AI in Mechanical Design: The Fourth Generation of Transformation

Another technological disruption is transforming mechanical engineering software as artificial intelligence (AI) closes gaps and advances what’s possible.

What’s really going to change is still under speculation, but history can tell us one thing: Change is inevitable as limits are met and new foundations are built.

With a focus on PTC, let’s take a look at the history of MCAD from the 1980s to the very near future.

The First Generation of MCAD: From Paper to Parametric

In the 1980s, PTC introduced a modern parametric 3D design system to a field still largely working in 2D. The idea was to develop a technology that could define a 3D shape through relationships and constraints. With parametric design, one dimension could change, and the computer model accurately responds.

It was the first time a modeling system could think, even a little, accelerating drafting work and even how engineers thought about design itself.

The Second Generation of MCAD: From Workstations to PCs

Later, CAD work moved off expensive workstations and onto the PCs already sitting on desks. Software like SOLIDWORKS made professional 3D design accessible to a much wider audience.

The underlying system, however, stayed the same, with files living on local machines, using check-in/check-out PDM vault processes to work on a design. The philosophy to expand accessibility did not move with the hardware.

File-Based CAD Software Limitations

For years, file-based CAD software worked well enough that teams absorbed its frustrations as part of the job. They learned to live with version confusion, broken file references, software update mismatches between collaborators, IT overhead that compounded with every new seat, crashes that erased hours of unsaved work, and IP sitting in a supplier’s inbox long after the contract ended.

Still in the 2020s, CAD users are bogged down with the same issues. A survey of thousands of engineers quantified what most already felt: 7.1 hours lost per engineer, per week, to these friction points alone.

The file-based model had hit its ceiling.

The Third Generation of MCAD: Cloud-Native CAD

Mechanical engineering software has spent years running behind software in other industries. Over the last two decades, most professional tools have moved from local files to the cloud. Google Docs reshaped expectations for Microsoft Word.

CAD was the last major professional software category to make the jump, and now it has.

PTC’s Onshape, built on AWS infrastructure, rethinks CAD and PDM as a database-driven system. It runs in any browser, but the foundation bears no resemblance to a file-based tool. Every design lives in the cloud as a continuous, timestamped record of every action ever taken, tracking every sketch, extrude, chamfer, branch, merge, and even the missteps along the way.

Teams can collaborate on the same model, at the same moment, with granular permissions, so there are no copies flying around over email. Branching and merging, borrowed from software development, let engineers experiment on parallel paths and integrate their work without conflict. It's the first time a CAD system could support that workflow. And because compute runs on AWS rather than local hardware, an assembly with thousands of components can open on a MacBook, for instance, without a hassle.

The Fourth Generation of MCAD: The AI Transformation

The history makes one thing clear: Every previous shift in CAD produced better tools. This generation is the next iteration of mechanical engineering, with AI driving the greatest changes.

AI is only as effective as the data behind it, so it needs centralized, structured, continuously updated data that it can actually reach and reason about. That is precisely what cloud-native CAD provides.

This is where Onshape’s architecture stands apart. Because it’s database-driven, Onshape provides a rich, high-fidelity dataset that captures how a design was built, how it changed over time, what failed, and how it was fixed. That depth gives AI something meaningful to learn from and something useful to act on.

When that full history lives in a cloud database, an AI agent can query it just as a teammate can. Additionally, agents will only get more powerful precisely because of the collaboration capabilities Onshape already has.

The same three things that let people work together let you work alongside an agent. Scoped sharing means an agent is invited into a specific project with specific permissions, rather than handed the keys to your entire application. Real-time editing means you and the agent are in the same live model at the same moment, not trading copies back and forth. And full traceability means every action the agent takes is timestamped, visible, and reversible, so if it does something you don’t like, you can restore to any point, exactly as you would after a human collaborator’s change. You’re handing AI a defined task inside a system built for working together and not bolting a chatbot onto a file.

Onshape is developing agents that work alongside engineering teams in real time, for instance, fixing failing features or updating parameters and geometry against defined goals.

The tedious work gets handed off so engineers can stay on the work that requires judgment. Underneath it sits the Onshape MCP (Model Context Protocol) server, an open toolkit that lets teams connect their own agents, trained on their own data and tuned to their own standards.

None of it is possible when data lives in files. Competitors still managing files will be wrestling with version control, references, and check-in/check-out long after the rest of the industry has moved on.

Setting a Strong Foundation

PTC introduced parametric 3D CAD to the industry four decades ago. With Onshape, it’s leading the transition that may prove the most consequential yet.

The history of CAD is a history of removing constraints. Parametric design removed the constraint of manual drafting. The PC removed the constraint of expensive hardware. Cloud removed the constraint of the file.

The foundation for AI in mechanical engineering is already built, the data already structured, the architecture already in place, and the only question left is whether your design data lives somewhere AI can actually reach it.