Users don't need a smart system. They need a clear one.
The first time I used the system, I stopped.
Not because it was complex, but because it demanded more thinking than the task deserved. Every step implicitly asked are you sure? - instead of helping me move forward.
It was a government system in Saudi Arabia. The experience wasn't obviously bad - and that was part of the problem. Obviously bad gets fixed. This was full of small, repeated friction:
Steps longer than they needed to be
An unclear sequence
Terminology that didn't match how users speak
A constant demand for extra attention
Why this kind of failure survives
Small friction never gets reported. Nobody files a ticket to say a step felt slightly tiring. Users push through, feel drained, and can't name the cause.
So these systems look healthy in reporting: completion rates are acceptable, complaints are absent. The cost shows up elsewhere - in time-to-complete, in call-centre volume, in users who postpone the task until they have no choice.
We chose not to start with design
The instinctive request here is a redesign. We declined, because redesigning a system whose failures you haven't precisely diagnosed produces a different-looking system with the same problems.
We started with a full evaluation - not of the interface, but of how users actually move through it.
We drew on established UX principles:
Hick's Law - more options slow down decisions
Miller's Law - working memory is limited; users can't absorb everything at once
Jakob's Law - people expect patterns familiar from other systems they use
But principles alone aren't enough, and they're often used to justify decisions already made. What settled it was observing real usage instead of assuming.
What surfaced
The problem wasn't one screen. It was how the system thought - its assumption that the user held the full context at every moment.
When we moved to solutions, the question wasn't "how do we redesign this?" It was "where do we reduce the load?"
We shortened steps
We resequenced the flow
We simplified the language
So that every step moved the user closer to their goal rather than further from it.
Same system, completely different experience. Clearer, faster, lighter. And most importantly, coherent.
A significant part of the work was AI-assisted, particularly in accelerating ideas and building early versions to test.
The takeaway
Government and enterprise systems tend to be pushed toward intelligence: more options, more validation, more features.
But users aren't asking for a smart system. They're asking for one where they know where they are, what to do, and what happens next.
Clarity isn't a cosmetic layer applied after the logic is built. It is the logic.