AI models don't see text as words or letters – they see it as tokens. A token is roughly 3–4 characters in English. "automation" might be 3 tokens. "AI" is 1 token. "What is artificial intelligence?" is about 8 tokens.
Why should you care? Because most AI model APIs bill by token count – separately for input tokens (prompt + context) and output tokens (generated response).
Indicative prices (OpenAI GPT-4o, May 2026): input ~$2.50 / 1M tokens, output ~$10 / 1M tokens.
For typical business use, this means: processing 1,000 customer emails per month with a reasonably configured system will cost a few dollars, not hundreds.
Token limits also relate to the context window – window size is measured in tokens. When we build AI solutions for clients, we always calculate projected token consumption so you have realistic operational cost estimates.