What does ADE actually do for Bayer's users?

ADE is a natural language interface over Bayer's Connected Decision Engine. Breeders, analysts, and researchers ask questions about workflow status, results, model behavior, and failures, and a supervisor agent routes each request to a specialist agent.

How many users have tried ADE since launch?

Bayer launched ADE to a gated group rather than the whole organization. Since launch, more than 100 users across Bayer Crop Science have used it, including breeders, analysts, researchers, data scientists, and engineers.

What concrete efficiency gains did Bayer report?

Exposing APIs through AgentCore Gateway rather than a dedicated server per API cut per-API integration overhead by more than 99 percent. Caching the stable prompt prefix cut input-token cost by 20–30 percent.

Large Language ModelsAI Business & IndustryAmazon AI Blog14m agoOct 24, 2026, 04:00 JST

Bayer builds ADE on Amazon Bedrock AgentCore with 22 agents

Send on LINE

3 Key Points

What happened
Bayer Crop Science built the Agentic Decision Engine (ADE), a multi-agent natural language interface over its Connected Decision Engine, running 22 specialist agents on Amazon Bedrock AgentCore; in its first phase ADE resolves about 80 percent of user questions without an engineer in the loop.
Why it matters
That resolution rate suggests a production multi-agent deployment can move routine operational questions away from engineers, while Bayer's five design decisions show the work sat in durable state, identity boundaries, and tool scoping rather than in prompts.

WHO IT HITSThis matters for enterprise data and platform teams who field internal questions about analytics workflows, and for AI engineers deciding whether a managed agent runtime can replace self-built agent hosting, session isolation, and tool access layers.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, every morning.

Get these free every morning

Context & Analysis

Bayer's Connected Decision Engine already handled the governed execution of ML workflows across genomics, field trials, imagery, and environmental data. The shift came as adoption grew: explaining a result meant piecing together context from APIs, documentation, metadata stores, dashboards, and logs, so the bottleneck moved from running workflows to understanding them. ADE was built as a natural language layer over CDE, not a replacement, and CDE remains the governed execution layer for analytics.

The production work described sits mostly outside the prompts. Bayer had to externalize LangGraph checkpoint state to DynamoDB so workflows could resume across scale-out events and session boundaries, define identity mechanisms at each trust boundary rather than reuse one token, and scope tools by specialist agent because tool schemas themselves consume input tokens. A later model version was evaluated but not adopted for the release, because tool use and structured output were not compatible with how the agents were implemented.

The AWS Unified Operations engagement supplied a designated Technical Account Manager and a generative AI Domain Specialist Engineer who worked as part of the ADE team, and Bayer tracked 35 blockers to resolution before launch. Bayer is now exploring hybrid cloud agent interoperability, and plans to evaluate Amazon Bedrock Guardrails as an additional layer on its existing custom safety checks.

FAQ

What does ADE actually do for Bayer's users?: ADE is a natural language interface over Bayer's Connected Decision Engine. Breeders, analysts, and researchers ask questions about workflow status, results, model behavior, and failures, and a supervisor agent routes each request to a specialist agent.
How many users have tried ADE since launch?: Bayer launched ADE to a gated group rather than the whole organization. Since launch, more than 100 users across Bayer Crop Science have used it, including breeders, analysts, researchers, data scientists, and engineers.
What concrete efficiency gains did Bayer report?: Exposing APIs through AgentCore Gateway rather than a dedicated server per API cut per-API integration overhead by more than 99 percent. Caching the stable prompt prefix cut input-token cost by 20–30 percent.

Amazon AI BlogRead Original Article

Send on LINE

AI news that matters for your work, delivered every morning.

Pick your industry and the AI tools you use, and get news related to your work every day.

Get it free

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.