Adaptive UI
Most software ships screens. Everyone gets the same ones, and the work of deciding what matters today is left to the person. An interface that is composed rather than shipped moves that work to the machine — which raises an obvious objection.
Most software ships screens. Someone designs them once, everyone gets the same ones, and the work of deciding which part matters today is left entirely to the person looking at it.
That is a strange division of labour. The machine knows what happened this week. It knows what you did last time. It knows what is about to expire. And it uses none of that to decide what to put in front of you — it just lays out every drawer and lets you search.
Composed, not shipped
The alternative is that there is no screen. What you see is assembled each time, from your own record and the situation you are actually in.
A grower opening it during harvest and a cook opening it on a Tuesday evening are not looking at the same product with different data in it. They are looking at different arrangements, because the arrangement is derived rather than designed.
In practice that means a few things vary: which mode the thing is in, what context gets loaded before it composes anything, and — most importantly — what it decides not to show you at all. Most of what happens in a food business is routine. An interface that surfaces all of it is not being helpful, it is being lazy.
The obvious objection
An interface that changes is an interface you cannot learn.
This is a real cost and it is usually underweighted by people who find adaptive systems exciting. Consistency is not a limitation of old software; it is most of what makes software usable. If the thing rearranges itself, you can never build a habit, you can never point at it over someone's shoulder, and you are permanently a beginner.
Plenty of adaptive interfaces have failed exactly here. Menus that reorder by frequency, home screens that guess. They were not wrong about the data. They were wrong about what a person needs from a tool.
What has to stay still
The resolution is that two different things are being confused.
The grammar is fixed. The filling adapts.
There is a small set of shapes, and they never change. Something the machine noticed. A choice, always with the option to write your own answer instead of picking. A proposal waiting on your word before anything irreversible happens. A record of what it did and why.
You learn those shapes once, in the first week, and they are the same shapes forever. What varies is which ones appear today and what is inside them.
That is the same trick as a language. English has a fixed grammar and you have never heard most of the sentences you will hear this year. You are not confused by them, because the structure was learned long ago and only the content is new.
The test
You should be able to predict what it will do, even when you cannot predict what it will say.
If you know that nothing costing money happens without you saying yes, that every choice can be answered in your own words, and that you can always ask why it did something — then it can surprise you with content all day and you will never feel lost.
An adaptive interface that fails that test is not adaptive. It is unreliable, which is a different thing wearing the same clothes.