Business diagnostics and operational insights for sales and ops teams on Clover POS

The Challenge

Sales and operations teams running on Clover POS had transaction data everywhere but no systematic way to diagnose where the business was actually underperforming — labor patterns, inventory friction, day-to-day operational drag. Retrospectives surfaced symptoms, not root causes. Analysts spent hours in spreadsheets triangulating across reports. The client needed CASS to guide sales and ops people through structured discovery conversations and deliver operational insights they could act on — not a chatbot that regurgitated documentation.

Our Approach

We designed a multi-stage agentic pipeline where each session is an active diagnostic, not a Q&A exchange. Intent classification routes each query to the right reasoning chain. A query reformulation step converts ambiguous inputs into tractable retrieval targets. A ReAct loop drives iterative reasoning — forming hypotheses, checking them against retrieved context, and refining conclusions before presenting output. Redis provides session persistence so diagnostics can resume across interactions without losing context.

What We Built

• •Multi-tenant RAG pipeline with per-tenant knowledge base isolation
• •Intent classification and query reformulation pre-processing layer
• •ReAct reasoning agent with tool-augmented retrieval and hypothesis validation
• •Redis-backed session persistence for multi-turn diagnostic conversations
• •Stripe-integrated usage billing per tenant with configurable tiers
• •REST API for embedding the diagnostic engine into existing enterprise portals