Glossary

Plain definitions of customer-service AI terms. One question per page. What it is, where it gets used, where it misleads.

• What are AI guardrails in customer service?

  AI guardrails are the rules and checks that limit what an AI agent can say or do. Plain explainer of the three layers, what they prevent, and the trade-off they create.

• What is agent-assist in customer service?

  Agent-assist is software that helps a live support agent during a contact — suggesting replies, surfacing knowledge, drafting wrap-ups. Plain explainer.

• What is AHT (Average Handle Time) in customer service?

  AHT is the average time to handle one customer contact end-to-end. The headline efficiency metric in contact centres — useful and misleading depending on what you do with it.

• What is an AI agent in customer service?

  An AI agent handles a customer message end-to-end — answers, looks up, refunds, escalates — without a human stepping in. Different from a chatbot in one key way: it takes actions, not just text.

• What is CCaaS (Contact Centre as a Service)?

  CCaaS is cloud-hosted contact-centre software: routing, IVR, agent desktop, reporting, and increasingly AI. Here's what's in it and what it isn't.

• What is containment rate in AI customer service?

  Containment rate is the share of customer conversations that finished without ever reaching a human agent. Useful directional signal, easy to misread on its own.

• What is deflection rate in AI customer service?

  Deflection rate is the share of customer contacts an automated system handles without escalating to a human. Useful, but easy to misread.

• What is escalation rate in AI customer service?

  Escalation rate is the share of AI conversations handed off to a human. The mirror of containment — and the one that tells you whether the handoff actually works.

• What is FCR (First Contact Resolution) in customer service?

  FCR is the share of customer issues resolved in the first contact, without the customer coming back. The metric closest to actual customer value — and the one most teams measure laziest.

• What is hallucination in customer-support AI?

  Hallucination is when an AI agent confidently invents an answer that isn't true. Plain explainer of why it happens, where it shows up in support, and how to reduce it.

• What is RAG (retrieval-augmented generation) in customer support?

  RAG is the system that lets an AI agent read your help articles before answering, so it stops making things up. Plain explainer of what it is, how it works, and where it fails.

• What is recontact rate in customer service?

  Recontact rate is the share of customers who contact you again about the same problem after the first ticket was closed. The metric that exposes the gap behind deflection.

• What is resolution rate (and how is it different from deflection)?

  Resolution rate is the share of customer conversations where the actual problem was solved. Stricter than deflection or containment, and the number a CFO should care about.

• What is voice AI in customer service?

  Voice AI in customer service is software that answers calls, understands intent, and speaks back. Here's what it does, where it works, and how to test it.

14 entries.