AWS re:Invent 2025 - Autodesk Assistant: Building an Agentic Platform on Bedrock (ISV322)

Summary of AWS re:Invent 2025 - Autodesk Assistant: Building an Agentic Platform on Bedrock (ISV322)

Introduction to Autodesk and Industry Challenges

• Autodesk is a global leader in 3D design, engineering, and entertainment software
• Autodesk's customers, including architects, engineers, designers, and creators, face unprecedented disruption and demand
• Manufacturers struggle with supply chain delays and labor shortages, while needing to constantly push out new products
• Design and construction industries face pressure to deliver more complex projects faster, with increased sustainability demands
• Production studios struggle to keep up with consumer demand and the need for fresh content

The Rise of Agentic AI

• Agentic AI represents a fundamental shift in how AI systems function in the design and manufacturing space
• It enables a "describe and do" model, where user intent becomes the direct action, rather than a point-and-click approach
• Agentic AI unlocks new levels of creative velocity, allowing customers to go from idea to impact in moments, not weeks
• It redefines what's possible, shifting from manual tasks to highly connected, automated, and accessible workflows

Introducing Autodesk Assistant

• Autodesk Assistant is an Agent AI partner that speaks the language of design and manufacturing
• It can automate manual tasks, connect disconnected workflows, and deliver real-time insights in collaboration with human users
• The demo showcases Autodesk Assistant's capabilities in a Revit building information modeling (BIM) software context:
  • Answering questions about the building model
  • Modifying window designs through natural language commands
  • Automating the creation of floor plans, view templates, and cover sheets

Autodesk Assistant Architecture

• The architecture is built on a foundation of MCP (Model Composition Platform) servers, which bridge the gap between language models and Autodesk's product APIs
• Autodesk has a variety of agents, including product-specific and general-purpose agents, that provide the core capabilities of Autodesk Assistant
• The Assistant Backend acts as a shared platform service, handling guardrails, context management, observability, and other cross-cutting concerns
• The Conversation Proxy Router is responsible for routing requests to the appropriate domain agent, using a combination of language model-based and rule-based approaches
• The Assistant UI provides a consistent user experience, passing user prompts and relevant context to the backend for processing

Leveraging AWS Bedrock and Agent Core

• Autodesk recognized the need for a secure, scalable, and governed runtime environment for their agents and MCP servers
• They leveraged AWS Bedrock and Agent Core services to address key challenges:
  • Secure compute runtime for agents and MCP servers
  • Centralized context management and identity/access governance
  • Seamless integration and discovery of custom MCP servers and cross-domain agents
  • Auditability and observability of the entire system
• Agent Core Runtime, Agent Core Gateway, and Agent Core Memory were key components used by Autodesk to build their platform

Strands Agent Framework

• Autodesk adopted the Strands agent framework as a common foundation for building agents across the organization
• Strands provides native integration with Agent Core, supports various open-source protocols, and accelerates the development cycle for building agents

Key Takeaways

• Autodesk recognized the need for a scalable, secure, and governed runtime environment for their agentic AI platform
• They leveraged AWS Bedrock and Agent Core services to address key challenges and build a unified platform for their industry teams
• The Strands agent framework was adopted as a common foundation for building agents, promoting consistency and accelerating development
• Autodesk's experience showcases the importance of a platform-centric approach when building complex, enterprise-grade agentic AI systems

Call to Action

• Explore the AWS Agent Core services and samples to learn more about building scalable and secure agent-based applications
• Check out the Autodesk Assistant in action to see the capabilities of agentic AI in design and manufacturing workflows
• Consider the Autodesk Agent Omni Mesh, an open-source framework for multi-agent collaboration, as a starting point for your own agentic AI projects