What is a token?
A token is a small chunk of text, often part of a word, that an AI model reads or writes and that providers use to measure usage and price. OpenAI's rule of thumb for English is that one token is about four characters, and 100 tokens are about 75 words. Both the text you send and the text the model writes back count.
Why it matters to a revenue leader
Tokens turn AI from a fixed software line into a variable cost that grows with use. Many model providers price per million tokens, with separate rates for input and output; on Anthropic's published price list, output tokens cost more than input tokens for every model shown. Long documents, full CRM histories and agents that loop through many steps all burn tokens, so in a PE-backed company watching EBITDA, AI spend needs an owner and a forecast, the same as fuel or overtime. Tokens also set limits: each model can only hold so many at once, which caps how much account history it can consider in one pass.
Example
Fictitious numbersA field services company has an AI tool summarize every customer email thread before renewal season. Each summary reads about 6,000 tokens and writes 500. Across 2,000 agreements that is 12 million input tokens and 1 million output tokens, so finance can price the job before it runs.
"What do our AI tools spend in tokens each month, and who signs off when that number doubles?"
Sources
- Understanding and counting tokens checked 9 Oct 2026
- Pricing - Claude Platform Docs checked 9 Oct 2026