AI for CAD Tools
SolidWorks, Onshape, Siemens NX, Teamcenter, and Creo all shipped built-in AI in 2026. Here is what each one actually does, and does not, reach.
·
⏱
8 min read
Michelle Ben-David
Michelle Ben-David is a mechanical engineer and Technion graduate. She served in an IDF elite technology and intelligence unit, where she developed multidisciplinary systems integrating mechanics, electronics, and advanced algorithms. Her engineering background spans robotics, medical devices, and automotive systems.
BOTTOM LINE
Every major CAD and PLM vendor now ships some form of built-in AI, and each one is genuinely useful for the problem it was built to solve. SolidWorks and CATIA's AURA reaches across the 3DEXPERIENCE platform. Onshape's AI Advisor and FeatureScript MCP Server help inside Onshape's own kernel. Siemens splits the job between NX Design Copilot and Teamcenter Copilot. PTC's Creo AI is still rolling out its three planned tiers. What none of them do is cross vendor lines: a team running SolidWorks, Teamcenter, and a shared network drive still has three separate pockets of knowledge, each with its own assistant. Leo sits above all of them as a single intelligence layer, so the answer to a question does not depend on which tool happens to be open.
Every major CAD and PLM vendor has now shipped some version of built-in AI. SolidWorks talks about AURA. Onshape has an AI Advisor and a new FeatureScript MCP Server. Siemens split its effort between NX and Teamcenter. PTC is rolling out Creo AI in three stages. For an engineering team comparing tools, it is a fair question whether any of this closes the gap that a separate assistant like Leo is built to fill. The honest answer takes a vendor-by-vendor look, because each of these tools was built to solve a narrower problem than it sounds like from the press release, and the pattern that connects all of them matters more than any single feature.
What SolidWorks and CATIA's AURA Actually Reaches
Dassault Systemes ships a set of virtual companions across its SolidWorks and 3DEXPERIENCE products, and AURA is the one most engineers have heard of. Dassault's own material describes AURA and its siblings as an AI-powered workforce for the 3DEXPERIENCE platform that can reason, plan, and act on industrial problems, with some features still in beta as of the 2026 release. The scope depends on which product you are in. Inside desktop SolidWorks, AURA's reach is still being defined as more of the software ties into the cloud-connected 3DEXPERIENCE platform. Inside 3DEXPERIENCE CATIA, the picture is clearer: Dassault describes the platform itself as a single source of truth for requirements, geometry, simulation, and analytics, and AURA is built to draw on that shared data model rather than one open file.
That makes AURA's reach the broadest of any tool in this comparison, but it is a reach bounded by one thing: the 3DEXPERIENCE platform itself. AURA does not describe any capability for reading a Teamcenter vault, a shared network drive, a resource-planning system, or a second CAD tool a team might also run. Dassault also ships generative features under the same 2026 release that can turn a sketch, a scanned part, or a written requirement into solid features, part geometry, or a first-pass assembly, again grounded in whatever already lives inside that one platform's data model. For a team fully committed to 3DEXPERIENCE end to end, that is a genuinely wide net. For the more common case of a shop running SolidWorks alongside a separate PDM system, a supplier's native files, or years of tribal knowledge sitting in email and shared folders, AURA has no stated path to any of it. A closer look at AURA's specific SolidWorks capabilities is in our full AURA review.
IN PRACTICE
It integrates directly with PLM and existing workflows, making past designs, standards, and calculations instantly available. The result is fewer errors, faster decision-making, and a more consistent process across teams.
- Sergey G., Board Member
Onshape's AI Advisor and the New FeatureScript MCP Server
Onshape's AI Advisor is the more modest of the two tools it shipped this year, and Onshape is direct about what it is: a conversational assistant grounded in Onshape's own library of training material, built to answer questions about how the platform works. Onshape's own feature page lists what AI Advisor explicitly does not do: it does not generate designs, does not predict errors in a model, does not make engineering decisions, and does not support generative design. It answers questions and points to documentation, embedded in the workspace so a new user does not have to leave the tool to look something up.
The more interesting release is the FeatureScript MCP Server, which PTC (Onshape's parent company) introduced to let an AI agent write, test, debug, and refine custom CAD features in Onshape's native scripting language, then share the finished tool across a team without repeating the same AI conversation for every engineer. That is a real automation surface, and it is a different kind of capability than a chatbot. But it is scoped to writing code inside the CAD kernel itself. Neither Onshape tool, as documented by Onshape and PTC, is built to search a company's document history across the workspace, read a separate PDM system, or reach resource-planning or standards content that lives outside Onshape entirely. Both tools make Onshape itself easier to use and extend. Neither makes the knowledge sitting outside Onshape any more reachable, which is the same boundary we found when we looked at what Onshape's own API hands an AI agent.
