AI for CAD Tools
PTC added Advise, Assist, and Automate to Creo 13. Here is what each one actually does today, and where the built-in AI's reach ends.
·
⏱
7 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
Creo 13’s Advise is live and useful for in-session guidance today, Assist is a promising beta that reads model geometry directly, and Automate is an ambitious alpha still some months from broad production use. All three are scoped to the model open in the current session and to Creo’s own documentation, which means the institutional knowledge sitting in a PDM vault, an ERP system, or an engineer’s memory stays out of reach. Teams do not have to choose one AI over the other. The native assistant handles in-file questions well, while a connected layer like Leo handles the retrieval work across PDM, PLM, and standards that a single CAD session was never built to do.
PTC spent 2026 building artificial intelligence directly into Creo, and with Creo 13 and Creo+ 13.3 it shipped enough of it to actually evaluate. The work is organized into three tiers PTC calls Advise, Assist, and Automate, each at a different stage of readiness. Advise is live for every Creo user today. Assist is in beta. Automate is still in alpha. For a mechanical engineering team weighing how much of its AI strategy to leave to the CAD vendor and how much to bring in separately, the interesting question is not whether Creo’s AI is useful. Early coverage says it is. The interesting question is where its reach ends inside the model and the session, and what an engineer still has to go find somewhere else, usually in a PDM vault, a supplier spec, or a colleague’s memory of a project from three years ago.
What Advise, Assist, and Automate Actually Do
PTC’s rollout has three parts, sometimes referred to as its "Triple A" AI framework, and they are not equally available yet. Advise is the one every Creo 13 user already has. It is a chat interface trained on the product’s own documentation and decades of internal best practice, and it answers questions like where a command lives, why a feature failed to regenerate, or how a tolerance stack should be built for a given part family. It replaces a search through help files, not a design decision. It also sits alongside Creo’s existing generative design tools, which already handle topology optimization on their own track.
Assist is newer and still in beta, expected more broadly by late autumn 2026. It reads the active model directly rather than only answering from documentation, so it can pull every dimension and radius on a part into a report in seconds, work that would otherwise take an engineer a good part of an hour, or apply a parametric change across several related features at once. This is the tier that starts to look like a real productivity gain, because it is reasoning over the geometry in front of the engineer instead of a manual.
Automate is the most ambitious tier and the least finished. Still in alpha and targeted for release before the end of 2026, it takes a design objective stated in plain language, such as a target flow ratio or an acoustic constraint, and proposes modified geometry to meet it. PTC shows every Automate change in a sandbox first, with a before and after comparison, before anything touches the working file. The company has been explicit that human review stays in the loop by design, partly to manage the risk of a generative model producing a confident but wrong answer inside a CAD file that other systems depend on.
IN PRACTICE
It surfaces the relevant internal material, previous design decisions, past calculations, and backs everything with a cited source I can actually click on and verify.
- Yuval F., Clalit
Where the Native AI’s Reach Ends
All three tiers share the same boundary. They operate on the model open in the current Creo session, and on Creo’s own documentation set. Advise cannot tell an engineer what a similar part cost to manufacture, because that number lives in an ERP system, not in the CAD file. Assist cannot tell an engineer that a bracket with the same load case was designed eighteen months ago by a different team, because that history sits in a PDM vault, not in the open assembly. Automate cannot check whether a proposed geometry change conflicts with a customer-specific compliance note, because that note is usually a document sitting in a project folder, not a searchable field inside Creo.
The gap widens further in organizations that run more than one CAD system, which is common when a legacy program was built in SolidWorks or NX and the current one runs on Creo. A native assistant scoped to one vendor’s files has no way to reach a part sitting in another tool’s format, even when that part is the closest existing match to the one being designed today. It is the same boundary that shows up when comparing third-party AI tools built for Creo against the vendor’s own assistant: both are still working from inside one CAD environment.
None of this is a criticism of what PTC built. A CAD vendor’s AI is naturally scoped to the CAD file, in the same way a word processor’s spell checker is scoped to the document open in front of it. The gap is structural rather than a bug PTC is likely to close quickly, because closing it means reaching into every other system an engineering organization runs: the PDM or PLM vault, the ERP, the standards library, and whatever local or network directories hold the last decade of design decisions.
The Retrieval Gap: PDM, Standards, and Institutional Memory
Most of what a senior engineer actually knows does not live in the part file. It lives in the PDM system’s revision history, in old engineering change orders that explain why a tolerance was tightened, in a supplier’s datasheet that never made it into the bill of materials, and in the memory of the person who ran the analysis on a similar bracket two design cycles ago. None of that is visible to an AI feature that only reads the open Creo session, however good its model of the geometry is.
The practical cost shows up as repeated work. A team re-derives a tolerance stack that was already solved, or re-specifies a fastener that was already rejected for corrosion reasons on an earlier program, because the reasoning behind that earlier decision was never connected to the new part. Standards add another layer: a native CAD assistant can tell an engineer how to build a feature, but it is generally not trained to check that feature against the specific clause of a standard like ASME Y14.5 or ISO 1101 that governs the callout, or to flag that a company’s own internal design guide overrides the general standard for that part family.
This is the piece that a single CAD session, however AI-assisted, was never built to hold. It requires a system that treats the PDM vault, the standards library, and the ERP data as sources to search, not as separate destinations an engineer has to remember to check. Other ecosystems show the same pattern from a different angle: what the Teamcenter API actually gives an engineering team is real, but it is still an access point that something else has to query and connect, not a native chat assistant answering from inside the session.
Where an AI Layer Above Creo Picks Up
This is the specific gap Leo is built to sit inside. Leo is an AI assistant for mechanical engineers that connects to an organization’s full knowledge base, including its PDM and PLM systems, local and network directories, and ERP data, on top of a base of more than a million pages of standards, handbooks, and technical references. Where Creo’s built-in AI can only reason over the model open in the session, Leo can pull in the change order behind a tolerance, the supplier datasheet a part depends on, and the standard clause a callout needs to satisfy, and surface all of it with a citation an engineer can click through and verify.
Leo offers integrations with leading PDM and PLM platforms, including SolidWorks PDM, Autodesk Vault, PTC Windchill, Siemens Teamcenter, Arena PLM, and others, so it sits above whichever CAD and data management system a team already runs rather than asking them to replace it. That is the same principle behind connecting an AI agent to a CAD system’s own API or building on top of a Model Context Protocol server for SolidWorks: the value comes from reaching across systems, not from staying inside one of them. For a team running Creo, that means Advise, Assist, and Automate keep handling the questions they are good at inside the session, while the retrieval and institutional memory work happens in the layer above.
How to Decide What to Rely on Case by Case
A useful rule of thumb is to separate questions the model can answer alone from questions that need history. If the question is about the geometry currently open, such as a missing dimension, a broken parametric relationship, or a feature that will not regenerate, Creo’s built-in AI is the right first stop and will likely keep getting faster at it as Assist and Automate mature. If the question needs anything outside that single file, such as why a tolerance was set the way it was, whether a part has already been designed somewhere in the organization, or what a specific standard requires for a callout, that is a retrieval problem rather than a modeling problem, and it needs a system built to search across PDM, ERP, and documentation rather than one file at a time.
Teams that evaluate both layers together tend to get more value than teams that treat the choice as either-or. PTC’s own tiers are explicit that a person reviews every proposed change before it lands, and that same discipline holds for any AI layer sitting above Creo: a useful answer still needs a source an engineer can check, not just a confident one.
FAQ
See What Leo Finds That Creo Cannot
Connect your PDM, standards, and past designs to one AI layer.
Leo sits above Creo and your PDM to surface the design history, standards, and supplier data your CAD session cannot see on its own.
Schedule a Demo →
#1 New AI Software Globally - G2 2026
Enterprise-grade security
Trusted by world-class engineering teams
