Spend enough time observing engineers process change orders, create product variants and update drawings, and familiar patterns begin to emerge. Although many teams believe their work is too complex or unique to automate, much of their daily activity follows the same repeatable processes.
These insights come from years of working alongside engineers using SolidWorks, NX, CATIA, Creo and Inventor in real manufacturing environments. Some observations are lighthearted while others expose inefficiencies that impact engineering capacity, knowledge retention, design quality and delivery timelines.
The following 27 lessons capture common challenges seen across engineering organizations and illustrate where design automation can deliver measurable value.
27 Things I've Learned Watching Engineers Design the Same Part Over and Over
By Dan Grabowski, Design Automation Practice Head, IndX
I've spent the better part of two decades sitting behind engineers, watching them work in SolidWorks, NX, CATIA, Creo, and Inventor. Not in a creepy way. In a "show me how you actually process a change order" way.
After hundreds of these sessions, patterns emerge. Some of them are funny. Most of them are expensive. Here are 27 things I've learned, in no particular order except the order I got annoyed enough to write them down.
- Nobody thinks their process is repetitive until you count the clicks. Every team tells me "our products are too custom to automate." Then we watch a designer build a "custom" order and 80% of the clicks are identical to the last one. Custom is usually 20% of the work wearing the other 80% as a costume.
- The most dangerous file in your company is a spreadsheet named FINAL_v7_USE_THIS_ONE.xlsx. It contains your sizing logic, your pricing rules, or your BOM mappings. It lives on one person's desktop. It is your actual PLM system. Nobody will admit this.
- Engineers don't hate documentation. They hate documentation that dies. A wiki page describing a design rule is dead the day it's written. A design rule enforced by software is documentation that fights back.
- "It only takes me ten minutes" is the most expensive sentence in engineering. Ten minutes, forty times a month, across six designers, is 40 hours. You've hired an invisible engineer whose entire job is renaming drawing title blocks.
- Your senior engineer's brain is a single point of failure with a two-week notice period. Every team has one person who knows why the flange is 12mm and not 10mm. When they retire, that knowledge doesn't retire gracefully. It just disappears, and you find out six months later, in the field, expensively.
- Copy-paste is the leading cause of procurement surprises. A part number gets re-typed from CAD into a BOM into an ERP entry. Somewhere in that chain, an 8 becomes a B. Nobody notices until 400 of the wrong bracket show up on a truck.
- The second-most repetitive task in engineering isn't modeling. It's fixing the drawings after the model changed. Update the model, then spend the afternoon chasing dimensions, notes, and views that didn't come along for the ride. This is almost entirely automatable, and almost nobody automates it.
- Automation projects fail for cultural reasons, not technical ones. The API can do it. The question is whether Gary, who has done it manually for 19 years, believes the tool won't make him look replaceable. (Spoiler: the companies that automate don't fire Gary. They finally let Gary do the interesting work he was hired for.)
- Say the quiet part out loud: automation is not a headcount reduction play. In 20 years I have almost never seen a design automation project result in layoffs. I have constantly seen it result in the same team quoting twice as many orders. Engineering capacity is the constraint, not the cost.
- Start with the boring stuff. Everyone wants to automate the sexy thing — generative design, AI-driven optimization. Meanwhile their designers spend three hours a week exporting PDFs one at a time. Automate the PDF export first. It'll pay for itself before the AI whitepaper finishes printing.
- A $5k script that saves 2 hours a week beats a $500k platform that saves nothing. Some of our most beloved deliverables are embarrassingly small. A batch drawing exporter. A title-block updater. A hole-pattern checker. Clients name them things like "The Robot" and defend them like family.
- If your sales team promises delivery dates based on engineering "having a look," you don't have a sales problem. You have a design automation problem. Product configurators exist so that a valid configuration, BOM, and price can be generated without pulling an engineer off real work to answer "can we do it in blue, but longer?"
