The Pen Clip in Qatar
The pin was bent.
We were standing in a mechanical room in the middle of the Qatari desert, five years into a project that had almost broken everyone involved. Multiple project managers had come and gone. The schedule had stopped and started so many times we’d lost count. We’d traveled sixty hours round trip to be there for this commissioning window, and now a bent pin inside a VFD card slot was threatening to send us all home with nothing to show for it.
The mood was immediate and heavy. Everyone knew what a bent pin meant. You can’t just order a replacement PLC card slot and have it show up in the Qatari desert by Tuesday. We were looking at weeks of delay, minimum. Another sixty-hour trip back. More schedule slippage on a project that had already eaten years of people’s lives. The VFD wouldn’t run without that software card seated properly, and the card wouldn’t seat with a mangled pin.
I looked at the damaged slot. Looked at my tool bag. Looked at the team.
“Fuck it. If it’s already broken, I can’t make it brokener.”
I pulled the card out and went to work. It was like performing surgery in a mechanical room — tiny pin, tight clearance, hundred-degree heat, and the knowledge that if this didn’t work we were done for the day, possibly for the month. I went through my tools one by one, looking for anything with the right geometry to reach the pin and bend it back to an angle where the card would seat.
About twenty-five minutes in, I found it. A pen clip. Just a standard pen clip that I’d pulled off and stuck in my bag at some point. It had exactly the right curve, the right thickness, and the right amount of flex to slide into the slot, hook the pin, and coax it back into position.
I slipped the software card back in. It seated. Fired up the VFD.
Back in business.
Later that day, we ran the world’s largest vertical wind tunnel for the first time. I was one of the first five people to fly in it.
I tell that story a lot. The reason is not the obvious one.
The story sounds like a lesson in improvisation, and to a certain extent it is. But the part of the story that actually matters is not that I improvised. It’s that the most valuable thing in that mechanical room wasn’t the pen clip. It was the years of practice that taught me, when faced with a damaged pin and a tool bag, to see the geometry that would work.
The judgment was the load-bearing thing. The pen clip was just an implement.
That distinction is the one I want to talk about, because we are in the middle of a profession-wide conversation about AI tools that has it almost exactly backwards.
In the last two years, AI tools have become genuinely useful for engineers. They can generate calculations, write code, run simulations, draft designs, produce reports. The output is often plausible. Sometimes it’s correct. Sometimes — and this is the part that should keep you up at night — it’s confidently wrong in ways that look right to engineers without judgment.
The pitch you’re hearing from every direction is learn to use these tools. Get the prompts right. Stay relevant.
That pitch is a half-truth, and the half it gets wrong is the half that matters.
AI tools do not make you smarter. They make you faster.
That is a profound difference. Speed is an accelerant. An accelerant on a good engineering process — clear problem definition, strong verification habits, a calibrated nose for when an answer is wrong — produces astonishing results. I designed a complete dry-cooler product line in forty-eight hours last year with an AI team. That’s real. That’s the upside.
But an accelerant on a bad process produces a different result. It produces wrong answers, faster. It produces confident-looking garbage at machine velocity. The engineer who can prompt an AI to generate a 200-page design report in an afternoon is the most dangerous professional in the industry — if that engineer hasn’t built the judgment to know whether the report is correct.
The engineer who has built the judgment, and uses AI to deploy it, is the most valuable professional in the industry. By a wider margin than at any point in the last fifty years.
This sets up a question I think every engineer needs to answer for themselves in the next twelve months: am I building judgment, or am I substituting AI for judgment?
Three quick tests.
Test one: do you trust AI’s first answer? If the answer is yes, you are not building judgment. The right answer is no, I do not trust AI’s first answer, I treat it as a starting point that I verify against deterministic tools and field experience.
Test two: when an AI gives you an output, can you say what would make it wrong? If the answer is no, you are using AI as a smart-friend who tells you the answer. That is not engineering. That is faith. Engineering is the discipline of being able to articulate, in advance, what would make an answer wrong, and then checking for those failure modes.
Test three: what are you doing in the field? Not at your desk. Not in front of a chat window. In a mechanical room. On a job trailer. At a commissioning. Walking pipes. Standing in front of equipment. Verifying with your eyes and hands what your spreadsheets and AI agents say should be true. If the honest answer is not much, your judgment is atrophying, and AI tools are masking the atrophy.
The pen clip story matters because the engineering wasn’t the pen clip. The engineering was the years that made the pen clip useful.
Those years were not glamorous. They were a high-rise mechanical job in New York where I learned what a properly designed riser actually felt like under load. They were eighteen months mapping hundred-year-old piping inside the New York Public Library, where I learned that drawings lie and only the field tells the truth. They were nights at a job trailer in the desert when nothing worked and somebody had to figure out why. They were bad bosses, good mentors, mistakes that I had to own and fix.
You cannot speed-run that. AI cannot give it to you. A four-year degree barely starts it. The only way through is in.
The good news — the actual good news of this AI moment — is that the engineers who do go through it now have access to a multiplier their predecessors never had. The same sixty-hour Qatar trip, with AI agents in the loop, would have produced different work. Better work. The pen-clip moment would have been the same — that’s a human standing in a mechanical room with their hands — but the analysis, the verification, the documentation around it would have all happened at five times the speed.
That’s the bargain on offer right now. Build the judgment, the slow boring way. Then deploy it at machine velocity, the new way.
The engineers who take that bargain will own the next decade of the profession.
The engineers who try to skip the first half — who treat AI as a substitute for the years of practice rather than a multiplier for them — are going to produce work, fast, that nobody can verify and that eventually causes a project to fail in a way nobody can explain.
The choice is the same as it’s always been. AI just made the consequences arrive faster.
That’s the prologue of a book I just wrote called The AI-Era Engineer. It’s a field guide for engineers building careers in this moment — written from twenty years of practice across building systems, vertical wind tunnels, CFD, and product design.
If the argument here lands, you can read the whole book at theaieraengineer.com.
If you’ve been around the profession long enough to have your own pen-clip story — I’d love to hear it.
— Kyle
Originally published at https://beneshengineering.com/the-pen-clip-in-qatar/
Kyle Benesh, PE is a mechanical engineer licensed in New York and Texas, and the managing partner at Benesh Engineering. He is the author of The AI-Era Engineer: A Field Guide for the Next Generation.