AI Practices AI实践 8h ago Updated 2h ago 更新于 2小时前 43

How Mobileye transformed support operations using Amazon Bedrock AgentCore Mobileye如何利用Amazon Bedrock AgentCore转型支持运营

Mobileye deployed an AI Support Agent on Amazon Bedrock AgentCore to automate routine internal ticket status inquiries, cutting response times by 90% and exceeding 95% accuracy targets The solution leverages the Model Context Protocol (MCP) to give the agent real-time access to on-premises drive-data processing platform APIs, enabling it to act as an intelligent investigator rather than a simple classifier A hybrid architecture bridges Mobileye's on-premises ticketing system with AWS cloud compo Mobileye利用Amazon Bedrock AgentCore部署AI支持Agent,将工单响应时间缩短90%,准确率超过95%,且无需基础设施管理 采用混合架构设计:本地组件处理工单提取与响应回写,AWS云端提供无服务器AI推理,解决内部系统无法从AWS访问的难题 Model Context Protocol (MCP)是关键使能技术,赋予Agent实时访问数据处理平台API的能力,实现从简单分类到智能调查的跃升 基于Anthropic Claude模型,通过内部LLM Gateway进行治理和配额管理,确保企业级安全合规 成功案例推动AgentCore转型为全公司自助服务平台,支持各

58
Hot 热度
65
Quality 质量
60
Impact 影响力

Analysis 深度分析

TL;DR

  • Mobileye deployed an AI Support Agent on Amazon Bedrock AgentCore to automate routine internal ticket status inquiries, cutting response times by 90% and exceeding 95% accuracy targets
  • The solution leverages the Model Context Protocol (MCP) to give the agent real-time access to on-premises drive-data processing platform APIs, enabling it to act as an intelligent investigator rather than a simple classifier
  • A hybrid architecture bridges Mobileye's on-premises ticketing system with AWS cloud components, using a Local Orchestrator to handle ticket extraction and response posting while AgentCore's serverless runtime manages agent execution
  • Mobileye's internal LLM Gateway provides governed, quota-managed access to Anthropic Claude foundation models on Amazon Bedrock, ensuring enterprise-grade security and compliance
  • The success of the proof of concept led Mobileye to transform AgentCore into a self-service platform, enabling other teams across the company to deploy their own AI agents

Why It Matters

This case study demonstrates a practical blueprint for enterprises seeking to scale AI agents while maintaining strict governance, security, and hybrid infrastructure requirements—common pain points for organizations with legacy on-premises systems. The integration of MCP for real-time API access during inference represents a significant advancement over static, rule-based automation, showing how agentic AI can handle complex, context-dependent workflows that previously required human intervention.

Technical Details

  • Model & Governance: The agent uses Anthropic Claude foundation models accessed through Mobileye's internal LLM Gateway, which provides governed, quota-managed access to models on Amazon Bedrock, ensuring enterprise compliance and controlled usage.
  • Model Context Protocol (MCP): MCP serves as the critical enabler, granting the agent real-time access to the drive-data processing platform's APIs during inference. This allows the agent to query session status, retrieve processing logs, and pull diagnostic information dynamically, transforming it from a ticket classifier into an intelligent investigator.
  • Hybrid Architecture: On-premises components include a Local Orchestrator (handles ticket extraction and response posting) and the internal ticketing system. AWS cloud components include AgentCore Runtime (serverless agent execution with automatic scaling), eliminating infrastructure management overhead.
  • Performance Metrics: The proof of concept targeted 95% accuracy in ticket classification with sub-2-minute response times; production deployment exceeded the 95% accuracy target and achieved a 90% reduction in response times.
  • AgentCore Platform: A fully managed platform for building, connecting, and optimizing AI agents at scale, supporting multiple agentic frameworks, comprehensive built-in observability, and serverless infrastructure that auto-scales to handle support volume spikes.

Industry Insight

  • Enterprises with on-premises legacy systems can adopt a hybrid agent architecture that keeps sensitive data and ticketing infrastructure local while leveraging cloud-based AI compute, reducing the security and compliance risks associated with full cloud migration.
  • The Model Context Protocol (MCP) is emerging as a key standard for enabling AI agents to interact with real-time external systems during inference, moving beyond static training data to dynamic, context-aware decision-making—a capability essential for production-grade support automation.
  • Mobileye's pivot from a single-use agent to a self-service platform illustrates a scalable governance model: once enterprise-grade agent infrastructure is proven, it can be productized internally, accelerating AI adoption across teams while maintaining centralized control over model access, observability, and security policies.

TL;DR

  • Mobileye利用Amazon Bedrock AgentCore部署AI支持Agent,将工单响应时间缩短90%,准确率超过95%,且无需基础设施管理
  • 采用混合架构设计:本地组件处理工单提取与响应回写,AWS云端提供无服务器AI推理,解决内部系统无法从AWS访问的难题
  • Model Context Protocol (MCP)是关键使能技术,赋予Agent实时访问数据处理平台API的能力,实现从简单分类到智能调查的跃升
  • 基于Anthropic Claude模型,通过内部LLM Gateway进行治理和配额管理,确保企业级安全合规
  • 成功案例推动AgentCore转型为全公司自助服务平台,支持各团队自主部署AI Agent

为什么值得看

本文为企业级AI Agent落地提供了完整的实践范本,展示了如何平衡企业治理要求与技术创新灵活性。其混合架构方案对拥有遗留系统的传统企业具有直接参考价值,证明了AI Agent可在不牺牲安全合规的前提下实现规模化部署。

技术解析

  • 混合架构设计:本地部署Local Orchestrator负责从内部工单系统提取新工单并将AI响应回写,AWS云端运行AgentCore Runtime提供无服务器计算。这种设计解决了内部系统无法从AWS直接访问的安全约束,同时享受云端的弹性扩展能力。

  • Model Context Protocol (MCP):作为关键使能技术,MCP赋予Agent实时访问数据处理平台API的能力,使其能够查询会话状态、检索处理日志、获取诊断信息。这使Agent超越了简单的工单分类,成为能够主动调查问题的智能体。

  • LLM治理架构:通过内部LLM Gateway集中管理对Amazon Bedrock上基础模型的访问,实现治理策略和配额控制。Agent使用Anthropic Claude模型,确保企业级安全合规的同时获得强大的推理能力。

  • 无服务器Agent运行时:AgentCore提供完全托管的部署环境,团队无需配置服务器即可部署和迭代Agent,自动扩展应对支持量峰值,支持多种Agent框架,通过单一API调用即可集成现有系统。

行业启示

  • 混合架构将成为企业AI落地的主流范式:对于拥有大量遗留系统的企业,完全云原生方案往往不现实。Mobileye的本地-云端混合架构证明,通过合理的组件分工,可以在不改造现有系统的前提下实现AI能力集成。

  • 从POC到生产的治理是关键分水岭:许多企业AI项目止步于概念验证,核心障碍在于缺乏企业级治理框架。Mobileye通过LLM Gateway实现集中治理,同时保持Agent的灵活性,为规模化部署提供了可复制的路径。

  • 自助服务平台是AI Agent规模化部署的必然选择:单一团队的成功案例往往难以持续,Mobileye将AgentCore转型为全公司自助平台,使各团队能够自主部署AI Agent,这种模式将显著加速企业内部的AI能力普及。

Disclaimer: The above content is generated by AI and is for reference only. 免责声明:以上内容由 AI 生成,仅供参考。

Agent Agent Autonomous Driving 自动驾驶 Deployment 部署 LLM 大模型