- Every "quick check" a senior engineer does is a design rule that hasn't been written down yet. "Let me just eyeball the wall thickness" is a rule. "That weld looks wrong" is a rule. Automated design checking is just the act of writing these down and letting software be the pedant so your senior people don't have to be.
- The tribal knowledge problem gets worse, not better, as CAD tools improve. Modern parametric CAD lets your best engineer encode enormous cleverness into a model — and nowhere else. Sophisticated models built by one person are just tribal knowledge with a nicer UI.
- Model-based definition is coming for your drawings whether you like it or not. OEMs are increasingly shipping models, not prints. If your downstream process still starts with "wait for the 2D drawing," you're going to feel this. Better to feel it on your own schedule.
- You cannot automate a process you can't describe. If the answer to "how do you size this component?" is "well, it depends," we're not ready to write code yet. But here's the good news: the act of preparing to automate — mapping the rules, the exceptions, the inputs — is valuable even if you never write a line of code. It's the process documentation you've been meaning to do since 2019.
- The exceptions are where the money is. Any process is 90% happy path and 10% weird cases. Teams say "we can't automate because of the weird cases." Wrong lesson. Automate the 90%, and route the 10% to a human — who now has time for it, because they're not doing the 90%.
- Engineers will trust an automation tool exactly as far as they can inspect it. Black boxes get abandoned. Tools that show their work — here's the rule I applied, here's why — get adopted. Build for skeptics. Engineers are professional skeptics; it's a feature.
- "We tried automation once and it didn't work" usually means "a vendor sold us something generic in 2017." Off-the-shelf configurators bounce off real engineering environments because real environments are weird. The weirdness isn't a bug to be worked around. The weirdness is the spec.
- Measure time-to-first-article, not lines of code. The best metric for a design automation project is: how long from customer order to released, manufacturable design? If that number isn't dropping, the project is decorative.
- Watch the workflow, not the workshop. You learn more in 45 minutes of watching one designer process one real change order than in three days of workshops about "the process." The official process and the actual process are usually different documents. Only one of them is true.
- The ROI math on design automation is boring, and that's why it works. A scripted automation runs $5k–$15k. A full configurator, $25k–$75k+. An engineer costs what an engineer costs. When automation saves each engineer 2+ hours a week, the payback typically lands in 6–9 months. No hockey sticks required. Just arithmetic.
- Beware the automation that nobody owns. A macro written by an intern in 2021 is currently load-bearing in more manufacturing companies than anyone wants to admit. If your automation has no owner, no source control, and no documentation, you haven't removed a single point of failure. You've just given it a file extension.
- Standardize before you automate, but don't wait for perfect standards. If every designer models a bracket differently, automation will faithfully reproduce the chaos at higher speed. But teams that wait for the Grand Unified CAD Standard before automating anything wait forever. Standardize the one part family you're about to automate. Then the next. That's it. That's the method.
- Your CAD platform's API can do more than you think. SolidWorks, NX, CATIA, Creo, Inventor — every one of them has a deep API that most teams have never touched beyond recording a macro once and getting scared. The capability has been sitting in your license this whole time, like a gym membership.
- The best automation ideas come from the most annoyed person on the team. Not the manager. Not the architect. The designer who audibly sighs when a certain order type comes in. Find the sigh. Automate the sigh.
- The hardest part is knowing where to start. Which is, honestly, why we built a whole offering around exactly that.
The Soft Pitch, Clearly Labeled As Such
We run a free 90-minute Design Process Assessment. A senior solution architect sits with your team (onsite or virtual), watches a designer do a real task — a change order, a new variant, a drawing update — and maps every redundant click and re-entry point in real time.
You walk away with a documented list of opportunities, sorted into quick wins vs. strategic plays, with honest budgetary numbers attached. No summary deck. No "we'll follow up." No obligation to ever talk to us again.
Worst case: you get a free, outside-eyes audit of your design workflow and a list you can act on yourselves. We're weirdly okay with that outcome.
Book your free 90-minute assessment →
Dan Grabowski leads the Design Automation practice at IndX. He has watched more engineers rename drawing files than any human should, and would like it to stop.
