Why it matters for your business
Context window size decides how much a model can consider at once: a long contract, a full email thread or many retrieved documents. Large windows are useful, but filling them has costs. Every token sent is billed, longer requests take longer, and models can pay less attention to details buried in very long inputs. For most business systems, sending the right few pages beats sending everything. Long-running agents and chat sessions also need a plan for summarizing or trimming history as they approach the limit.
Example
A law firm wants to compare a 120-page vendor agreement against its standard terms. The full agreement fits in the model's context window, so the review runs in one request. For a question across hundreds of past contracts, the same approach would not fit or would cost far more, so the firm uses retrieval to send only the relevant clauses.