> For the complete documentation index, see [llms.txt](https://help.filed.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://help.filed.com/reference/products/chat-and-analyst/prompting-best-practices.md). # Prompting best practices ## Start broad, then go deep Before opening individual documents, ask Chat for an overview: "What's this client's tax position?" or "What's in this binder?" surfaces a summary grounded in the client's tax position data - taxpayer profile, income sources, and estimated position, as described in [What the assistant knows about this client](/reference/products/chat-and-analyst/client-chat.md#what-the-assistant-knows-about-this-client). Starting here builds context before diving into individual source documents, and often surfaces which documents are worth opening first. ## Be specific about what you want back A vague question gets a general answer. Naming the form, the box, the client, or the year narrows the assistant's search and produces a more precise citation. Compare "did their income change?" with "how did Schedule C income change from last year?" - both are valid questions, but the second points the assistant at a specific comparison instead of a general summary. ## Use dictation for hands-busy review The message box supports voice input - useful when a document is open and typing would mean switching focus away from it. Dictate the question, review the transcribed text before sending, and the assistant answers the same way it would a typed message. ## Ask for output in the format you need Chat and Analyst can produce more than a written answer. Asking for a table, a CSV, or a spreadsheet - for example, "give me a CSV of every client missing a K-1" - returns the answer as a downloadable file instead of only prose. This is useful for anything you plan to hand off, file, or process outside of Filed rather than just read on screen. ## Verify before you rely on an answer Every claim grounded in a document comes with a citation. Click it before treating the answer as final - especially for anything that will inform a filing decision. See [Cited answers and navigating to sources](/reference/products/binder/chat.md#cited-answers-and-navigating-to-sources) in Binder Chat and [How the assistant cites its answers](/reference/products/chat-and-analyst/client-chat.md#how-the-assistant-cites-its-answers) in per-client Chat. ## Let the assistant take the next step, but review the proposal Both per-client Chat and Binder Chat can propose Binder actions - adding a note, placing a flag, signing off an item - directly from the conversation. These show up as a proposal to approve, not an action already taken. Read the proposal the same way you would review a flag: confirm it matches your intent before approving it. ## Use separate conversations for separate lines of inquiry Starting a new conversation (the **+** control) keeps an unrelated question from cluttering the thread you are actively using to work through a specific issue. Conversation history is saved per client, so nothing is lost by starting fresh - see [Conversation history](/reference/products/chat-and-analyst/client-chat.md#conversation-history). ## Choose the right surface first Before prompting, confirm you are in the surface that can actually answer the question - see [Choosing the right surface](/reference/products/chat-and-analyst.md#choosing-the-right-surface). A cross-client question asked in per-client Chat will come back empty; the same question in the Analyst works because it has workspace-wide scope. --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://help.filed.com/reference/products/chat-and-analyst/prompting-best-practices.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.