Glossary
AI customer support, term by term
32 terms that come up when you compare support tools, fill in a security questionnaire or read a support dashboard. Each one says what it means, when it matters and where the definitions split.
- AI support agentSoftware that answers customers in chat from your own content, takes actions such as an order lookup, and hands over to a person when it cannot help.
- ChatbotAny program that holds a text conversation, either by following scripted buttons and keyword rules or by generating replies with a language model.
- AI agentSoftware built on a language model that chooses actions, runs them through tools and reads the results before deciding what to do next.
- CopilotAn AI assistant for your human agents rather than your customers, drafting replies and summarizing conversations inside the inbox.
- Large language model (LLM)A neural network trained on vast amounts of text to predict the next token, which lets it write replies, summarize and follow instructions.
- HallucinationA confident answer that is false or unsupported by its source, such as an invented refund policy; NIST calls it confabulation.
- Tool callThe model's structured request to run one operation you defined, such as an order lookup, with typed inputs and a result it reads before replying.
- Knowledge baseThe private collection of pages, files and snippets an AI support agent searches before every reply, not the same thing as a public help center.
- Knowledge sourceOne item in a knowledge base, such as a crawled website, an uploaded file or a connected Notion workspace, and often the unit a plan limits.
- Retrieval-augmented generation (RAG)Searching your content for passages relevant to a question and handing them to the model as it writes the answer.
- EmbeddingA list of numbers that represents the meaning of a text, so that texts with similar meanings get similar numbers and can be matched by meaning.
- GroundingMaking the model answer from source text supplied as it replies, not from training memory, so every claim traces back to your content.
- CrawlAn import that starts at one address and follows a site's links to collect pages for the knowledge base, usually with a page and depth limit.
- System promptThe hidden text an application places before every conversation, telling the model what it is, what to answer from and what never to do.
- Agent instructionsThe part of the system prompt you write yourself: persona, tone, focus and refusals, layered on top of the vendor's base rules.
- Welcome messageThe first message a chat widget shows before the visitor types, also called a greeting, which should say in a line or two what the agent can do.
- Suggested questionsTappable buttons shown with the greeting, also called conversation starters or quick replies, that start a conversation with that text.
- Fallback messageWhat a chatbot says when it cannot give a grounded answer, because the content is missing, the question is unclear or a rule blocks it.
- Human handoffThe transfer of a live conversation from an AI agent to a person, with the transcript and collected details, so the customer need not repeat anything.
- Escalation triggerA condition that makes the AI stop and hand over, such as a request for a person, no relevant content, a failed action or visible frustration.
- Team inboxThe shared view where your team watches conversations live, takes over the ones the AI hands off, and closes them.
- Away messageWhat the agent tells a visitor who asks for a person while nobody can reply live, with the next step and when to expect a reply.
- TicketA support request recorded in a helpdesk, with a number, status, owner and history, which an AI agent can open when a chat needs follow-up.
- First response time (FRT)The time from a customer's first message to the first substantive reply, reported per channel; definitions split on whether an instant AI reply counts.
- Deflection rateThe share of would-be tickets that never reach a person because the customer found an answer first; a figure that can only be estimated.
- Containment rateThe share of bot conversations that ended without escalating to a person; it shows where chats ended, not whether anyone was helped.
- Resolution rateThe share of AI-handled conversations that ended with the problem solved, by each vendor's own definition, which is where the figures diverge.
- Customer satisfaction score (CSAT)The share of customers who rate an interaction positively in a one-question survey, such as 4 or 5 out of 5 or a thumbs-up.
- Verified visitorA chat visitor who has proved control of an email address in the current session, usually with a one-time code, which unlocks personal features.
- One-time codeA short secret sent to a channel the customer controls and accepted once, within minutes, to prove they control that email address or phone.
- Meeting memoryA feature that lets an AI agent recall a returning, verified customer's past meetings with your team, such as commitments and summaries, from notes.
- Prompt injectionText written to make a model ignore its instructions, typed into the chat or planted in a page or file the model reads.
Want the one-page version? The blog has 60 terms with one-sentence definitions, including pricing units and compliance vocabulary.