Siemens Split the Job in Two: NX Design Copilot and Teamcenter Copilot
Siemens is a useful case because it did not ship one built-in AI, it shipped two, for two different products solving two different problems. NX Design Copilot, announced as part of Siemens' 2025 to 2026 NX releases, is a natural-language help assistant. Siemens describes it as a way to accelerate learning by helping users find answers to technical queries, best practices, and documentation, aimed at both new and experienced NX users. It is a documentation and onboarding tool, not a tool described as reading the specifics of an open assembly or a project's history.
Teamcenter Copilot is a separate product built for a separate job. Siemens describes it as answering questions from a knowledge base, defined as a curated collection of documents already managed inside Teamcenter's PLM system: PDFs, Word and Excel files, scanned drawings, and structured data indexed out of NX and Solid Edge drawings and Teamcenter requirement specifications. Answers come with citations back to the source document, and Siemens is explicit that existing access permissions still apply, so a document a user cannot open does not appear in a Copilot answer either. That is a meaningfully more enterprise-grade scope than a documentation chatbot. It is also explicitly bounded by what one condition: the document has to already be managed inside Teamcenter. Siemens' own material does not describe Teamcenter Copilot reaching into resource-planning records, a standards library outside the PLM system, or a second PDM system a company might run for a different product line, a boundary we detail further in what the Teamcenter API actually gives engineering teams.
PTC's Creo AI Is Rolling Out in Three Stages, All Scoped to One Session
PTC has been the most explicit of any vendor covered here about sequencing its built-in AI. Its own newsroom material describes three tiers, sometimes called the "Triple A" framework: Advise, which is live in current Creo and Creo+ releases; Assist, targeted for a beta later in 2026; and Automate, still at an early alpha stage targeted before the end of 2026. Advise answers questions about the part or assembly currently open and about how to use Creo itself. The later tiers are meant to move from answering questions toward carrying out modeling steps, but neither has shipped broadly yet.
What PTC's own scoping language makes clear, across all three tiers, is the boundary: Creo AI is grounded in the model open in the current session plus Creo's own documentation. It is not described as reaching a PDM vault, a PLM system, a resource-planning record, or a standards reference outside Creo itself, whether that vault happens to be PTC's own Windchill or something else entirely. That is a sensible place to start, since a session-scoped assistant is far easier to build and validate than one that has to reconcile data across systems. It also means the gap is structural rather than a beta limitation that later tiers will simply close. Assist and Automate are described as doing more inside that same session, not as reaching further outside it. An intelligence layer that sits above the CAD tool, connecting to PDM, PLM, local and network directories, and resource-planning data the way Leo does, is solving a different problem than any of the three tiers, not a more advanced version of the same one. Where third-party tools fit around Creo today is covered in our roundup of AI tools for Creo.
The Pattern: Every Vendor's Built-In AI Stays Inside That Vendor's Own Walls
Line up all five and the shape is consistent. AURA's widest documented reach stops at the edge of the 3DEXPERIENCE platform. Onshape's AI Advisor stops at Onshape's own documentation, and its FeatureScript MCP Server stops at code written inside Onshape's kernel. NX Design Copilot stops at NX's own help content. Teamcenter Copilot stops at documents already managed inside Teamcenter. Creo AI, in all three of its planned tiers, stops at the model open in the current session plus Creo's own documentation. None of the five vendor tools covered here describe reaching into a second vendor's CAD system, a shared network drive, a resource-planning system, or a standards library that sits outside their own platform.
That is not a criticism of any single tool. Each one is a reasonable answer to the problem its own vendor set out to solve, and each genuinely helps inside that boundary. It is the reason a real engineering knowledge management gap remains even after every vendor in a company's stack ships its own assistant. Most engineering teams do not run one system. They run a CAD tool, a PDM or PLM vault, a shared network drive full of legacy files, and a resource-planning system for BOM and procurement data, often accumulated over a decade of acquisitions and tool changes. Leo is built as an intelligence layer that sits above that whole stack rather than inside any one piece of it, connecting to PDM, PLM, local and network directories, and resource-planning data at once, so a question about a part gets answered the same way regardless of which native tool happens to have the file open. Vendor AI and Leo are not competing for the same job. One makes a single tool easier to use. The other makes the knowledge scattered across every tool a team already owns actually reachable. For the conceptual difference between an agent and a copilot in CAD, see our breakdown of agents versus copilots.
FAQ
Sourced from Dassault Systemes' own 3DEXPERIENCE and CATIA R2026x release material on AURA and its Generative Experiences; Onshape's AI Advisor feature page and PTC's own newsroom announcement of the Onshape FeatureScript MCP Server; Siemens' NX Design Copilot announcement and its Teamcenter Copilot product documentation; and PTC's newsroom material on the Creo Advise, Assist, and Automate AI tiers.
See What Leo Adds to Your CAD Stack
One layer across every CAD and PLM tool your team already uses.
Built-in AI stays inside one vendor's walls. Leo connects your PDM, PLM, network drives, and resource-planning data, so your team gets answers no single tool can give alone.
Schedule a Demo →
#1 New AI Software Globally - G2 2026
Enterprise-grade security
Trusted by world-class engineering teams
