AI and agents
Tokens
A token is the unit of text a language model reads and writes, usually a word or part of a word, and the unit most AI providers bill by.
How it works in practice
Models do not see letters or words directly. Text is split into tokens by a tokeniser, and a common rule of thumb for English is that one token is about three quarters of a word. Providers charge per million input tokens and per million output tokens, with output usually priced higher.
Token counts drive the running cost of any AI build. A long system prompt sent on every call, full email threads where a summary would do, or retrieved documents that are not needed all add up. We estimate token use per task during the blueprint so the monthly AI bill is known before the build starts.
A worked example
A firm classifies 2,000 support emails a day. Trimming each email to the latest message and dropping signatures cut the average input from about 1,800 tokens to about 400, which cut the classification cost by roughly three quarters.
Examples in this glossary describe typical builds for companies of 1 to 1,000 people. They are not named client results.















