Agentic RAG in Production: Orchestration, Evaluation & ROI - Rohit Bhardwaj

This session introduces the concept of Agentic Retrieval-Augmented Generation (RAG), addressing common challenges that lead to stalled initiatives, such as hallucination and lack of measurable business value. The speaker discusses how to leverage Salesforce Agentforce, Salesforce Data Cloud, and LangGraph to create robust, self-correcting RAG systems. Additionally, the talk explores evaluation frameworks like the RAG Triad, TruLens, and Ragas to assess accuracy, fairness, and ROI. The presentation concludes with a live demonstration of a Smart Checkout Helper scenario that showcases grounded retrieval, agent-driven correction, and quantified business impact.