This is an AI customer support agent I designed and built to handle routine questions on its own and pass anything uncertain to a person. It is my own project, not a client engagement.
The problem
Most support inboxes are full of the same questions: where’s my order, what’s your return policy, how do I change my booking. Answering them by hand is slow. But a chatbot that guesses when it doesn’t know does more damage than no chatbot at all.
What I built
- Document search. The agent answers from the company’s own documents, using FAISS with FastEmbed embeddings to find the right passages.
- Order lookups. It can query order data through tools, so it can answer “where is my order?” with real information.
- Human hand-off. When it can’t answer with confidence, it doesn’t guess. It passes the case to a person with the conversation attached, so the customer never has to repeat themselves.
The agent is built with CrewAI.
Three rules behind the design
- Only the tools it needs. The agent can search documents and look up orders. Nothing else.
- Everything is logged. Every action is recorded, so you can see exactly what happened in any conversation.
- Risky or uncertain goes to a person. The agent handles routine questions; people handle the rest.
These rules come straight from my years running enterprise networks: least access, full logs and a plan for failure.
Where this applies to your business
Any business with a steady stream of repeated questions: ecommerce, clinics, home services, agencies, training providers.
Frequently asked questions
What does this AI customer support agent do?
It answers routine questions from the company’s own documents, looks up order information through tools, and passes anything it is not confident about to a person.
What happens when the agent does not know the answer?
It does not guess. It hands the case to a human with the full conversation attached, so the customer never has to repeat themselves.
Which technologies does it use?
CrewAI for the agent, FAISS with FastEmbed embeddings for document search, and tool connections for order lookups.
Which businesses is it suited to?
Any business with a steady stream of repeated questions, such as ecommerce stores, clinics, home services, agencies and training providers.
Try it free: take the 3-minute AI readiness quiz, see what missed calls cost you with the AI savings calculators, try the in-browser RAG chatbot demo or listen to the AI receptionist demo.
Related services: AI chatbot development · AI customer support automation · AI agent development
Book a free consultation to see how this would fit your support process.
AI/ML engineer building LLM agents, RAG systems and automation for growing businesses.
