Writing · September 29, 2026
Some problems are never quite solved
Every few years the software industry discovers the answer to everything, and it always arrives with a new name. At one time it was Document Management: get your files into one system and your troubles were over. When the troubles came back, the answer was Knowledge Management. Then Content Management, then Enterprise Content Management once that stopped impressing anyone, then Big Data, which was going to turn the exhaust of daily operations into gold, and then Digital Transformation, which was going to make every company a technology company. Each one showed up as a category you could buy, sold as the thing that would finally make the mess make sense.
I've built inside several of these waves, and they run the same way every time. A new word arrives with a demo that genuinely dazzles. The platform ships. People move their work onto it. And a few years on, the same complaints come back wearing the same faces, and the industry, already a little bored, goes looking for the next word. The word this year is AI. You already know how this goes.
The technology was never the point. Some of it was real, some of it wasn't, and it made no difference either way. What mattered is that the names were market moments, and the problems underneath them don't care what we call the wave. A few of those problems have outlasted every name on that list, and they'll outlast this one too. I want to talk about just one of them, because it's the one that shows up on every budget: the systems you buy to get control end up controlling you, and billing you for it every year.
It never looks like a trap when you sign. You buy a system to get a handle on something — your documents, your customers, your spending — and for a while it helps. Then your work moves inside it. Your data takes its shape. Your processes bend to fit it. By the time it stops serving you, leaving costs more than staying, so you renew — not because it's good, but because the exit is priced higher than the ransom. The bill arrives every year whether the software earned it or not.
This scales further than people expect. A company loses track of what it's paying for, finds three tools doing the same job, and keeps all three because untangling them is someone's whole quarter. Push it far enough and you get major US cities that spend so much on software they can no longer say who bought what, or why, or whether it's still needed — only that the invoices keep clearing. They need to cut and can't, held in place by past investments and by systems that have quietly started working against the people who depend on them. The lock-in isn't a clause in a contract. It's the accumulated weight of everything you already put inside.
Some governments have decided this is serious enough to organize around. Germany calls it digital sovereignty — the idea that a country shouldn't have its records, its communications, and its institutions living inside systems it doesn't control and can't leave. It's the same problem, seen from a great height. A CIO doesn't need the word. They see the renewal.
The way out is not the next system that promises everything, and it is certainly not the one with AI in the name. It is a quieter, more boring discipline, and it starts with a question the hype never asks: when this wave recedes like all the others, what do you still hold? The answer worth building toward is data you own outright — files you can open without the vendor, formats that outlive the system that wrote them, work that stays yours when the platform is gone. It doesn't demo well. It just doesn't leave you holding an empty box when the music stops. Some problems are never quite solved. That's why they're the ones worth building for.