Skip to content

Case study UX review

Bad design is felt.

How a senior care founder’s AI-built platform got the interface its foundation deserved.

By Steven & Anthony5 min read

“I opened your email, followed the link and then got lost in the rich feedback. It was exactly what I needed.”

Cameron ParsonsCEO, Parsons Group

Is this ready for real users?

Cameron runs several senior care facilities, and he saw a need for a more affordable at-home care option that doesn’t cut corners on care. He has no software development background, so he rolled up his sleeves and built the platform himself with AI. Ambitious, but so is Cam. By the time he came to Aspenglow, much of the pilot was already built. His question was the one we hear most often: is this ready for real users, and am I building on a solid foundation?

It’s a common worry for founders building with AI. AI is good at making something work on one screen or one machine. Once real customers and real protected health data arrive, you want a human who has done this before to tell you whether you’re headed in the right direction. Valuable software takes two kinds of expertise: knowledge of the industry, which Cam has, and the experience to know where to look and what tends to kill software projects as they grow. That second kind is what we bring.

You know it when you feel it

The good news came first: the foundation was sound. Cam’s approach to AI-assisted (or vibecoded) development produced a platform that was fast, secure, and structured so he could keep building on it. What Cam wasn’t happy with was the interface. Good interface design (called UX in the software biz) is rarely noticed, but bad design is felt, even if the user can’t name why. That’s what Cam was feeling in his app.

  • 0.5 sto sign in
  • 2.1 sslowest page we opened
  • 4tools handed to his AI team

So we logged in as a care coordinator and went through every page of the app. We mapped the information architecture (every screen and the actions a user takes on each), traced the journeys behind the major tasks, and wrote down the job each page is there to do and how well it does that job. That written record is what turns “it feels off” into specific fixes.

A strong visual hierarchy matters

The app opens on a dashboard where remote care coordinators see the tasks they need to do for their patients. Functionally, it worked. What Cam felt was a common AI design problem: AI doesn’t see a page with human eyes. People need a visual hierarchy, cues that tell the eye where to go first on every page, and AI rarely builds one on its own.

Hierarchy comes from titles, subtitles, and labels that each look different. A title orients the user to the page’s purpose and has a different size (and usually font) than the rest of the page. Subtitles explain how to use the page, and labels make clear how each piece works, again with their own sizes and weights. Done well, it lets the eye grok a page in a second and know what to do.

Typographic example of visual hierarchy: “YOU WILL READ THIS FIRST” in huge type, “Then you will read this next” beneath it, a smaller callout you will likely read before the paragraph, and a dense body paragraph you read last.

Without that hierarchy, AI-built interfaces tend to feel… flat. You’ve probably felt this on an AI-designed page. On busy pages like a dashboard, your eyes bounce around looking for an anchor. You can’t focus on anything, and you can’t say why.

Getting things shaped up

The fix was simple: a clear type scale that sets the order in which you take in the page. We wrote detailed specs for each level of the scale and built clickable HTML mockups of the dashboard after the fix. Cam builds with a team of AI agents, so we made everything ready for them to use: the specs, the mockups, our walkthrough with Cam, and a detailed prompt he could hand straight to his lead AI developer.

The review didn’t end with a laundry list of everything wrong. It ended with the few things that mattered, and the tools to fix them without us in the room. That’s the point. Cam doesn’t need a software team looking over his shoulder. He needs someone who has seen enough projects to know which problems are cosmetic, which are structural, and which will quietly stall growth a year from now. In his case, the structure held. The interface just needed to catch up to it.

Experience still matters

We’re seeing this pattern more and more. Founders with deep industry expertise are building real software with AI, and they’re building it better than most people would expect. What they’re missing isn’t talent, just a second set of experienced eyes before real users, real data, and real stakes arrive. Aspenglow’s founders, Steven and Anthony, have over 20 years of experience building software. That experience is where our taste and opinions come from, and it’s why we can spot what’s wrong quickly, even when it just feels off.