AI customer support glossary
Plain-language definitions of the vocabulary buyers, support leaders and vendors use to describe AI customer support — defined for the category first, then grounded in how Evoriqa applies each one.
AI customer support has arrived with a vocabulary of its own, and much of it is used loosely. The same word can mean two different things on two vendors' pricing pages, and a metric one team treats as a headline number another treats as a footnote. This glossary defines the terms the way the wider industry uses them — independent of any single product — so you can read a pricing page, a contract or a comparison and know exactly what is being claimed.
The terms fall into a few clusters that build on each other. One set describes what an AI agent is and does: an AI support agent, the grounded answers it produces, the retrieval-augmented generation that keeps those answers tied to your content, and the agentic actions it takes when a reply alone is not enough. A second set describes where support happens: omnichannel support and the unified inbox that pulls every channel into one queue. A third set is about the moment automation reaches its limit — the knowledge gap that exposes what your content cannot answer, the human handoff that passes the conversation to a person, and the conversation handoff context that decides whether that person starts cold or fully briefed.
The last cluster is how any of this gets measured. Ticket deflection, deflection rate, containment rate, first response time and CSAT are the numbers people quote to prove an AI agent is working — and they are the easiest to misread or inflate. We define each one, note where the definitions diverge in practice, and are deliberate about one thing: we publish no benchmark figures. There is no honest "industry average" deflection rate, because the only rate that describes your support is the one measured on your own traffic. Each metric page explains how to measure it yourself instead.
Two more terms cover the commercial side: white-label SaaS, the model that lets an agency resell a platform under its own brand, and AI credits, the usage-based unit that meters what the models actually cost. Every definition below stands on its own; where a term touches Evoriqa, a short, clearly separated section says how — and links to the product page that puts it to work.
- Agentic Action
An agentic action is a task an AI agent performs inside a conversation, such as booking, order lookup or ticket creation, rather than only replying with text.
- AI Credits
AI credits are a usage-based billing unit that meters each AI action against its real model cost, so you pay for what you use rather than a flat per-seat fee.
- AI Support Agent
An AI support agent is software that answers customer questions in natural language from a defined knowledge source, and can act on the user's behalf.
- Containment Rate
Containment rate is the share of support conversations an automated system handles end to end, without ever escalating to a human agent.
- Conversation Handoff Context
Conversation handoff context is the history and customer detail passed to a live agent at escalation, so they continue without the customer starting over.
- CSAT (Customer Satisfaction Score)
CSAT, or customer satisfaction score, is the percentage of surveyed customers who rate an interaction positively, usually just after it closes.
- Deflection Rate
Deflection rate is the percentage of incoming support conversations closed by automation or self-service without escalating to a human agent.
- First Response Time (FRT)
First response time is how long a customer waits between sending a support message and receiving the first human or automated reply to it.
- Grounded Answer
A grounded answer is an AI reply built from a verified knowledge source and traceable back to it, rather than generated from the model's general training.
- Human Handoff
A human handoff is the moment an automated support conversation is passed to a live agent, ideally with full context so the customer need not start over.
- Knowledge Gap
A knowledge gap is a question customers ask that your support content cannot answer, leaving the agent, human or AI, without a grounded response to give.
- Omnichannel Support
Omnichannel support is a model where a customer moves between channels like chat, email and phone while the conversation and its context stay continuous.
- Retrieval-Augmented Generation (RAG)
Retrieval-augmented generation (RAG) fetches relevant passages from a knowledge source and feeds them to a language model so its answer stays grounded in fact.
- Ticket Deflection
Ticket deflection is the share of support questions resolved by self-service or automation before a human agent ever opens the conversation.
- Unified Inbox
A unified inbox is a single queue where messages from every support channel — chat, email, social and more — arrive and are handled by one team.
- White-Label SaaS
White-label SaaS is software one company builds and another resells under its own brand, so end customers see the reseller, not the original vendor.
See grounded AI support on your own content
Point Evoriqa at your website and help docs, and your agent starts answering questions, capturing leads and booking appointments across every channel. Free to start, no credit card.