AI will not replace the CAD engineer. It will replace their 3 hours of clicking

When I say this sentence in a meeting, the same thing always happens. First there is silence. Then someone, almost always the person with the most years on the team, starts making the list.

The list

The list is the set of tasks no engineer ever put on their resume. Filling in Vault properties file by file. Checking 200 components before a release. Exporting BOMs and reconciling them against the ERP by hand. Generating the same report every week. Renaming and reclassifying documents because the standard changed.

Added up, those tasks eat hours of every week of a person who was hired to design. It is not an attitude problem: it is operation that landed on the engineer because they are the only one who knows the system well enough not to break it.

Why operation ends up with the engineer

PDM and CAD systems concentrate critical information, and that is exactly why they are scary. Nobody wants an intern touching the lifecycle states of 500 files. The result is perverse: the most mechanical tasks in the department end up in the hands of the most expensive and scarcest person, precisely because they are delicate.

That is where a well-designed agent changes the equation. Not because it is smarter than the engineer, but because it can execute the delicate operation under hard rules: validation against the system's real schema, dry-run before every write, explicit confirmation and an audit trail for every step. That is a combination a tired human cannot guarantee on a Friday.

What does not get replaced

On the other side is what I do not see being replaced on any near horizon: deciding a tolerance, choosing between two assembly architectures, knowing from experience that a supplier will not deliver, spotting that the customer requirement is wrong before modeling anything.

That is not information: it is judgment. It is built over years of your own and other people's mistakes, and it is exactly why you hire a senior engineer. An LLM can rank options and draft justifications, but responsibility for the technical decision is not something you delegate.

What this means if you run a team

The useful question is not whether AI will take jobs. It is how many hours of your team's week go into repetitive operation, and what it is worth to give them back to design. My recommendation is to start by measuring it: one week of honest logging usually reveals 3 to 5 hours per person.

Then automate exactly that, with validations and audit trails, and leave judgment where it always was. The agent takes the list. The engineer goes back to engineering. That is the division of labor I defend, and the one I build in every project.

I have watched it work in practice with VaultPilot, the agent we developed together with Avant Leap for Autodesk Vault Professional 2026: the tasks it absorbs are exactly the ones on the list, and decisions keep going through the same people they always did. Nobody on the team lost their job over it. What they lost were the afternoons of clicking.

Next time someone asks you whether AI is coming for the engineers, run the list exercise. The answer is usually right there, and it is almost never scary.