The short version: At Autodesk University 2026 in Las Vegas (September 15 to 17), Autodesk previewed AI features for Fusion, Inventor and Vault that rebuild imported geometry as editable parametric parts, help automate assemblies and let engineers search their data in plain English. Autodesk says the rollout schedule comes in 2027, so for now this is a roadmap, not a release.
What Autodesk showed
Autodesk is positioning Autodesk Assistant, its natural-language AI, as the front door to most of its software. CEO Andrew Anagnost described it as the company’s AI interface for design and make, built to bring project context into everyday work while engineers stay in control. A new Agent Builder will let companies feed their own data into the assistant.
For mechanical design and manufacturing, the features that matter most are the ones aimed at repetitive work:
- AutoConstrain and AutoTimeline. AI adds constraints and a feature history to imported geometry, turning a “dumb” STEP or neutral file into a part you can actually edit. Autodesk’s data solutions lead compared it to working out the recipe behind the geometry.
- AutoAssemble. Partly automates putting assemblies together.
- Manufacturing prep. AI-assisted machine setup, toolpath generation and production preparation.
- Inventor. The assistant will use Inventor’s APIs and iLogic to analyze models, change parameters and components, and create model states. Inventor is also becoming the first “Fusion Connected Client,” linking the desktop program to cloud product data.
- Vault. Users will be able to search existing design data by asking in plain English.
When engineers will get it
Autodesk Assistant already runs in its main design and manufacturing products, but the new capabilities were previews. Engineering.com reported that Autodesk expects to share the rollout schedule in 2027, and All3DP pointed out that the expanded assistant is not generally available yet. Teams planning around these features should treat 2027 as the earliest realistic date.
What changes in the day-to-day work
Ask any mechanical designer where the hours go and the answer is rarely “designing.” A lot of the week is spent cleaning up supplier models that arrive as STEP or IGES files with no history, rebuilding them so they can be changed, hunting through Vault for the right revision, and redoing setup work that someone has already done on a similar part. Those are exactly the jobs Autodesk is aiming at.
- Supplier and legacy files. If AutoConstrain and AutoTimeline work as previewed, a designer who receives a dumb solid from a vendor would get back a part with constraints and a feature tree they can edit, instead of remodeling it by hand.
- Finding data. Plain-English search in Vault means asking for “the latest bracket revision for the conveyor frame” rather than remembering file names and folder paths.
- Inventor automation. Many Inventor shops already use iLogic rules to drive configurations. Having an assistant that can call iLogic and the Inventor API could put that kind of automation in reach of engineers who never learned to script it.
- Shop handoff. AI help with machine setup and toolpaths shortens the gap between a released model and a part on the machine, which matters most for job shops and fast-turn teams like the race shop Autodesk highlighted.
None of this removes the need for an engineer to own the result. A constraint the AI guessed wrong, or a search that pulls an obsolete revision, still ends up on a drawing unless a person catches it. The work shifts from building models to checking them.
Who should pay attention
Teams running Fusion or Inventor with Vault get the most from this release, especially those that receive a lot of outside geometry: OEM suppliers, contract manufacturers, machine builders and product design firms. Shops on SolidWorks, Creo or NX will see similar features from their own vendors, since every major CAD company is adding an AI assistant, but the specific tools in this story are Autodesk’s.
What it means for hiring
AI that cleans up imported geometry takes grunt work off designers, but it raises the bar on judgment. Someone still has to check that the rebuilt constraints match design intent before a part goes to the shop.
- For Inventor shops, iLogic and API experience becomes more valuable, since that is what the assistant will drive.
- Clean, well-organized Vault data is what makes plain-English search useful. PDM discipline is worth asking about in interviews.
- Strong CAD fundamentals still matter most. Ask candidates how they would verify a model the AI rebuilt.
Hiring designers or mechanical engineers? See our mechanical engineering recruiters and CAD staffing pages, or our guides on how to hire a mechanical engineer and how to hire a CAD designer.
Interview questions for designers working with AI CAD tools
- A supplier sends a STEP file and the software rebuilds it with constraints. How do you confirm the rebuilt part matches the original intent before you change it?
- Tell me about an iLogic rule or API script you wrote, or one you would write, to automate a repeat task.
- How does your team keep Vault or PDM data clean enough that searches return the right revision?
- When a tool suggests an assembly or toolpath, what would make you reject it?
Frequently asked questions
Is Autodesk’s new AI available in Fusion now?
Autodesk Assistant already runs in Autodesk’s main design and manufacturing products, but the features shown at AU 2026, including AutoConstrain, AutoTimeline and AutoAssemble, were previews. Autodesk said rollout details will come in 2027.
Will AI CAD tools replace mechanical designers?
Not based on what Autodesk has shown. The features target repetitive work like rebuilding imported geometry and searching for files. Someone still has to define design intent, check the output and sign off on drawings.
What skills should a CAD designer build now?
Solid modeling fundamentals and GD&T come first. After that, iLogic or API scripting for Inventor users, good PDM habits, and the judgment to review AI-generated models are the skills most likely to set candidates apart.
More stories like this are on our AI for Engineers hub.
