DentBot: WhatsApp AI automation in production
How Pablo Costa built and operates DentBot, connecting WhatsApp, business workflows and human operators, with deterministic control over external actions.
Overview
DentBot is an AI customer-service product with a dental-clinic deployment in production. I have built and operated it since December 2025. It connects WhatsApp conversations, business workflows and human operators.
The engineering problem
A conversational system needs to do more than generate a reply. It must maintain conversation state, route work and support a handoff to a human operator. Fragmented incoming messages also need to be handled without producing competing responses.
Architecture and responsibilities
The production system connects WhatsApp, Chatwoot, n8n and LLM APIs to PostgreSQL/Redis-backed state. Structured model outputs support intent classification and handoff decisions. Permissions, routing and external side effects remain under deterministic software control: explicit application rules decide which actions are allowed.
Handling fragmented messages
Message debouncing consolidates fragmented inputs before responding. This helps prevent competing responses when a person sends several short messages in succession.
Production system and backend modernization
DentBot Core is a separate Python/FastAPI modernization focused on typed contracts, domain rules and testability. It should not be confused with a claim that every component of the existing production deployment already uses this architecture.
Evidence and limits
The public engineering case study and modernization repository provide additional context. This portfolio does not publish customer conversations or claim a measured conversion uplift, response-time improvement or uptime percentage.