AWS re:Invent 2025 -How Toyota Built an AI Platform that Revolutionizes the Dealer Experience-IND320

AWS re:Invent 2025 - How Toyota Built an AI Platform that Revolutionizes the Dealer Experience

Toyota's Enterprise AI Approach

• Toyota formed an "Enterprise AI Team" focused on building AI accelerators and "AI teammates" to augment their existing workforce
• The team is primarily composed of engineers who focus on:
  • Exploring novel AI use cases
  • Experimenting and educating business teams on AI capabilities
  • Enabling the adoption of AI across Toyota's business units
• Toyota takes a "build, configure, buy" approach to developing AI solutions:
  • Build custom capabilities from scratch when needed
  • Configure off-the-shelf products to fit their requirements
  • Buy SaaS solutions when available to avoid reinventing the wheel

Enhancing the Dealer Experience with Generative AI

• Toyota identified an issue where customers would research vehicles extensively online, but dealers lacked the same level of product knowledge
• To address this, Toyota built a platform that:
  • Ingests and summarizes all Toyota vehicle data (specs, pricing, features, etc.) using generative AI
  • Provides dealers and customers with accurate, up-to-date product information
  • Includes contextual legal disclaimers to ensure compliance
• This system is currently deployed to over 7,000 Toyota dealers, handling 7,000+ interactions per month

Technical Architecture

Version 1 (Current)

• Front-end makes requests to an Enterprise AI account
• Intent router identifies the vehicle being queried and routes the request
• Prompt Guard checks for malicious activity
• Semantic search is performed against a vector database to retrieve relevant documents
• Bedrock is used to generate responses based on the retrieved documents and business requirements
• Responses are post-processed to include required disclaimers and images
• Logs are exported to MongoDB for compliance reporting

Version 2 (Future)

• Transitioning to an agent-based, agentic platform using Amazon Bedrock and Agent Core
• Orchestrator will manage agent routing based on an agent registry
• Agents (e.g., Product Expert, Product Support) will be deployed in Agent Core runtime
• Agent Core services will be leveraged for caching, identity management, observability, and more
• Aim to reduce infrastructure overhead and improve scalability, flexibility, and responsiveness

Key Takeaways

• Toyota is at the forefront of enterprise-scale adoption of generative AI and agentic platforms
• Their approach focuses on building reusable AI accelerators and "AI teammates" to augment their workforce
• The dealer experience platform demonstrates how generative AI can be used to provide accurate, up-to-date product information to customers and dealers
• Toyota is transitioning to a more flexible, agent-based architecture using Amazon Bedrock and Agent Core to improve scalability and responsiveness

Technical Details

• Technologies used: Amazon Bedrock, Agent Core, Strands, Dynamo DB, OpenSearch, MongoDB, AWS Lambda, ECS, EKS
• Metrics: 7,000+ dealer interactions per month with the current system
• Key features: Semantic search, natural language summarization, compliance-aware response generation, caching, and observability

Business Impact

• Enhances the dealer experience by providing them with comprehensive, up-to-date product information
• Enables faster access to information for both dealers and customers, improving the sales process
• Allows Toyota to quickly adapt to changes in product data and customer needs
• Lays the foundation for further AI-powered enhancements to the dealer and customer experience

Examples

• Automating the analysis of 300,000 contracts, reducing the time from 30,000 per year to near real-time
• Deploying the dealer experience platform to over 7,000 Toyota dealers, handling 7,000+ interactions per month