How Bayer Crop Science built an agentic decision engine on Amazon Bedrock AgentCore with AWS Unified Operations
Bayer Crop Science built the Agentic Decision Engine (ADE), a multi-agent platform on Amazon Bedrock AgentCore that lets breeders, analysts, and researchers ask natural language questions about their analytics platform. This post covers the ADE architecture, five design decisions, and how the team prepared it for production with AWS Unified Operations.
First seen by Oriel on .
AI summary
Bayer built the Agentic Decision Engine (ADE), a multi-agent natural language interface over its governed analytics platform, resolving about 80 percent of user questions without engineering help in its first phase.
- ADE runs on Amazon Bedrock AgentCore with LangGraph orchestration, 22 specialist agents, and AgentCore Gateway tools connected to CDE APIs and data sources.
- A troubleshooting workflow combines CloudWatch telemetry with live CDE status to diagnose failures in under 10 minutes, down from over an hour of manual investigation.
- The team addressed durable workflow state, identity across trust boundaries, model compatibility, scoped tool catalogs, and promoting complex capabilities into specialist agents.
- CloudWatch task-level metrics, request-level cost attribution, least-privilege read-only access, and production reviews helped prepare the service for launch; prompt caching reduced input-token cost by 20–30 percent.
- More than 100 Bayer Crop Science users have used ADE, and exposing APIs through AgentCore Gateway reduced per-API integration overhead by more than 99 percent.
#llm agents #amazon bedrock #langgraph #machine learning
Generated by gpt-5.6-sol. Check the original for details.
Aliases
| Source | External ID | URL |
|---|---|---|
| AWS Machine Learning Blog | b1f9be62942b6b77e09ee44eda01d25e5a58cfc4 | https://aws.amazon.com/blogs/machine-learning/how-bayer-crop-science-built-an-agentic-decision-engine-on-amazon-bedrock-agentcore-with-aws-unified-operations/ Read discussion |