Every conversation in Ask Double draws from your firm's shared pool of AI credits. The habits below help that pool go further. None of them change the quality of your results — they just avoid paying for context Ask Double doesn't need.
For how your firm's credit pool is calculated and how to set an overage budget, see AI Credits: Usage, Budgets, and Adding Credits.
Keep each chat focused on one task
Ask Double sends the full conversation history with every new request — that's what lets it follow along across a multi-step workflow. The tradeoff is that earlier messages and any uploaded files are re-sent each time you send a message, so a long chat costs more per request than a short one.
Best practice: Stay in the same chat while you're working on the same task or workflow. Start a new chat when you switch topics, and always start a new chat once you've finished working with an uploaded file.
For example, stay put while reviewing a set of invoices; start fresh before moving on to a bank reconciliation. Uploaded files matter most here — they stay in the conversation's context and keep getting re-processed long after you're done with them.
Running a skill automatically starts a new chat.
Upload only the data you need
File processing is the most credit-intensive thing Ask Double does. A large workbook containing data unrelated to your request still has to be processed in full.
Best practice: Trim files before uploading. Upload a single worksheet rather than a multi-sheet workbook, remove columns and rows that aren't relevant, and avoid attaching a large supporting file when only a small part of it matters.
Choose the right model for the task
By default, Ask Double uses the model we consider the best balance of capability and reliability for accounting work.
Best practice: Switch to a lower-cost model for simple or repetitive work, summaries, and straightforward prompts. Keep the default for analysis, diagnostics, cleanup scoping, and anything involving multi-step judgment.