> ## 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. # Glossary > A practical guide to the terms you'll encounter when building and managing an AI Experience with Agentsy. # Glossary This glossary explains the language used throughout the Agentsy Help Center. It includes Agentsy concepts and general AI terms so you can make informed decisions as you build your experience. > [!NOTE] > > Some terms describe general AI concepts rather than a specific feature available in every Agentsy plan or configuration. ## A ### AI assistant An AI-powered conversational guide that responds to visitor questions according to its instructions and the information available to it. In Agentsy, the assistant is part of a broader branded experience. ### AI Experience A branded, interactive destination where people can explore an expert's or organization's knowledge through an AI-powered conversation. An AI Experience can help visitors understand an offering and decide whether to take the next step with a human. ### AI Experience Builder A guided process for defining an AI Experience's purpose, audience, conversation approach, and supporting knowledge before configuring it for launch. Designing an experience and publishing it are distinct steps. ### AI model The underlying technology that generates the assistant's responses. Different models may vary in capabilities, speed, cost, and supported features. ### Analytics Information about how visitors engage with an experience, such as conversations and usage patterns. Available metrics depend on the account's configuration and enabled features. ## B ### Branding The visual and verbal identity of an AI Experience, including its name, logo, colors, imagery, and website copy. ## C ### Call to action (CTA) A prompt that invites a visitor to take a next step, such as booking a conversation, visiting a website, or contacting the organization. ### Chat session An individual conversation between a visitor and an AI assistant. A session may contain multiple messages. ### Conversation design The planning of how an assistant welcomes visitors, asks useful questions, explains concepts, handles uncertainty, and suggests next steps. ### Conversation insights Patterns and signals that can help an organization understand what visitors are asking and what they may need. Insights should be interpreted with appropriate attention to privacy and context. ### Custom domain A web address owned by an organization that can be connected to its branded AI Experience when supported by the account and properly configured. ## D ### DNS (Domain Name System) The system that directs a domain name to the correct online service. Connecting a custom domain commonly involves updating DNS records with the domain provider. ## F ### File search A method that helps an AI assistant retrieve relevant information from uploaded or connected knowledge sources when responding to questions. It does not guarantee that every answer will be correct. ### Follow-up A next action after a conversation, such as scheduling a meeting or requesting additional information from a person. ## G ### Guest chat A conversation experience that a visitor may be able to try before completing full registration, depending on how access is configured. ### Guardrails Instructions and design constraints that define what an assistant should and should not do, including how to handle uncertainty, sensitive information, and requests outside its scope. ## H ### Hallucination An AI-generated statement that sounds plausible but is incorrect, unsupported, or fabricated. Clear instructions, reliable knowledge, and testing can reduce this risk but cannot eliminate it completely. ### Human handoff The point at which an AI assistant directs a visitor to a real person for assistance, advice, a transaction, or a decision that requires human involvement. ## I ### Instructions (system instructions) The configuration that defines an assistant's role, objectives, tone, boundaries, and behavior. Instructions guide responses but do not replace accurate knowledge or ongoing testing. ### Intent What a visitor is trying to accomplish—for example, learning about a topic, comparing options, or requesting a consultation. ### Invite-only access An access setting in which participation is limited to authorized or invited visitors, when supported and enabled. ## K ### Knowledge base The collection of materials made available to an assistant to support its answers, such as documents, guides, policies, methodologies, and frequently asked questions. ### Knowledge retrieval The process of identifying relevant information from a knowledge source to help answer a visitor's question. ## L ### Lead A person who may be interested in an organization's expertise, products, or services. A conversation alone does not necessarily mean a visitor is a qualified lead. ### Lead identification The process of associating an interaction with information a visitor has provided, subject to registration settings and applicable privacy requirements. ### Landing page The public-facing page that introduces an AI Experience and invites visitors to interact with it. ## M ### Message A single contribution to a conversation, from either the visitor or the assistant. Plan limits may count messages according to the applicable billing rules. ### Multitenancy A software architecture in which different organizations use the same platform while their accounts and data are logically separated. ## P ### Private experience An AI Experience whose access is restricted rather than openly available to everyone. The exact restrictions depend on the selected access settings. ### Prompt Text supplied to an AI model to guide a task or response. A visitor's question is a prompt; system instructions provide additional behavioral guidance. ### Publishing Making an AI Experience available at its intended web address after configuration and testing. ## R ### Registration The process by which a visitor creates an account or supplies required details before accessing a conversation, depending on the experience's access settings. ### Retrieval-augmented generation (RAG) An AI approach that retrieves relevant external information and uses it as context when generating a response. Retrieval can improve grounding, but retrieved material must still be accurate and current. ## S ### Sensitive information Information that requires extra care, such as health, financial, identification, or confidential business details. Avoid collecting it through an AI Experience unless there is a clearly defined, appropriate, and compliant process. ### Source document A file or other reference material added to a knowledge base to support an assistant's responses. ### SSL/TLS certificate Technology used to help secure connections between a visitor's browser and a website. Custom domains typically require valid HTTPS configuration. ## T ### Testing Checking an AI Experience before and after launch using realistic visitor questions, difficult cases, and requests that should trigger a human handoff. ### Tone of voice The communication style an assistant uses, such as approachable, professional, educational, or concise. ## U ### Usage limit A plan-based allowance or restriction, such as the number of messages available during a billing period. Refer to your current plan for exact limits. ## V ### Vector store A system that stores representations of content so related information can be retrieved by meaning rather than only exact keyword matches. It may be used behind the scenes in knowledge retrieval. ### Visitor A person who opens an AI Experience, whether or not they have registered or started a conversation. ## Related guides * [Welcome to Agentsy](/getting-started/welcome-to-agentsy) * [What is an AI Experience?](/getting-started/what-is-an-ai-experience) * [How Agentsy works](/getting-started/how-agentsy-works) * [Create your first AI Experience](/getting-started/create-your-first-ai-experience) * [Frequently asked questions](/getting-started/faq) This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.