Ecolab Builds Multi-Agent Retail Intelligence Platform on Databricks and Anthropic Claude, Cutting Compliance Report Time from Two Weeks to Two Minutes

Title

Ecolab Builds Multi-Agent Retail Intelligence Platform on Databricks and Anthropic Claude, Cutting Compliance Report Time from Two Weeks to Two Minutes

Content

Ecolab, the global leader in water, hygiene, and infection prevention, deployed a multi-agent Retail Intelligence platform on Databricks with Anthropic Claude, cutting per-location compliance report generation time from two weeks to under two minutes while serving approximately 600 frontline staff across North American retail and food service locations. Built as a native Databricks App with Lakebase Postgres, the system unifies nine previously siloed data sources — audits, health inspections, pest IoT telemetry, checklists, chemical usage logs, weather feeds, Yelp reviews, CDC neighborhood data, and the FDA food code — into a single governed lakehouse under Unity Catalog. Launched in mid-April 2026, the conversational agent returns cited, plain-language answers to FDA food code questions in seconds, and supports approximately twelve languages at roughly 98% accuracy. "We had nine different data sources, nine different intelligences, and no way to see the full picture for a single location" — Nicholas Dylla, Technical Lead at Ecolab. "What used to take two weeks, pulling data from nine systems to compile a single compliance report, now takes under two minutes with Claude on Databricks. Our frontline staff across 600 locations get cited, plain-language answers from a 700-page FDA food code in seconds" — Josh McCoy, Product Manager for Retail Intelligence, Ecolab. The deployment addresses a fundamental fragmentation problem: Ecolab monitors food safety, pest control, and water quality for thousands of retail and fast-food locations, but the underlying data lived across nine systems with no unified view per location. The platform surfaces cross-source insights — for example, a pest issue that is also a food safety violation is now flagged, investigated, and resolved once, rather than discovered twice in separate workflows. Architecture follows a multi-agent supervisor pattern orchestrated through Databricks Workflows. A Coordinator Agent breaks each user question into subtasks and delegates to specialized sub-agents: one retrieves relevant FDA passages via Vector Search, another queries structured compliance data through SQL and Unity Catalog Functions, and a third pulls pest telemetry from an external MCP server. Claude Sonnet handles complex reasoning, Claude Haiku handles fast summarization every three turns, and Gemini handles image analysis — all served through Databricks Foundation Model APIs inside the Databricks security perimeter. Five Judge LLMs evaluate every interaction, MLflow traces every execution path, and Databricks AI batch inference functions like ai_query() apply Claude across thousands of records for high-volume offline workloads such as retroactively scoring historical inspections. A dual-layer memory architecture makes the experience feel personal: short-term working memory carries the last ten conversation turns directly in the prompt, while long-term semantic memory is maintained by Claude Sonnet 4.6 as a per-user profile of role, preferences, recurring focus areas, and location context. Next, Ecolab plans to add MCP-powered automated actions — pest inspections, chemical reorders, food safety norms, and work orders triggered directly from the chat interface — turning the system from an intelligence layer into a full operational agent.