Customer service teams drown in repetitive questions, leads go cold outside business hours, and generic chatbots make things worse — they hallucinate answers and can't actually do anything.
What I built
Luísa is a production AI customer service agent. She answers from the business's real data via RAG, remembers each customer between conversations, books meetings directly into the calendar, and hands critical actions to a human for approval before anything irreversible happens.
RAG on business data — answers come from the company's own knowledge base, not from the model's imagination
Persistent memory — returning customers are recognized and context carries over
Scheduling — she checks availability and books meetings in the calendar (CRM + Calendar integration)
Human-in-the-loop — sensitive actions wait for human approval instead of running blind
Production-grade — error handling and monitoring, built with Claude + n8n
Production AI agent: answers from business data via RAG, remembers customers, books meetings, human-in-the-loop on critical actions. Built with Claude + n8n.