MAIRA Chat CRM

Growing customer conversations create missed follow-ups, fragmented context, duplicated replies, and poor coordination between automation and human operators.

A basic chatbot was not sufficient. The system needed AI automation with operational CRM control.

I designed the product around conversation ownership, customer state, AI context, and event-driven integrations.

• AI-assisted responses with controlled history and business context.
• Human takeover that disables AI ownership when staff is assigned.
• Follow-up timers reset from the customer's latest message.
• PostgreSQL for transactional state and Redis/worker patterns for async workflows.
• Webhook-oriented integration with external MAIRA services.

The difficult part was not calling an LLM API. It was making AI behavior respect business state.

Conversation ownership is part of the AI execution model, not just a UI status.

The system was hardened around API failures, duplicate events, message state, provider errors, authentication, role-based access, secure secrets, and production diagnostics.

Product workflow definition, architecture, AI orchestration, CRM behavior, API integration, database workflow, testing, deployment, and production troubleshooting.