Bayer Crop Science runs 22 read-only agents over its ML decision engine on AWS

Monday, 26 October 2026By Sridhar Mukkandi1 min read

AWS described on 23 October how Bayer Crop Science built an Agentic Decision Engine: a multi-agent, natural-language interface over its Connected Decision Engine, which manages more than 500 model configurations. A supervisor agent routes each question to one of 22 specialist agents, running on Amazon Bedrock AgentCore with LangGraph, with about 20 API tools exposed through AgentCore Gateway and scoped per agent.

Agents are read-only in production, each with an IAM role limited to its function, and the system launched to a gated group first. AWS reports about 80% of user questions resolved without an engineer in the first phase, troubleshooting cut from over an hour to under 10 minutes, and input-token cost down 20 to 30% from caching a stable prompt prefix. The team also kept its current model after a newer reasoning model broke tool use and structured output.

So what

Read-only agents with narrowly scoped tools, launched to a small group first, is a pattern worth copying. So is testing a new model against your own tool calls before you switch: newer isn't always compatible.

SourceAWS Machine Learning Blogaws.amazon.com/blogs/machine-learning/how-bayer-crop-science-bui Go deeper · EssayRead-only first: how much access should an agent get?

Drafted with AI from the original source, then checked and edited by Sridhar Mukkandi. Spotted a mistake? Write to hello@tarkika.com and we'll correct it here.

More from The Signal

Research Microsoft’s ThinkingBox turns database checks into rewards for training agentsChecking the state an agent leaves behind is a more honest test than reading its transcript, and the same check doubles as a training reward. Start with the three actions your agent takes most often. Microsoft and Hugging Face Tools AWS lets you write your own model-training loop on SageMaker without managing GPUsCustom reinforcement learning used to mean renting and babysitting GPUs. If you can write a check for a good outcome, you can now train on it without running the infrastructure yourself. AWS Machine Learning Blog Agents Anthropic donates Agent Skills to the Linux Foundation’s Agentic AI FoundationNothing changes for builders today: skills you write keep working. Neutral governance makes the format a safer long-term bet, and the planned security guidance for skill authors and installers is worth following before you install third-party skills. AnthropicAll stories