Bayer brings agronomic expertise to the edge with Azure AI Foundry | Microsoft Customer Stories

Bayer — the German life-science company — built E.L.Y. Crop Protection (E.L.Y. Mini), a small language model using Microsoft Phi and Azure AI Foundry, fine-tuned on proprietary label data, to give Bayer's agronomic advisors and partners precise, compliant answers to crop protection questions. Early users report 5-10% productivity gains, with complex questions resolved in under 30 seconds instead of days. Per the Microsoft Customer Stories article: Bayer's agronomic advisors faced growing complexity in delivering accurate crop protection guidance across diverse products and regulations, with frontline teams spending hours navigating lengthy labels to make critical decisions. The E.L.Y. Mini small language model (SLM), built on Microsoft Phi and deployed via Azure AI Foundry, addresses this by providing fast, label-grounded answers — fine-tuned on Bayer's proprietary label data to ensure compliance with product regulations. The motivation: agronomic advisors are responsible for making crop-protection recommendations across a wide range of products, regions, and regulatory regimes. With hundreds of label changes per year, advisors faced increasing difficulty staying current, and the latency of searching through lengthy product labels led to delays in critical decision-making. E.L.Y. Mini is built on Microsoft Phi — a family of small language models — and deployed on Azure AI Foundry. The model is fine-tuned on Bayer's proprietary label data, ensuring all answers are grounded in current product information. Through E.L.Y. Mini, Bayer provides precise, compliant answers to internal teams and partners. Operational scale: rolled out as Bayer's internal crop-protection Q&A tool for agronomic advisors and partners. Early users report 5-10% productivity gains; complex questions resolved in under 30 seconds vs days. Future direction: per Microsoft, E.L.Y. Mini "sets a new benchmark for responsible, scalable AI in agriculture" — Bayer plans continued iteration on the model and potentially expansion to other agricultural Q&A use cases.