Insight Sell

Context

Commercial conversations on WhatsApp contain valuable information about service, objections and conversion, but are rarely treated as structured data.

Problem

How can unstructured conversations become operational insight — useful metrics, patterns and reports — without depending on manual annotation?

Decisions

• Conversations are the primary data source; the CRM is not a substitute
• Relational storage in PostgreSQL through Supabase
• LLM analysis with validation to avoid incorrect inferences
• Dashboards as the operational reading layer

Architecture

WhatsApp → Ingestion → PostgreSQL (Supabase)
                          │
                          ├── Analysis (LLM)
                          │       └─ context, validation, categorization
                          │
                          └── Metrics → Dashboards
                                  └─ service, conversion, satisfaction

Role

I conceived and developed the project end to end, including system design, data modeling and AI integration.

Status

Details are restricted due to confidentiality.