Context window
The model's working memory: everything it can see at once, including your prompt, the documents you attach and the conversation so far.
The context window is the model’s working memory: the amount of text it can hold in view at one time, which includes your prompt, any documents you attach, and everything already said in the conversation. Modern models hold hundreds of pages, which is why a KAM can now drop in the annual report, the last three QBR decks and the call transcripts and ask a question across all of them.
Two consequences follow. Anything outside the window does not exist to the model, so a long conversation quietly loses its beginning and a question about last month’s meeting needs last month’s notes in front of it. And what is in the window is what the model reasons from, which is why grounding it in your own sources beats asking it what it happens to remember. RAG and the living account plan are two different answers to the same problem: what to put in the window, and how to keep it current.
Set out in full in: Brooks, AI Glossary of Terms (Cranfield KAM Best Practice, 2026).
Related terms
Prompt
The instruction you give an AI model. The single biggest factor in the quality of what comes back. Also, in the book, a structured question you ask yourself.
Retrieval-augmented generation
The model reads your files afresh to answer each question. It works, but nothing accumulates. Each answer starts from a blank page.
Hallucination
A confident, well-formed answer that is simply false. Worst on multi-step reasoning and specialised domains, which describes key account work exactly.
See the method run on one real account in three days: the One Page Proposition.