Autodesk University 2026 demonstrated two directions for bringing AI into engineering workflows. Autodesk Assistant Builder allows companies and partners to connect specialized knowledge and data to Autodesk Assistant, while Fusion MCP lets external AI systems access engineering capabilities. Beyond PLM argues that these developments raise a larger issue: connecting tools is not the same as maintaining a shared understanding of a product.
Model Context Protocol, or MCP, provides a standardized mechanism for AI applications to access tools and resources. But engineering decisions often involve CAD, PLM, ERP, supplier services, and other systems that describe the same product differently. A component might have one identifier in CAD and another at a supplier, while inventory records could reference an earlier revision. Successful tool connections alone cannot resolve these differences.
The article argues that AI workflows therefore need relationships, identity mappings, and explicit rules defining which source governs each piece of information. This becomes increasingly important as agents move from retrieving information to taking actions. Human engineers may recognize inconsistencies through experience, but AI agents need those distinctions represented in accessible data.
Beyond PLM suggests evaluating openness through complete engineering decisions rather than simply counting available connectors. A product-change workflow should identify the correct revision, examine supporting evidence, request appropriate approval, preserve the decision, and expose conflicts among systems instead of hiding them.
Autodesk AI Orchestrator adds another layer by selecting AI models suited to particular tasks. Yet choosing the right model does not solve the underlying data problem. Reliable decisions still require trustworthy product information and context.
The article ultimately emphasizes product memory. AI assistants and models will change, while engineering decisions may remain relevant for years. Companies therefore need to preserve the evidence, reasoning, responsibilities, source authority, and lifecycle records behind decisions. MCP can connect AI to engineering tools, but dependable AI workflows will depend on maintaining this context across applications, organizations, and changing assistants.
