Walmart's team built an LLM-powered operations assistant using Model Context Protocol (MCP) to help non-technical L2 support staff triage incidents across 5,000+ stores. By connecting GPT-4o and Claude to BigQuery, Prometheus, Splunk, and edge systems via MCP, support staff can ask plain-language questions like 'What's wrong with Store 1182?' and receive instant, actionable answers. This cut triage time from 15+ minutes to under two minutes — an 85% reduction — without adding headcount. The system was itself built using AI code generation tools like GitHub Copilot, compressing weeks of development into days. Next steps include a production UI, automated remediation actions, and proactive monitoring capabilities.
Table of contents
Your inbox, upgraded.Scaling support without scaling peopleBuilding an LLM-powered operations assistantMore like thisBuilding with AI, not just using itMeasurable impact of MCPWhat’s next?Key takeaways for engineering leadersSort: