If you've ever looked at AI pricing, you've seen the word "tokens." It's the core unit of AI consumption, and understanding it helps you predict costs.
What is a token? A token is a chunk of text — roughly three-quarters of a word. "The quick brown fox" is about 4 tokens. A typical business email is 200-400 tokens. A 10-page report is roughly 3,000-4,000 tokens. You're charged for both input (your prompt) and output (the response).
Why this matters for cost. Most business subscriptions include a generous monthly allowance. But if you're building AI into automated workflows — processing every support ticket or generating nightly reports — token costs become a real budget line. Processing 1,000 support tickets with a mid-tier model costs roughly $2-10.
The efficiency angle. Shorter, well-crafted prompts use fewer tokens and produce better results. A verbose, poorly structured prompt costs more and returns worse output than a concise one. Good prompting is literally cost-efficient.
The trend. The cost of getting a given level of AI performance has been falling roughly tenfold a year, according to a16z and Epoch AI. What costs $100 today may cost $10 a year from now. This deflationary trend is what makes increasingly ambitious AI applications economically viable for SMBs.