A magnet with a hole in it

An entrepreneur who went back into export trading, using AI to build websites that bring in overseas inquiries, told a story about a magnet.

A European client sent a drawing: a magnet fourteen centimetres long, with a hole through the middle. How hard can that be, he thought. He took the drawing to a supplier platform and asked factory after factory — and found that most of them couldn't do it. After several days he found one willing to, at eight hundred. He kept looking and found a small workshop that could do it for a little over five hundred.

He quoted the client a thousand. The client accepted without haggling. Over the following months, the orders kept coming back.

What the client was paying for

Getting from eight hundred down to five hundred earns a margin. Getting from five hundred up to a thousand earns something else.

His own explanation: he was local, spoke the language, had the platform, and was willing to ask one factory after another — and even so it was hard. For a client on the other side of the world, getting that object made would be several times harder. The thousand wasn't the price of a magnet. It was the price of someone who could get it done.

We've turned that sentence over for a long time, because it lands exactly on the thing most misunderstood about the AI era.

AI has made a great deal of "being able to do it" cheap: drafts, translation, analysis, code — work you used to hire for now arrives in an afternoon. So people ask: what's a consultant for now? What are my staff for? What is my company for?

The magnet answers: being able to do it got cheap; getting it done didn't. Between a drawing and an acceptable magnet lie finding the factory, comparing prices, negotiating specifications, chasing delivery, carrying the risk. None of it is magic. But someone has to string it together and answer for the result. That stretch is what clients pay for — without haggling.

Let AI ask the extra question

The same entrepreneur's second magnet client also looked small at first: on the client's initial request, a hundred-odd US dollars in total. Not worth the bother, he thought.

The AI nudged him: ask what the client is actually using them for. They're probably not just buying magnets — they're probably building something.

He asked, and the requirements grew: beyond the magnets, marking, printing a logo, winding coils. With the downstream processing and labour counted in, the order came to about fifteen hundred dollars, with more than a thousand in profit.

What happened there wasn't a mark-up. He caught the work the client would otherwise have had to find someone else to do. Same client, same project: ask shallowly and you see a materials cost; ask deeply and you see what the whole thing needs.

Notice who prompted him: the AI. It didn't do the business for him. It asked one more question at the moment he was about to give up. That may be AI's most underrated use inside a small company — not doing your work, but holding a follow-up question open for you when you're too busy, too experienced, too sure.

When the client talks price, talk technique

He has another line we'd recommend to anyone in business: "Whenever a client talks price with me, I talk technique with them."

This isn't changing the subject. When a client brings someone else's quote, he brings the conversation back to precision, specification, lead time and the services he provides that the other doesn't — so the client sees that the two quotes contain very different amounts of work.

If all you do is pass on a price, all you can be is compared on price. Help a client find the problem, improve the design and locate the right supply, and you've earned the right not to fight for the lowest number.

When output gets cheap, stop competing on output. Compete on the things that didn't get cheap — judgement, responsibility, the ability to get it done.

What a magnet with a hole in it is worth depends on who asks, how many layers deep they ask, and whether anyone is willing to spend a few days running around for it. That was true before AI and it's true after. The difference is that you now have a tireless assistant holding open the question: "and what else?"