I Gave Claude My Raw Notes for a Case Study. It Caught a Mistake I Missed in My Own Screenshots.

Real project, typo-filled notes, no cleanup. Here is what came back and what I had to fix.

Quick note before you dive in: every tool or feature I mention here, I’ll tell you straight up whether it’s free or paid as of today. I’ve lost count of how many times I’ve watched a video, gotten excited about some AI result, tried it myself, and only then found out the person was on a paid plan the whole time. Not doing that to you.

Writing case studies is the part designers put off the longest. The design is done, the client is happy, and then you have to explain two weeks of work in a way that makes sense to a total stranger. So it sits in a doc, half-written, for weeks.

I wanted to see if Claude could actually close that gap. So I ran a real test: I took a finished project, wrote my notes the messy way I actually think, typos and all, and handed it over with zero cleanup. What came back surprised me, and at one point it caught something in my own screenshots that I hadn’t noticed myself.

The project

A lash mapping app for a European beauty brand. I designed the whole thing from scratch, alone, in two weeks. You can see the full project on my Behance. The client uploads a face photo, AI scans her face and eye features, and the app suggests lash styles that suit her. She can adjust length, thickness, volume, color, and curl, then try the lashes on photos of her own eyes before ordering. After that she books a stylist for a date and time. There is a separate flow for stylists, who see incoming orders and respond to them.

What I gave Claude

I uploaded five screenshots from my Figma file and wrote my notes in one go. Here is a piece of it, exactly as I typed it:

“the idea is client uploads its face image. ai scans face and eye features. and show results to client. based on the eye and face features, app shows relevant lash style to client.”

And later:

“it is not a redesign. i built all from scratch. i do user research in Europe market what they like, ladies interest about beauty, etc. i just built the design as I was hired for and have no details about development.”

I also told it the research was general, pulled from Medium blogs and other sites, with no live interviews. I wanted to see if it would keep that limit honest or dress it up.

What Claude wrote first

The first draft came back with an overview, the problem, the audience, a step by step flow, design decisions, and research. Here is how it wrote the case study:

Lash extensions are a commitment. Once they are on, you can’t take them off and try a different style if they look wrong on your eyes. Most clients choose from a catalog or a stylist’s suggestion and hope for the best.

A European lash brand wanted to fix that, and I got to design the whole app from zero. I worked alone, and the project took two weeks.

The idea: The app lets a client see lashes on her own eyes before she orders them. She uploads a face photo, and AI scans her face and eyes, picking up eye shape, face shape, the distance between her eyes, and the distance between eyelid and eyebrow. The app shows lash styles that match those results. She adjusts length, thickness, volume, color, and curl, with curl options like D, C+, and J/A coming with an AI recommendation. She uploads photos of both eyes and sees the lashes applied, so she can compare styles on her own eye. When she has decided, she orders the style, picks a stylist, and books a date and time at the store.

A gallery keeps her uploaded photos and saved styles. The app also shows adhesive and temperature details, and the best conditions for applying lashes so they last longer.

Two sides of the app: After sign up, the app asks one question, are you a client or a stylist? The stylist side is built for daily work. A stylist opens the app and sees her client profiles, each with the client’s lash parameters, her face and eye details, and a comments screen where the client can write what she wants, a specific curl, layers, or volume. The stylist reviews all of it before the appointment and responds to the order.

Design decisions: Soft pink background, since the audience is women and the brand is beauty, a light pink feels familiar and welcoming. Marcellus typeface, which fits beauty brands well and gives the app a calm, elegant tone. Big type and lots of whitespace, so it works for young and older women alike. No shadows, keeping the interface flat and clean. Few screens, so each one has a single job. Plain labels, so every button and title says exactly what it does.

Research: I researched the European beauty market, what women like, what they care about in beauty, and what they expect from apps. This was general research from Medium articles and other websites, not live interviews. The client also shared some material, which I went through as well.

Result: The client was highly impressed with the final design. I delivered the Figma file along with a PDF of all screens. I handled the UI/UX only, so I don’t have details about development or whether the app has launched.

That’s close to how I’d say it myself, and it came out of a rambling note about lashes that can’t be removed. It kept my research limits exactly as I stated them, including the fact that I didn’t interview anyone. It didn’t invent details either, where it lacked information, it said so instead of guessing.

It also asked me questions: how long the project took, whether I worked alone, and whether the client gave feedback.

What it caught

This is the part I didn’t expect. Claude compared my claims to my screenshots and flagged two problems:

• I wrote that the design had no shadows, but the sign up fields in one screenshot had visible shadows
• One screen said Ms. Samantha Watson while another said Linda Josef

Checking back, the shadowed screenshot was from an older version I hadn’t swapped out, and the two names turned out to be leftover placeholder text from testing different client profiles, not something that should be sitting in final screens. Both are worth cleaning up in the Figma file before this project goes anywhere near a real portfolio or a client’s eyes. Small things, but exactly the kind of small things a stranger reviewing your portfolio would notice before you would.

The second draft

After I answered its questions (brand unnamed, solo, two weeks, stylist screens available, client highly impressed), I sent one more screenshot showing the stylist flow. The second draft added a section on the two sides of the app and described how a stylist reviews a client’s parameters, face details, and comments before responding to an order. It also became more specific about the timeline and the client’s reaction.

What I would change

A few of the design-decision lines read a little too polished for how I’d actually explain them out loud, “gives the app a calm, elegant tone” is the kind of line that sounds like marketing copy, not how I’d talk to another designer. I’d trim that down to something closer to “it just fits the beauty space better than a sans-serif would.” I’d also shorten the research section rather than let it stand on its own the way Claude wrote it, since the honest version is a couple of sentences, not a full paragraph, and stretching it out risks making general reading sound like more than it was.

My verdict

What worked:

• Turned messy notes into a clear structure
• Kept my limits honest, including the fact that I didn’t interview anyone
• Asked for missing details instead of making them up
• Caught a real inconsistency between my claims and my own screenshots

What to watch:

• It can only write what you tell it, so the case study is only as strong as your notes
• It can’t judge your screens unless you actually upload them
• You still have to check every claim against your own work, it caught my shadow and naming mix-up only because I gave it the screenshots to check against in the first place

If you’re a designer who already dreads writing up finished work, this is worth trying. If you’re expecting it to invent a polished story out of nothing, it won’t, and that’s the part that makes it actually useful instead of just another content generator.

If you want to try it yourself, give Claude the real problem, the real process, and the real limits, and let it ask questions. Then read every sentence and cut anything you wouldn’t say out loud.

I’m Usama, a UI/UX designer based in Islamabad. I write about what design tools actually do, not what they claim to do. If you’ve tried something like this yourself, I’d like to hear how it went, drop it in the comments. More of these coming.

I Gave Claude My Raw Notes for a Case Study. It Caught a Mistake I Missed in My Own Screenshots. was originally published in Bootcamp on Medium, where people are continuing the conversation by highlighting and responding to this story.