Bayer and EY help farmers use AI to unearth critical data with Azure OpenAI Service

Driven by its mission, “Health for all, hunger for none,” Bayer, a multinational agriculture pharmaceutical technology company, is working to solve complex issues related to healthy food production and food security. “A constraint limiting food supply throughout the world is a shortage of access to agronomic and crop input product knowledge,” says Daniel Kurdys, VP and Global Business Lead, GENAI at Bayer. “Critical information is often buried in bulky reference materials, and we want to change that.”

Tami Craig Schilling, VP and Agronomic Digital Innovation Lead at Bayer, agrees: “Every farmer, regardless of location, resources, or language, should have access to agronomic expertise. To make agronomy data more accessible, we needed a consolidated solution that could provide users with immediate guidance on everything from soil health to Bayer products for specific crops. We knew generative AI had a strong role to play.”

Conversational AI and crop science

Bayer Crop Science (BCS), a division of Bayer, is committed to providing tools and information that support farm operations into the future. The group wanted to create a proof-of-concept (POC) conversational AI assistant to provide easy access to its wealth of current agronomy knowledge from field work with farmers and in-house experts, along with past data from legacy systems.

With the help of Microsoft and Ernst & Young (EY) teams, long-standing Bayer collaborators and technology providers, BCS chose to build its POC on Microsoft Azure AI Foundry and use Azure OpenAI Service and Azure AI Search with the latest GPT-series models. “This is where Microsoft and EY really stepped up for us with generative AI,” says Kurdys. “We formed a trilateral partnership with a mission to build a prototype in 90 days—a very short timeline.” 

“We worked with the natural language capabilities of GPT-series language models in the trusted Azure environment, knowing all of our data was safeguarded,” continues Kurdys. “We wanted the generative AI system to comprehend the context of inquiries and respond to specific questions with accuracy and precision—empowering users to rapidly troubleshoot complex questions about things like a seed product’s performance in varying weather conditions.”

In addition to Azure OpenAI Service, BCS used Azure Data Manager for Agriculture and Microsoft Intelligent Data Platform with industry-specific data connectors and capabilities to pull farm data from disparate sources, including satellites, soil sensors, drones, and weather providers. “AI helps us fully utilize an array of high-quality datasets, internal and external, to support prescriptive insights,” explains Schilling. This helps BCS fine-tune guidance for farmers based on specific crops and geographic regions that can potentially increase yields and profits.

BCS relied on Azure Databricks for data processing and analytics used to train machine learning models for specific agricultural applications. Kurdys says Azure Data Lake also played a crucial role by providing a scalable and highly secure environment for storing and managing large volumes of data. “Azure Data Lake helps us streamline hundreds of years of agronomy knowledge housed in disparate data sources into a unified interface, which is essential for delivering actionable insights to farmers and other stakeholders,” says Kurdys. With support from EY and Microsoft teams, the BCS team implemented a retrieval-augmented generation (RAG) solution to dynamically source relevant information and respond in real time for rapid problem-solving. 

Together, the team also developed a sophisticated scoring system to evaluate the BCS app’s performance against open-source GPTs currently serving the agriculture market. The system showed that the app generated expert information within seconds across certain subjects. Prompt engineering was employed to correct inaccurate responses and speed times to align with BCS objectives.

Unlocking food productivity

BCS built a prototype in 90 days and shared it with 1,000 employees in 2024 to high acclaim. The solution will be available to several thousand BCS employees in 2025, and Schilling is confident it has a promising future. “Grounded in internal and trusted external data sources and in collaboration with Microsoft and generative AI, we’re creating a scalable foundation of expertise in applied agronomy across crop species and product categories that can support millions of farmers globally,” says Schilling.

“The possibilities are exciting,” adds Kurdys. “We think this is the next wave of unlocking food productivity for more people and better feeding communities across the world.”

Discover more about Bayer Global on Facebook, Instagram, LinkedIn, X/Twitter, and YouTube.