Mobileye deployed an AI support agent on Amazon Bedrock AgentCore that automated 66 percent of routine engineering tickets, cutting response times by 90 percent and hitting a 98 percent accuracy rate. The deployment removes skilled engineers from manual data lookups while maintaining strict security controls across on-premises and cloud environments.
The support bottleneck
As Mobileye's drive-data pipeline scaled, two-thirds of its internal support tickets became routine status inquiries. Engineers spent roughly 15 clicks navigating multiple backend systems just to check processing logs or validate outputs. Traditional scripting and rule-based workflows could not handle the variability of these requests. The team needed a solution that understood context without requiring constant manual intervention.
From proof of concept to production
Mobileye tested an intelligent agent using Anthropic Claude models routed through an internal gateway. A Model Context Protocol server gave the agent direct access to live production APIs, allowing it to query session statuses and retrieve diagnostic logs during inference. The prototype hit a 95 percent accuracy target and resolved inquiries in under two minutes.
After validating the approach, the team moved to Amazon Bedrock AgentCore for full deployment. They selected the platform for its serverless runtime, built-in observability, and ability to bridge their on-premises ticketing system with cloud services. The results exceeded initial targets, automating 66 percent of total ticket volume and reducing average response times to roughly one minute.
"Amazon Bedrock AgentCore enabled us to build a production-grade AI agent that delivers measurable business value while maintaining the security and operational standards our enterprise requires," said Shay Margalit, principal engineer at Mobileye. "The AI Support Agent has transformed our support operations, saving our engineers hundreds of hours previously spent on routine ticket lookups, and freed them to focus on what they do best."
Bridging on-premises systems and cloud infrastructure
An on-premises orchestrator extracts new tickets and forwards them to the cloud-based agent. The agent processes each request end-to-end, categorizing the inquiry, pulling real-time session data through the protocol server, and generating a formatted response with actionable steps.
The orchestrator posts the completed answer back into the local ticketing system. Engineering teams monitor every step using built-in observability tools that track latency, token usage, and decision traces. Engineers manage secure credentials through AWS Secrets Manager, which prevents direct handling of authentication keys.
Scaling across the organization
The initial success prompted Mobileye to build an internal deployment platform. Most developers lacked the AWS credentials or infrastructure access to run agents independently. The cloud infrastructure team created a managed service that provisions required resources automatically. Teams submit their agent code and specify capabilities like memory storage or browser tools.
The platform generates a configuration file and handles IAM roles, storage buckets, and monitoring alerts. Developers deploy the solution with a single command. This shift turned a single proof-of-concept into an enterprise-wide capability where teams launch secure, cost-tracked agents in minutes rather than weeks.
Why this matters for operations teams
Operations leaders should recognize that agentic automation does not require rebuilding legacy infrastructure. Mobileye kept its existing ticketing system intact while routing queries through a serverless cloud layer. The protocol server handled real-time data retrieval without custom middleware. Teams can replicate this pattern by isolating routine, high-volume inquiries and attaching them to governed AI runtimes. This approach reduces backlog, standardizes response quality, and keeps sensitive on-premises systems offline from direct internet exposure.
Your membership also unlocks:
