Why My AI Service Advisor Forgot Everything
A practical memory strategy for LangGraph workflows, using an automotive service advisor to explain conversation context, execution state, persistent knowledge, and memory governance.
Production capability
Enterprise AI
Enterprise AI succeeds when model behavior, system architecture, governance, and business workflow all reinforce each other.
PoC-to-production architecturePlatform strategy, LLMOps, and engineering enablementBusiness workflow alignment and measurable outcomes
Related articles
Research notes in this topic
From Automotive Diagnostics to Banking Complaints: What I Learned Building a Local LLM Evaluation Platform
A second domain experiment showing why workflow behavior, retrieval shape, and reusable evaluation architecture matter more than simply adding more context.
From Baseline to RAG to Fine-Tuning: What Actually Works for Automotive AI
A practical experiment comparing baseline, RAG, fine-tuning, and combined approaches for automotive warranty and service diagnostics.