You Could Build It Yourself. Should You? The Story of the $640 Toilet Seat

In 1985 the Pentagon paid Lockheed $640 apiece for toilet seats on a Navy patrol plane. It became the symbol of waste. The fuller story is more useful: the money bought a custom-molded toilet assembly, made a few dozen at a time by a company built to make aircraft. A plane maker can build a toilet seat. It just pays plane-maker costs to do it. AI works the same way. You can build it yourself, and it may turn out fine. Count the hours your best people spend away from the work you are paid for, and decide from there.

The headline everyone remembers

In early 1985, Congress heard that the Navy had paid Lockheed $34,560 for 54 toilet units for its P-3C Orion patrol planes. That is $640 each. The Christian Science Monitor reported it that spring, right after the “now-legendary” $7,600 Air Force coffee pot. Around the same time came the $435 hammer.

Together they became shorthand for government waste. Of the three, the toilet seat is the one people still quote.

The part the headlines left out

The Pentagon said the $640 bought the whole molded plastic cover that fits around the plane’s toilet. Lockheed’s president still cut the price to $100 each and paid back $29,165, to “put to rest an artificial issue.”

The hammer shows how a price like that can happen. Steven Kelman, a former head of federal buying policy, called it “an accounting artifact.” The hammer came in one big order with many complex spare parts. The contractor spread its research and engineering costs evenly across every item, so a $15 hammer picked up the same $420 of overhead as an engine part. “The hammer got as much overhead as an engine,” Kelman said.

The records don’t say whether the toilet seat was priced the same way, so I won’t claim it was. Both stories still point at one thing. A company built to make one kind of thing carries its full costs into everything it makes. A plane maker builds with aerospace engineers, aerospace paperwork, and aerospace overhead, even when the part is a toilet cover.

Your business has toilet seats too

Most of the cost of an AI project is the work around the model, as I explain in why an AI project costs real money. When a manufacturer does that work in-house, the cost rarely shows up on one invoice. It looks like this:

• Your best people’s hours. The person who knows your data best is usually the one you can least spare.
• Learning from scratch. New tools, new mistakes, and time spent finding out which mistakes matter.
• Rework. The first version answers the wrong question, and you start again.
• One-person risk. It works, but only the person who built it understands it. When they leave, nobody can fix it.

That last one is the same risk we fix with Process Capture: know-how that lives in one person’s head is a risk, whether it runs your quoting or your AI.

Your skills earn your money

AI is amazing, but it doesn’t know what you want. Someone has to teach it, test it, and keep it honest. That takes skills your team may not have yet.

You could learn them. Given time, you might even do it better than an outsider, because you know your business. It would cost time and energy you would rather spend on the work that pays you. I get paid for my skills. You get paid for yours.

When you should build it yourself

Sometimes building it yourself is the right call. Build in-house when:

• You have the people. You already employ people who build software and data systems, and they have time to spare.
• It is core to how you compete. If the AI is the reason customers pick you, you want to own every part of it.
• It is small. A spreadsheet formula, a simple report, or a prompt your team uses to draft emails does not need a project.

If none of those fit, buying the skill usually costs less than growing it. Either way, count the hidden hours before you decide.

Bring one question. If you are weighing build against buy, book a call from the AI Builds page. I’ll tell you plainly if it is one to build yourself.