Years ago, I led a team of exceptional engineers supporting production. A newly introduced product was averaging 30 hours to assemble instead of the 5 hours management had confidently estimated during one of their “everything is under control” meetings. We weren’t just losing money, we were hemorrhaging it. The VP decided this catastrophe was now my personal responsibility. Nothing says career advancement quite like inheriting somebody else’s disaster.
I’d love to tell you I rode in like a technical genius and saved the day through sheer engineering brilliance. Reality was far less glamorous. I listened to the assemblers, asked what was slowing them down, and then spent most of my time convincing management to support the changes the employees had already suggested months earlier. Miraculously, once people doing the actual work were allowed to influence the process, assembly time dropped below 10 hours within a month and stabilized around 4 after three months.
Biggest mistake of my career.
The reward for success is always more work. I had forever doomed myself by demonstrating competence. I was immediately reassigned to reducing production costs across all product lines.
Not my idea of fun
None of this is particularly important except for one detail: it explains why I eventually found myself sitting in a conference room asking accountants how we calculated product costs.
Now let me say this clearly: I have enormous respect for accountants. There’s absolutely no way I could stare at columns of numbers all day without quickly losing my will to live. But there I was, trying to understand how we priced our products while everyone else in the room looked at me like a moderately bright preschooler. They were accountants, masters of all things accounting, I was only an overpaid engineer.
I’ll spare you most of the details because, frankly, I was bored enough for all of us. Eventually we arrived at something I actually cared about: components. Vendors change. Technologies evolve. Parts go obsolete. Some of our products had been in production for over 20 years, and the cost of individual components fluctuated constantly depending on suppliers, quantity breaks, availability, and whatever economic goat sacrifice was affecting the electronics industry that week.
Then they explained how total product cost was calculated.
A query.
That’s it. Somewhere in the distant past, some long-forgotten soul had written a query in our production software that totaled individual component costs. Nobody knew exactly how it worked. Nobody knew what costing method it used for the components. Nobody knew whether it was averaging costs, using FIFO, last cost, highest cost, or determining prices through astrological alignment.
So I asked.
“How exactly does the system calculate component cost?”
“We use the number the computer gives us.”
That sentence still haunts me.
Nobody in the room understood why this terrified me. The accountant who originally created the query had long since vanished into the misty wilderness of retirement or death. No documentation existed. Tribal knowledge had elevated this mystery query into sacred scripture. The computer had spoken, and therefore the computer must be right.
This was around the time I started mentally pricing whiskey by the gallon.
I’m old enough to remember when four-digit calculators cost over $100, and shady vendors made fortunes selling defective calculators at discount prices. They were wrong just infrequently enough to make you trust them over your own intuition. Muck later spreadsheet programs exploded across the market the same way weeds explode after a spring rain. Each with its own flavor and individual computing quirks. Most disappeared almost immediately, usually after destroying somebody’s inventory, payroll, or sanity.
That’s where “trust but verify” became welded into my engineering DNA.
I love spreadsheets. I use them for decisions, estimates, retirement calculations, and convincing myself an extra $50 payment on the mortgage somehow saves enough in the long run to buy that new CNC now. But I’ve also built enough bad spreadsheets to know they should never be trusted blindly. I always run test cases. Always. I honestly don’t understand how people can build a spreadsheet without understanding exactly what it’s doing underneath the hood.
And now we arrive at AI.
Great idea
This matters because we’re rapidly outsourcing more and more thinking to spreadsheets, apps, and now AI systems wrapped in sleek interfaces and marketing hype. Worse, modern AI is heavily optimized to be agreeable. It wants to encourage you. Support you. Validate your ideas. “Incredible insight, your plan to build a nuclear-powered waffle iron is both innovative and disruptive.”
What AI rarely says is:
“This is complete nonsense and you should stop immediately.”
Not exactly a great subscription model.
Nor is AI guaranteed to be correct. AI was trained on the Internet, which means humanity essentially dumped the contents of the world’s largest junk drawer into a machine and yelled, “Learn!” Unfortunately, the Internet contains both peer-reviewed research and a guy named Randy explaining why aliens built the interstate highway system.
AI cannot reliably distinguish between factual information and material best suited for lining the bottom of a bird cage. Add sarcasm, parody, propaganda, clickbait, and political nonsense into the mix, and things get even uglier.
Imagine an AI trained entirely on Fox News. Worse, imagine one trained entirely on Congressional records. At that point the system wouldn’t need artificial intelligence, it would need artificial life support.
Sure, valuable information exists in those mountains of noise. But separating the rare gems from the endless avalanche of nonsense? That’s the kind of task that makes quantum supercomputers quietly reconsider their career choices.
AI is an incredibly powerful tool. But it’s still just a tool. You need to understand its limitations, verify its output, and occasionally apply common sense before accepting what it tells you.
You wouldn’t hand a chainsaw to a six-year-old and expect handcrafted furniture.
Yet somehow many people are doing exactly that with AI.
Trust but verify. It’s more important than ever in the AI age.
Today’s song from Songer… Trust But Verify
© 2026, Byron Seastrunk. All rights reserved.
