Mobileye Cuts Support Response Times by 90% with Amazon Bedrock AgentCore

Mobileye, a pioneer in autonomous driving with over 230 million EyeQ system-on-chips deployed, faced a support bottleneck as 66% of tickets were routine status inquiries requiring manual steps across multiple systems. To address this, they built an AI Support Agent using Amazon Bedrock AgentCore, which leverages Anthropic Claude foundation models accessed through Mobileye's internal LLM Gateway. The agent uses Model Context Protocol (MCP) to query live production data during inference. After a successful proof of concept, Mobileye deployed the agent in production with a hybrid architecture bridging on-premises ticketing and AWS cloud services. The results exceeded targets: 98% success rate (above 95% target), response time reduced from hours to about 1 minute (90% improvement), and 100+ tickets processed monthly. The Cloud Infra team then built an internal platform enabling developers across the company to deploy their own AI agents on AgentCore with pre-configured infrastructure, security, and observability.