, or choosing models more carefully. It also says some large tech companies have already tightened access to third-party AI tools in certain cases, highlighting the pressure to control spending.

The broader significance is that as AI moves from novelty to enterprise infrastructure, pricing may become a major obstacle for adoption. The BBC says both providers and customers are still searching for a model that reflects the value of AI while remaining workable in a fast-changing market.