> ## 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. # Generate Your Configuration > Turn your AI Experience plan into a practical configuration you can review and implement in Agentsy. # Generate Your Configuration Once you have defined your AI Experience, described your audience, designed the conversation, and prepared your knowledge, the next step is to turn those decisions into a usable configuration. The **AI Experience Builder** helps organize your choices into a coherent starting point for an assistant. The resulting configuration is a **draft to review**, not a guarantee that an AI Experience has already been created or published in your Agentsy account. ## Before you generate Make sure you can explain: * **Purpose:** What should visitors be able to understand or accomplish? * **Audience:** Who is the experience designed to help? * **Assistant role:** Is it an educator, guide, advisor, or another clearly defined role? * **Conversation approach:** How should the assistant welcome visitors, ask questions, and respond? * **Knowledge:** Which documents, methodologies, and approved facts can support its answers? * **Boundaries:** What should the assistant avoid claiming or doing? * **Human handoff:** When should it suggest contacting the expert or business? > [!NOTE] > > You do not need a perfect first draft. Start with clear decisions and improve the configuration through testing. ## Step 1: Review your experience brief Revisit the decisions you made in the earlier Builder guides. Look for contradictions or missing details. For example, if the experience is described as *educational*, its instructions should not also tell it to make definitive professional recommendations without appropriate safeguards. Use these guides if you need to refine your plan: * [Define Your AI Experience](/ai-experience-builder/define-your-ai-experience) * [Describe Your Audience](/ai-experience-builder/describe-your-audience) * [Design the Conversation](/ai-experience-builder/design-the-conversation) * [Prepare Your Knowledge](/ai-experience-builder/prepare-your-knowledge) ## Step 2: Generate a structured draft Use the AI Experience Builder to turn your planning inputs into a structured configuration. Depending on the Builder workflow available to you, this may be presented as generated instructions, a configuration document, or organized content to transfer into Agentsy. A useful draft should address the following areas: | Configuration area | What it should define | | - | - | | Identity and purpose | Who the assistant represents and what it helps visitors do | | Audience and intent | Whom it serves and how to recognize different visitor needs | | Communication style | Tone, language, clarity, and response structure | | Conversation guidance | How to welcome, explore, explain, and follow up | | Knowledge usage | Which approved sources to rely on and how to handle missing information | | Safety and boundaries | Topics, claims, and actions that require limits or caution | | Human handoff | When and how to direct visitors to the appropriate person or next step | > [!WARNING] > > Do not assume generated content is automatically accurate, compliant, or aligned with your business. Review it before using it with real visitors. ## Step 3: Review the system instructions System instructions guide the assistant's behavior. Read the generated instructions as if you were a first-time visitor interacting with the experience. Check that they: 1. Describe the assistant's role without exaggerating its authority. 2. Put visitor needs ahead of unnecessary sales prompts. 3. Encourage helpful, natural conversation rather than a rigid questionnaire. 4. Require the assistant to acknowledge uncertainty and avoid inventing facts. 5. Specify when the visitor should speak with a human. 6. Avoid requesting sensitive personal information that the experience does not need. For deeper guidance, see [System Instructions](/create-and-configure/system-instructions). ## Step 4: Check the knowledge plan A configuration can describe what the assistant *should* know, but it does not automatically provide that knowledge. Confirm that the documents and reference materials mentioned in the draft actually exist, are approved for use, and are suitable for the intended audience. Remove outdated or conflicting material before uploading it. For example, an assistant for a boat advisor can explain vessel types, ownership considerations, and buying processes using prepared knowledge. It should **not** claim to know which boats are currently available unless a verified, current inventory source is connected. Learn more in [Knowledge Base Overview](/knowledge-management/knowledge-base-overview) and [Upload Documents](/knowledge-management/upload-documents). ## Step 5: Decide what belongs in Agentsy The Builder's output is a starting point. Implement the relevant pieces in the appropriate areas of your Agentsy account: * **Assistant configuration:** Identity, role, behavior, conversation guidance, and boundaries. * **Knowledge management:** Approved documents and reference materials. * **Branded website:** Name, copy, logo, colors, images, and calls to action. * **Access settings:** Public or restricted access and visitor registration, as applicable. * **Human follow-up:** A clear next step, such as a contact or scheduling option, where configured. The exact field names and available controls may vary by account and product updates. Follow the labels shown in your Agentsy dashboard. > [!NOTE] > > Generating a configuration is **not the same as publishing an AI Experience**. You still need to apply the configuration, test the assistant, and complete the launch steps in Agentsy. ## Example: A leadership consultant A consultant wants prospective clients to explore their leadership methodology before scheduling a meeting. Their configuration might define: * **Purpose:** Help visitors identify leadership challenges and understand relevant concepts. * **Style:** Thoughtful, practical, conversational, and concise. * **Conversation:** Begin with the visitor's challenge; ask one useful follow-up at a time. * **Knowledge:** Use the consultant's approved framework, articles, and examples. * **Boundaries:** Do not promise outcomes or claim to have diagnosed a workplace problem. * **Handoff:** Suggest a conversation with the consultant when the visitor wants personalized support or implementation. This gives the consultant a clear foundation for testing without turning the experience into an automated sales script. ## Step 6: Validate the draft before launch Test the configuration against realistic questions: * Can the assistant explain its purpose in plain language? * Does it help a curious visitor without pushing immediately for a meeting? * Does it handle a question outside its knowledge honestly? * Does it respond appropriately when a visitor asks for a real person? * Does it avoid presenting unverified availability, pricing, or professional advice as fact? * Does it respect the visitor's language and level of expertise? Revise the instructions and knowledge as needed. Testing often reveals gaps that are difficult to spot in a configuration document alone. ## Common mistakes to avoid > **Treating generated instructions as final** > > The draft needs human review, particularly for business-specific facts, professional boundaries, and regulatory requirements. > **Putting every fact into system instructions** > > Keep behavior rules in system instructions and substantive reference information in the knowledge base where appropriate. > **Writing overly aggressive conversion rules** > > Give visitors useful answers first. Introduce a human next step when their needs or interest make it relevant. > **Assuming the Builder uploads documents or publishes the site** > > Confirm each action in the Agentsy platform. A generated plan does not by itself establish that documents were uploaded or the experience is live. ## What's next? Continue to [Review and Launch](/ai-experience-builder/review-and-launch) to validate the experience and prepare it for visitors. You can also follow [Create Your First AI Experience](/getting-started/create-your-first-ai-experience) for the broader platform setup journey. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.