How AI agents work
An AI agent receives a customer message, retrieves relevant context from a knowledge base and connected systems (CRM, order data, billing), then drafts a response or takes an action like issuing a refund or updating a contact record.
Modern AI agents use retrieval-augmented generation (RAG) to ground their answers in a company's actual documentation, reducing hallucinations. They also use tools — small functions exposed by the support platform — to perform actions instead of just generating text.
AI agent vs chatbot
A traditional chatbot follows a fixed decision tree: if the user picks option A, show response A. An AI agent uses a language model to interpret free-form messages, take multi-step actions, and recover from unexpected inputs.
In practical terms: chatbots are good at structured flows (book a meeting, route a ticket), AI agents are good at open-ended support (answer a refund question, troubleshoot a setup issue, update an account).
Common use cases in customer support
Resolving FAQ-type questions (shipping, returns, account access) without human involvement, drafting suggested replies for human agents to review, summarizing long conversation threads before handoff, and tagging or routing tickets based on intent.
AI agents are typically measured on deflection rate (share of conversations resolved without a human), CSAT on AI-handled conversations, and AI containment time.
AI agent in Savena
Savena ships an AI agent in every plan. It is trained on your help center, past conversations, and connected integrations like Shopify, HubSpot, and Stripe. It resolves common questions, takes actions like looking up orders, and escalates to humans when confidence is low — included in the seat price.