The AI customer service glossary
36 terms you'll meet when you put an AI agent in front of customers, defined in a sentence or two.
A
- AI agent
- Software that understands a customer's message, finds the answer in approved sources and can take actions such as looking up an order, then hands off to a person when needed. Learn more
- AI chatbot
- A program that replies to messages automatically. Older chatbots follow fixed scripts; AI chatbots use language models to understand free text. Learn more
- AI credits
- A usage allowance for AI work such as replies, retrievals and actions, included in a plan and consumed as the agent works.
- AI resolution rate
- The share of conversations the AI agent closes without a person stepping in. Learn more
- AI video avatar
- A computer-generated presenter that speaks and appears on a video call, driven by an AI agent. Learn more
- AI voice agent
- An AI agent customers talk to out loud in real time, with speech recognition and a synthetic voice. Learn more
- Allowed domains
- The list of websites where a chat widget is permitted to load, which stops others reusing your key.
B
- Business-initiated message
- On WhatsApp, a message a business sends first, which must use an approved template. Learn more
C
- Canned response
- A saved reply a support agent can insert quickly. In LiveCart, reviewed replies play this role for the AI agent too.
- Chat widget
- The chat window and launcher button embedded on a website. Learn more
- Citation
- A reference to the source document an AI answer was drawn from.
- Co-browsing
- A support technique where an agent sees and guides a customer's browser session.
- Containment rate
- Another name for AI resolution rate: conversations contained by automation.
- Conversation funnel
- The stages a conversation moves through, for example started, engaged, qualified and converted. Learn more
- CSAT
- Customer satisfaction score, usually from a short survey after a conversation.
D
- Deflection
- Answering a question before it becomes a ticket for a person.
E
- Escalation
- Moving a conversation to a person or a more senior team. Learn more
F
- First response time
- How long a customer waits for the first reply.
- Fallback reply
- What an AI agent says when it can't answer, instead of guessing.
G
- Grounding
- Restricting an AI agent's answers to approved sources so it doesn't invent facts. Learn more
- Guardrails
- Rules that keep an AI agent on topic and safe: blocked subjects, limits and required handoffs. Learn more
H
- Hallucination
- When an AI model states something that isn't supported by its sources.
- Human handoff
- The moment an AI agent passes a conversation to a person, ideally with context attached. Learn more
I
- Intent
- What the customer is trying to do, such as track an order or start a return.
K
- Knowledge base
- The collection of documents and data an AI agent answers from. Learn more
L
- Large language model (LLM)
- The kind of AI model that reads and writes natural language, used inside AI agents.
O
- Omnichannel support
- Serving customers across website chat, messaging apps and calls with a shared history. Learn more
P
- Persona
- The identity, tone and goal you give an AI agent.
R
- Retrieval-augmented generation (RAG)
- A method where the system first retrieves relevant documents, then the model writes an answer from them.
- Routing
- Rules that decide which person or team gets a conversation.
S
- Story reply
- A message someone sends by replying to an Instagram story. Learn more
- Sub-agent
- A focused AI agent that handles one part of a larger job, called by a main agent.
W
- Warm transfer
- Handing a call or chat to a person while passing on what's already been said.
- WISMO
- Short for 'where is my order?', the most common post-purchase support question. Learn more
- Workflow
- A sequence of steps an AI agent follows to complete a task, such as verifying a customer before showing an order. Learn more
Keep exploring
- BlogHow to stop your AI agent making things up: grounding with a knowledge baseHow to build a knowledge base that keeps your AI agent accurate: which sources to add, how to organise and test them, and how to stop made-up answers.Learn more
- Blog10 guardrails to set before your AI agent talks to customersThe guardrails every customer-facing AI agent needs: grounding, blocked topics, verification, confirmation steps, limits, disclosure and review.Learn more
- BlogLive chat vs AI agent: the real difference, and why you want bothLive chat connects customers to people; an AI agent answers and acts on its own. Here's how they differ, where each wins, and how to combine them.Learn more
- BlogMeasuring AI agent performance: beyond “conversations handled”The metrics that show whether your AI agent is working: resolution rate, handoff reasons, accuracy reviews, first reply time and revenue influence.Learn more
Try it on your own store
Paste one snippet. Ask it what your customers ask.