We treat information architecture as a tidiness virtue. It's really a capacity constraint: your navigation is the ceiling on how much product you can ship.
Three different squads brought the same problem to crit in the same week, and I almost missed that it was one problem. Each had a new AI-driven pattern they wanted to introduce: an assistant entry point, a contextual surface, a place for the model to live in the product. Each had a thoughtful design. And each of them, when they got to the question of where it actually goes, ran into the same wall. Where to put it wasn’t entirely clear.
The instinct, mine included, is to treat that as a layout puzzle. Squeeze it into the existing nav, add an item, nest it under something that’s already there. But the third time I heard it that week, the shape of the problem changed. This wasn’t three teams with placement problems. It was one structure that had run out of room, and the structure was the information architecture.
Here’s the reframe: We talk about IA as a quality concern, the thing that makes a product feel coherent and findable, a tidiness virtue you get points for. That framing badly undersells it. Information architecture is a capacity constraint. It sets a hard limit on how much product you can add before the whole thing stops making sense, and you hit that limit long before you run out of ideas or engineers. The nav isn’t a container you decorate. It’s the ceiling on the roadmap.
Information architecture is a capacity constraint.
Which means the IA work that’s easy to defer, because it never has an urgent ship date and it always competes with features that do, is quietly governing whether those features have anywhere to land. A team can spend a quarter designing a genuinely good AI capability that the architecture has no room to expose, and discover the bottleneck only at the end, when the work is done and there’s still nowhere to put it. The constraint was upstream the whole time. We just don’t look at it until something can’t fit.
I’m treating this as a planning input now, not a cleanup task. Before a roadmap commits to a wave of new surfaces, the honest question isn’t “can we design these,” it’s “does the architecture have room for them, and if not, is restructuring the nav the real first project.” That moves IA from housekeeping to a gating dependency, which is what it’s been all along.
The teachable part
Information architecture isn’t a tidiness concern, it’s a capacity constraint. Your navigation sets a hard ceiling on how much product you can add before it stops cohering, and you hit that ceiling well before you run out of features to build. When new work has nowhere to go, the answer usually isn’t a cleverer layout. It’s that the structure itself is the next project, and it was the gating dependency all along.
On the shelf
A short stack if you want to get deliberate about this. Information architecture is one of the few design subjects with a real canon, and most of us have worked through less of it than we’d admit.
The shelf
