> ## 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. # Measure Engagement > Understand how visitors engage with your Agentsy AI Experience and use meaningful signals to improve conversations and human follow-up. # Measure engagement Publishing your AI Experience is only the beginning. Understanding how visitors use it can help you improve the experience, identify recurring questions, and create better opportunities for human conversations. The goal is not simply to increase the number of chats. It is to understand whether visitors are finding value and progressing toward useful next steps. > [!NOTE] > > The metrics and reports available to you depend on your Agentsy account and configuration. Use the analytics and conversation views available in your workspace; not every measure described below is necessarily an automated dashboard metric. ## 1. Start with a clear goal Choose one primary outcome for your AI Experience before reviewing its performance. | Experience goal | Useful questions to investigate | | - | - | | Educate prospects | Are visitors asking relevant questions and receiving useful explanations? | | Help visitors explore options | Do conversations reveal needs, constraints, or preferences? | | Build trust in an expert | Do visitors engage with the expert's methodology or point of view? | | Create human conversations | Do visitors express readiness to speak with someone or use the contact or booking option? | | Learn about market demand | Which questions and challenges appear repeatedly? | A successful educational experience may have different signals from one designed to help visitors request a consultation. ## 2. Understand the engagement journey Consider engagement as a sequence rather than a single number: 1. **Visit:** Someone opens your branded AI website. 2. **Start:** The visitor begins interacting with the assistant. 3. **Explore:** The visitor asks questions or follows up on answers. 4. **Reveal needs:** The conversation provides context about what the visitor wants to accomplish. 5. **Take a next step:** The visitor registers, requests human help, or uses a relevant call to action, depending on your configuration. Visitors do not have to complete every stage. Someone may get the answer they need in one exchange, while another person may return with more complex questions. ## 3. Review quantitative signals Where the relevant data is available, track signals such as: | Signal | What it can tell you | Important limitation | | - | - | - | | Website visits | How many visits your experience attracts | Visits alone do not show meaningful engagement | | Conversations started | Whether visitors choose to interact | A started chat does not establish satisfaction | | Messages exchanged | How much conversation activity occurs | More messages can also indicate confusion | | Registrations | How many visitors create an identifiable account | Registration is not the same as qualification | | Contact or booking actions | Whether visitors attempt to continue with a person | An action does not guarantee a completed meeting | | Repeat questions | Which topics deserve clearer explanations | Repetition can signal demand or missing information | Some signals may require your own website analytics, booking system, or manual review rather than being available directly in Agentsy. ### A simple engagement worksheet You can record observations weekly: | Period | Visits (if available) | Conversations | Registrations | Human follow-up requests | Main topics | | - | - | - | - | - | - | | Week 1 | — | — | — | — | — | | Week 2 | — | — | — | — | — | Only compare numbers collected with consistent definitions and tracking methods. ## 4. Look beyond the numbers Conversation quality often matters more than message volume. Review conversations you are authorized to access and ask: * Did the assistant understand the visitor's actual question? * Did it provide accurate, relevant information? * Did the visitor ask a meaningful follow-up? * Did the conversation reveal an unmet need or knowledge gap? * Was a human handoff suggested when appropriate? * Did the assistant avoid pressuring someone who only wanted information? > [!WARNING] > > Treat conversation content as potentially personal information. Review it only with appropriate permissions, avoid copying sensitive details into spreadsheets, and follow your organization's privacy and retention policies. ## 5. Distinguish interest from commercial intent A visitor asking about a service is not automatically ready to buy. For example, a leadership consultant might observe these interactions: | Visitor says | Possible interpretation | Appropriate response | | - | - | - | | “What is executive coaching?” | Early exploration | Provide a helpful explanation | | “How would your approach help a new manager?” | Exploring relevance | Explain the methodology with an example | | “Can this work for our leadership team of 20?” | More specific organizational need | Clarify the context and relevant options | | “Can we schedule a conversation about our team?” | Explicit interest in human follow-up | Point to the available booking or contact option | These are **contextual clues**, not automatic lead scores. Avoid assuming budgets, urgency, authority, or purchase readiness unless the visitor explicitly provides that information. For more detail, see [Lead Identification](/conversations-and-leads/lead-identification) and [Understanding User Intent](/conversations-and-leads/understanding-user-intent). ## 6. Identify opportunities to improve Use what you learn to make targeted changes. ### Visitors rarely start a conversation Review the landing page headline, supporting text, and primary call to action. Make it clear what visitors can ask and what value they will receive. ### Visitors start but quickly stop Test the welcome message and first answer. Check whether the assistant asks too many questions, gives generic replies, or produces unnecessarily long responses. ### Visitors repeatedly ask questions the assistant cannot answer Review your knowledge documents. Add missing explanations or clarify where the assistant should acknowledge limitations and recommend human assistance. ### Visitors show interest but do not take a next step Check whether the booking or contact option is visible, relevant, and working. Avoid adding aggressive prompts; make the next step easy when the visitor expresses readiness. ### Conversations contain incorrect or outdated information Correct the source material, review system instructions, and retest. Do not optimize conversion at the expense of accuracy. ## 7. Run a simple improvement cycle **1. Review recent activity** Look at the engagement data and conversations available to your organization. **2. Identify one recurring issue** Choose a specific pattern, such as unclear opening copy or unanswered questions. **3. Make one focused improvement** Update the landing page, knowledge base, assistant instructions, or human handoff as appropriate. **4. Test the experience** Run realistic visitor questions before and after the change. **5. Compare results over time** Look for a consistent improvement in usefulness, clarity, or relevant next-step actions. Avoid drawing conclusions from very small samples. ## 8. Example: A boat-buying advisory experience A boat broker launches an educational AI Experience. After reviewing permitted conversations, the team notices that visitors frequently ask about the differences between new and used boats and then ask to see currently available models. The team can improve the experience by: 1. Expanding educational content about new versus used boats. 2. Making it explicit that the AI does not have access to live inventory unless a real-time inventory integration is available. 3. Offering a natural handoff to a broker when visitors want actual listings or current availability. 4. Testing whether the assistant explains those boundaries accurately. This turns engagement observations into a better visitor experience without inventing listings or treating every question as a sales opportunity. ## Engagement review checklist * [ ] Define the main outcome of the AI Experience. * [ ] Identify which metrics are actually available and reliable. * [ ] Review recurring visitor questions and themes. * [ ] Check answer accuracy and knowledge coverage. * [ ] Distinguish registration from genuine buying intent. * [ ] Verify contact and booking actions work correctly. * [ ] Protect visitor privacy during analysis. * [ ] Make one improvement at a time and test it. ## Related guides * [Conversation Insights](/conversations-and-leads/conversation-insights) * [Lead Identification](/conversations-and-leads/lead-identification) * [Understanding User Intent](/conversations-and-leads/understanding-user-intent) * [Lead Generation Strategies](/share-and-grow/lead-generation-strategies) * [Improve Conversions](/share-and-grow/improve-conversions) * [Troubleshoot AI Responses](/knowledge-management/troubleshoot-responses) This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.