> For the complete documentation index, see [llms.txt](https://docs.filed.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.filed.com/reference/products/chat-and-analyst/analyst.md). # Analyst ## What is the Analyst? The Analyst is a firm-wide conversational assistant accessible from the left navigation. It is scoped to the entire workspace and can answer questions that span all clients: who is missing documents, which returns are ready to file, who is at risk of an underpayment penalty, and similar pipeline questions. It also supports general tax research, fetching current IRS figures and rule details using live web search. The Analyst maintains a persistent, searchable thread history. ## Opening the Analyst The Analyst is in the left navigation. The landing view shows a hero area, a message composer, and a set of starter prompt chips sampled from the full prompt pool. You can also reach the Analyst from the Home page: typing a question in the "Ask a question" box and submitting it opens a new Analyst thread automatically and navigates you there. ## What data the Analyst can reach The Analyst queries a synthesized index called the Dataview, which covers every client-year pair in the workspace that has been processed. For each client and tax year, the index includes: * Client identity and return type. * Filing attributes (filing status, entity type, tax year). * Income-source flags (for example, whether the client has W-2 income, self-employment income, rental income, or K-1s). * Deduction and credit flags. * Business-related flags. * Key dollar amounts from the return. * A synthesized text narrative describing the client's overall tax position. The index does not include raw source documents. For document-level questions such as "What does their W-2 say?", use per-client Chat instead. ## Firm-operations questions The Analyst is well suited to questions about your filing pipeline and client book. Sample questions: * "Which clients are still missing documents?" * "Who's going to owe this year?" * "Which returns are ready to file?" * "Who should I put on extension?" * "Who's at risk of an underpayment penalty?" * "Which clients are waiting on K-1s?" * "Who had a big income jump this year?" These queries run against the Dataview index. Results are at the client level, not the document level. A client appears in results only after their binder has been created. ## Tax research questions The Analyst can answer general tax questions by fetching current information from the web. This means thresholds, rates, phase-outs, and deadlines are sourced from live data rather than training-data recall. Sample questions: * "What's the bonus depreciation rate this year?" * "Who qualifies for the QBI deduction?" * "What are the estimated-tax safe harbors?" * "What are the crypto reporting rules now?" * "How do the passive activity loss rules work?" * "What counts as reasonable S-corp compensation?" For any answer based on an external source, the Analyst will cite where the information came from. ## Saved threads and conversation history The Analyst maintains a persistent list of saved threads. Threads are listed in a panel and are searchable, so you can return to a prior conversation or pick up where you left off. A specific thread can be linked to directly using the `?thread=` URL parameter in the Analyst URL, which is useful for bookmarking a running conversation. ## Interpreting Analyst answers A few things to keep in mind when reading Analyst responses: * Aggregate counts (for example, "23 clients are missing documents") are based on the Dataview index. A client appears in results only after their binder has been created. * Dollar figures come from return data at the time of last processing. * For tax-law answers sourced from the web, look for source attribution in the response. * If a result seems off, ask the Analyst to show its reasoning or the underlying data it used. * The Analyst counts by distinct client, not by tax year, to avoid double-counting clients who have multiple years in the workspace. ## Analyzing across related entities The Analyst can answer questions that span several clients in the same owner group. You do not need to build a consolidated "client group" client or combine documents into one record; open the Analyst without selecting a client and ask across the group. Questions of this shape work well: * "Show me all entities owned by the Smith family group." * "Compare K-1 income across these three LLCs." * "Which entities had a loss this year?" Each entity must have been through Tax Prep or binder ingestion at least once, so that Filed has extracted data to work from. Answer quality depends on how complete that extraction is for each client. Two limits are worth knowing up front. Filed has no formal consolidated entity structure, and a single Tax Prep or AI Review run covers one client only. AI Tax Planning also produces per-client results, so multi-entity work such as aggregating QBI or coordinating losses across entities has to be done through the Analyst rather than inside a plan. ## Limitations and scope boundaries * The Analyst does not have access to raw source documents. For document-level detail, use per-client Chat. * The Dataview index is populated by binder creation. Clients whose binders have not yet been processed will not appear in Analyst results. * The Analyst cannot navigate to an individual client's pages or change the product view to a specific client. * The Analyst cannot file or transmit anything to a tax authority. ## Skills and firm protocols The Analyst's behavior can be shaped by skills: saved instruction sets that load when relevant to the type of question being asked. Firm admins can author and publish custom skills to define how the Analyst responds to specific categories of questions, for example a firm-specific checklist for extension decisions or a preferred format for client summaries. --- # 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://docs.filed.com/reference/products/chat-and-analyst/analyst.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.