> ## Documentation Index > Fetch the complete documentation index at: https://help.agentsy.pro/llms.txt > Use this file to discover all available pages before exploring further. # Organize your knowledge > Structure your expertise, methodologies, FAQs, and reference materials so your Agentsy assistant can find and use relevant information. A useful AI Experience starts with useful source material. Organizing your knowledge helps your assistant locate relevant information, explain your approach consistently, and recognize when it does not have enough information to answer. You do not need to create a large library before launching. Start with a small set of accurate, focused documents, test them with realistic questions, and expand as you learn what visitors need. > [!NOTE] > > **Knowledge is different from instructions.** Your knowledge documents explain *what the assistant knows*: your services, methodology, educational content, and policies. [System instructions](/create-and-configure/system-instructions) define *how the assistant behaves*: its role, tone, boundaries, and when to suggest speaking with a person. ## 1. Start with the questions visitors ask Before organizing files, write down the questions your audience is likely to bring to the experience. For example, a consultant might hear: * What problems do you help solve? * How does your methodology work? * How is your approach different? * What should I consider before engaging a consultant? * When would it make sense to speak with you directly? Group these questions into topics. Each topic can become a section or a focused knowledge document. ## 2. Organize documents by knowledge role A practical starting structure is: | Knowledge role | What it contains | Example document | | - | - | - | | About and approach | Who you help, your philosophy, and the scope of your work | `01-about-and-approach.md` | | Core concepts | Definitions and educational foundations | `02-fundamentals.md` | | Options and use cases | Different situations, approaches, or categories | `03-options-and-use-cases.md` | | Your methodology | The steps, principles, and decision framework you use | `04-our-methodology.md` | | Decision guidance | Questions and trade-offs visitors should consider | `05-decision-guide.md` | | Common questions | Frequently asked questions and clear answers | `06-faq.md` | | Boundaries and next steps | What the assistant cannot verify and when a person should help | `07-boundaries-and-handoff.md` | This is a planning framework, not a required Agentsy folder structure. You can adapt the number and names of documents to your subject. ## 3. Keep each document focused A document should have a clear purpose. Avoid combining unrelated topics simply to reduce the number of files. **Less useful:** One long file containing your biography, all services, pricing policies, industry definitions, case studies, and FAQs. **More useful:** Separate documents for your approach, each major service, decision guidance, and common questions. Focused documents make information easier to review and maintain. They may also improve retrieval, although results depend on the content, search configuration, and the visitor's question. ### A simple document template Use a consistent structure when writing or revising a knowledge document: ```markdown theme={null} # Topic title ## Purpose Explain what this document covers and when it is relevant. ## Key concepts Define important terms in plain language. ## Guidance Describe the principles, steps, or options accurately. ## Examples Provide realistic situations that illustrate the guidance. ## Limitations Explain what is unknown, variable, or outside the scope. ## Related topics List other documents that provide useful context. ``` Write complete explanations rather than isolated keywords. Use descriptive headings, short paragraphs, and lists where appropriate. ## 4. Example: a consultant's knowledge library A business consultant might start with: 1. **Advisory philosophy:** how the consultant approaches business challenges. 2. **Services and outcomes:** the problems addressed and types of engagements offered. 3. **Assessment framework:** the questions used to understand a client's situation. 4. **Strategic options:** common approaches and their trade-offs. 5. **Implementation considerations:** what affects feasibility and timelines. 6. **Frequently asked questions:** recurring prospect concerns. 7. **Working together:** when to book a discovery conversation and what happens next. The assistant can use these materials to help visitors understand the consultant's thinking before a human conversation. ## 5. Example: an insurance education experience An insurance advisor offering educational guidance across multiple areas could organize documents into subject groups: | General guidance | Life insurance | Health insurance | Fixed annuities | | - | - | - | - | | Advisor's approach | Fundamentals | Plan types | Fundamentals | | Education principles | Term vs. permanent | Premiums and deductibles | Crediting methods | | Glossary | Beneficiaries | Provider networks | Liquidity and surrender | | Common FAQs | Coverage considerations | Enrollment basics | Time horizon and suitability | Keep product-specific information separate from general education. If the experience cannot provide regulated, personalized advice, state that limitation in the system instructions and relevant content. > [!WARNING] > > Do not upload protected health information, client records, identification documents, or other sensitive personal data unless you have established an appropriate, authorized process for doing so. An educational assistant should not request such information in chat. ## 6. Example: a boat-buying advisory experience A boat broker could create focused guides on: * Boat types and intended uses * New versus used boats * Budget and total cost of ownership * Inspections and due diligence * Ownership and maintenance * Safety and regulations * Buyer scenarios and checklists These documents support educational conversations. They **do not establish live inventory, current prices, or availability**. If the assistant has no verified real-time inventory source, it should explain that limitation and recommend speaking with a broker when a visitor wants actual listings or an active boat search. ## 7. Separate stable knowledge from changing information Some information changes slowly: your methodology, general concepts, and service philosophy. Other information may change frequently: prices, regulations, availability, schedules, or product offerings. For changing information: 1. Identify the person responsible for keeping it current. 2. Include a review date or version in the source document where useful. 3. Remove or replace outdated materials. 4. Avoid presenting historical figures as current facts. 5. Test the assistant again after important updates. > [!NOTE] > > Uploading a document does not make its contents automatically current. The assistant should not claim to have checked live information unless an appropriate live data source is actually connected and used. ## 8. Avoid contradictions and unnecessary duplication If two documents provide conflicting answers, the assistant may struggle to determine which one applies. Before uploading, check for: * Different versions of the same policy or methodology * Outdated service descriptions * Conflicting definitions or eligibility requirements * Duplicated FAQs with inconsistent answers * Unexplained regional or jurisdictional differences Where multiple rules legitimately apply, make the conditions explicit. For example, identify the country, state, customer type, or effective date associated with a rule. ## 9. Test the organization with real questions After adding documents to your Agentsy knowledge base, ask questions that should draw from different sources. | Test question | What to verify | | - | - | | "How does your approach work?" | The answer reflects your methodology accurately. | | "Which option might suit my situation?" | The assistant explains considerations without inventing facts. | | "What does this term mean?" | The definition is clear and consistent. | | "Can you show me what's available today?" | The assistant does not fabricate live availability. | | "I want to work with someone. What's next?" | The assistant gives an appropriate human next step. | | "What if the documents don't cover my question?" | The assistant acknowledges uncertainty rather than guessing. | If an answer is incomplete, inspect the relevant source document first. It may need a clearer heading, more explicit wording, or a dedicated FAQ entry. Then test again. ## Knowledge organization checklist * [ ] Each document has a clear topic and purpose. * [ ] Headings describe the information beneath them. * [ ] Core terms and methods are explained in plain language. * [ ] Frequently asked questions have direct answers. * [ ] Examples are realistic and clearly distinguished from current facts. * [ ] Conflicting and outdated documents have been resolved. * [ ] Sensitive or unnecessary personal information has been excluded. * [ ] Limits on advice, inventory, and other live information are clear. * [ ] The assistant has been tested with realistic visitor questions. ## Next steps Once your knowledge is organized, [upload your documents](/knowledge-management/upload-documents) and test how the assistant retrieves them. For a deeper explanation of retrieval, continue to [File search and retrieval](/knowledge-management/file-search-and-retrieval). You can also review [Knowledge base overview](/knowledge-management/knowledge-base-overview) and [System instructions](/create-and-configure/system-instructions) to understand how content and assistant behavior work together. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.