AWS launches AI equipment repair assistant using Bedrock AgentCore

Amazon announced an AI-powered equipment repair assistant built on Amazon Bedrock AgentCore, designed to help farmers and field technicians quickly diagnose equipment problems, identify required parts, and access manufacturer-approved repair procedures. The solution uses Amazon Nova 2 Lite as the foundation model, Bedrock Knowledge Base with retrieval-augmented generation (RAG) for indexing manuals and parts catalogs, and AgentCore Memory for persistent conversation history. Technicians interact via a React frontend hosted on AWS Amplify, authenticated through Amazon Cognito, and all requests go to the AgentCore Runtime endpoint. The Strands Agents SDK powers the agent, which uses a custom tool to query the knowledge base via the retrieve_and_generate API. OpenSearch Serverless vector search and Titan Embeddings handle semantic matching, while DynamoDB stores service tickets. CloudWatch and X-Ray provide observability.

The key innovation is how the agent seamlessly combines short-term and long-term memory: short-term for within-session context, long-term for cross-session facts, so technicians can ask follow-ups without repeating information. The system also provides source citations for every recommendation, enabling verification. This addresses a major pain point in heavy machinery repair — technicians often make multiple site visits because they lack the right parts or procedures, leading to extended downtime and financial losses, especially during harvest. By centralizing documentation and enabling natural-language queries, the assistant reduces diagnosis time, ensures parts are ordered correctly the first time, and gives access to manufacturer-approved procedures. Amazon's choice of Bedrock AgentCore (an emerging serverless agent platform) signals a shift toward more integrated, memory-aware AI assistants that can be deployed quickly with existing enterprise knowledge bases.