> ## 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. # Conversation settings > Design a welcoming, useful conversation experience for visitors to your Agentsy AI assistant. # Conversation settings Conversation settings shape how people begin and experience a conversation with your AI assistant. A thoughtful setup helps visitors understand what the assistant can do, ask questions comfortably, and recognize when speaking with a person is the best next step. > [!NOTE] > > Some conversational behaviors are controlled by your assistant's **system instructions**, while others may be configured through your Agentsy workspace or AI Experience website. Available controls can vary by account and product configuration. ## 1. Start with a clear welcome message Your opening message should introduce the assistant, explain its purpose, and invite the visitor to share what they need. A useful welcome message typically includes: * **Identity:** Who the assistant represents. * **Purpose:** What visitors can explore or understand. * **Invitation:** One natural way to begin. For example, a consultant might use: > Hi! I'm the AI guide for \[Business Name]. I can help you explore our approach and think through your current challenges. What would you like to work on? Avoid long introductions, exaggerated claims, or several questions at once. ## 2. Let visitor intent guide the conversation Visitors arrive with different goals. Some want a quick answer; others are comparing options, exploring a problem, or ready to speak with the business. Configure the assistant to: 1. Respond to the visitor's actual question first. 2. Ask clarifying questions only when they help provide a better answer. 3. Adapt the level of detail to the visitor's needs. 4. Offer a relevant next step when interest becomes clear. **Example:** If a visitor asks, “How does your coaching approach work?”, explain the approach before asking about their goals. If they then ask how to get started, explain the available way to contact or book with the coach. ## 3. Define the response style Use system instructions to establish a consistent voice. Consider: | Setting to define | Example guidance | | - | - | | Tone | Warm, professional, and straightforward | | Length | Start concise; expand when the visitor asks | | Structure | Use short paragraphs and lists when helpful | | Questions | Ask one meaningful question at a time | | Terminology | Explain unfamiliar terms in plain language | | Language | Respond in the visitor's language when supported | The assistant should feel like an extension of your expertise and brand—not a generic sales script. ## 4. Separate behavior from interface settings Not every part of the conversation is configured in the same place. | What you want to change | Where to start | | - | - | | Assistant identity, tone, and boundaries | System instructions | | How the assistant handles uncertainty | System instructions and knowledge sources | | Answers about your services or methodology | Knowledge management | | Welcome message and presentation | Relevant assistant or website settings, if available | | Public, private, or registration-based access | Access and registration settings | | Booking or contact options | AI Experience website and calls to action | > [!NOTE] > > The exact names and locations of controls may differ in your Agentsy dashboard. If you cannot find a particular setting, check the related guide or contact support rather than assuming it is available on your plan. ## 5. Plan for uncertainty and missing information A trustworthy assistant should acknowledge when it lacks sufficient information. For example, instruct it to: * Avoid inventing prices, availability, credentials, or policies. * Distinguish general guidance from verified business information. * Ask for clarification when the request is ambiguous. * Suggest contacting the business when a current or personalized answer is required. For changing information—such as live inventory, appointment availability, or individualized quotes—do not imply that the assistant has real-time access unless the relevant capability has actually been configured and tested. ## 6. Make human follow-up feel natural An AI Experience should help visitors understand their options, not prevent them from reaching a person. Identify moments when a human conversation would be more useful, such as when a visitor: * Requests a quote or personalized proposal. * Wants to discuss a specific engagement. * Asks for current availability or information the assistant cannot verify. * Requests professional advice beyond the assistant's permitted scope. * Explicitly asks to speak with someone. When a contact or booking action is available on your AI Experience website, the assistant can direct the visitor to it. Avoid claiming that a meeting has been booked unless a booking system confirms it. ## 7. Consider privacy and visitor access Decide what information the assistant genuinely needs to provide a useful conversation. Do not request sensitive personal information merely to personalize the experience. If your AI Experience uses visitor registration, guest access, or invite-only access, review those settings separately. Registration and access controls are not substitutes for appropriate data-handling practices. For regulated or sensitive topics, define clear boundaries and appropriate human escalation in the system instructions. ## 8. Test the conversation before publishing Run realistic conversations from the perspective of different visitors. * [ ] A new visitor asks what the assistant can help with. * [ ] A visitor asks a straightforward question about your expertise. * [ ] A visitor provides little context and needs gentle clarification. * [ ] A visitor asks several follow-up questions on the same topic. * [ ] A visitor asks about something outside the assistant's knowledge. * [ ] A visitor asks to speak with a person. * [ ] A visitor writes in another language you intend to support. * [ ] A visitor requests sensitive information or personalized professional advice. Review whether the assistant answers clearly, maintains context, avoids unsupported claims, and recommends human follow-up at the right time. > [!TIP] > > Test with real questions your prospects commonly ask. A short set of realistic scenarios is more useful than relying only on greetings such as “Hi” or “Hello.” ## Common issues ### The assistant asks too many questions Adjust the system instructions so it answers the visitor's question before gathering additional context. Limit follow-up questions to those necessary for useful guidance. ### Responses are too long or technical Specify the desired level of detail and ask the assistant to explain specialist terms simply. Test again with non-expert questions. ### The assistant pushes visitors to book too early Clarify that human follow-up should be offered when the visitor shows interest, requests it, or needs assistance beyond the assistant's scope—not after every response. ### The assistant gives inconsistent answers Review both the system instructions and the underlying knowledge materials. Conflicting, outdated, or incomplete sources can affect response quality. ## Related guides * [System instructions](/create-and-configure/system-instructions) * [Public and private access](/create-and-configure/public-and-private-access) * [Guest chat and registration](/create-and-configure/guest-chat-and-registration) * [Testing your assistant](/create-and-configure/testing-your-assistant) * [Design the conversation](/ai-experience-builder/design-the-conversation) * [Human handoff](/conversations-and-leads/human-handoff) This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.