CEO XB
An AI-assisted peer-accountability platform built and delivered for European life-sciences CEOs entering the US market.

Bottom line up front (BLUF) · 1 minBottom line up front
- CEO XB runs private peer cohorts for European life-sciences CEOs taking their companies into the US. Each cohort meets once a month.
- The previous version of its app was built around an eight-module online course, but members stay for their peers.
- Between meetings the app gave CEOs no reason to come back. The only pull was the two co-founders reaching out to each member by hand.
- That does not scale, and once a member drifts away, they rarely return. For a paid membership, every quiet month puts a renewal at risk.
- A live mobile app centered on the cohort
- A recommended peer on the home screen
- One-on-one coffee chats
- A monthly business update that AI helps each member write
- My role
- Product Designer & Sole Developer
- Timeline
- Spring and Summer 2026, about 20 weeks
- Team
- Four designers, working directly with CEO XB’s co-founders
- Owned the monthly update’s design and the entire build of the live mobile app
- Designed the AI drafting feature and the update flow it lives in, top to bottom
- Built the app end to end with Claude Code, from the React front end through the Anthropic API integration to the Vercel deployment, along with all six research prototypes
- Worked alongside the team on the rest of the design, from concept testing through usability testing
- 01Held every feature to one test: does it create a path to a 1:1 conversation?From concept testingThe appetite for 1:1 is real, but it needs a hook.
Two of our six concepts created that hook, so they carried into the product and the other four were cut.
- 02Made the AI an editor, so members stay the authors of their own updates75%of usability-study participants wanted to write their own update and have AI improve it
Members write first, and the AI improves what they wrote, drawing on the updates they posted in past months.
- 03Rebuilt a design system the client actually owns1,009elements in the inherited 2025 Figma file tied to a library the product didn’t own
Claude found them by reading the Figma file directly. I helped rebuild the system to match the code: 13 core components, 40 screens.
- Live, and handed off to CEO XB’s co-founders
- In usability testing, participants booked their first coffee chat without help, and each of the four problems the study surfaced changed the design
The cohort platform that turns a monthly business update into a conversation with a peer who can help
- A new member finishes onboarding and books their first coffee chat with no help
- Logging a monthly business update takes five minutes and feels like a win
- The client ends the project knowing what keeps a CEO coming back to the platform, proven in a working product
From the first screen to a coffee chat on a peer’s calendar
The monthly update, started from last month’s focus, then seen by the cohort
Everything above is deployed. Try the build yourself.
- Without CEO XB
You have to work out what comes first: regulatory approval, reimbursement, hiring, or raising money in the US
With CEO XBYour cohort has already worked through the same regulatory, reimbursement, and hiring questions
Nobody at your company has taken a product into the US, and your network is back in Europe
10 to 12 CEOs are taking their own products into the US right now, and some are ahead of you
You say what you will tackle next quarter, and nobody follows up on whether you did or not
You post a monthly business update, and the whole cohort sees what you committed to
When you get stuck, nobody around you has been stuck on the same thing
A peer who has made the same call books a coffee chat and tells you how it went
Introducing CEO XB.
A walkthrough of the live product. Julie has her second cohort meeting coming up, so she drafts her monthly business update, reads what her peers moved on since last month, and books a coffee chat with the one facing the same problem she is.
What brings a CEO back is the other CEOs
We tested two bets about what pulls a member back between sessions: accountability and networking. Each bet got three working prototypes, six in all. Around them, we ran stakeholder interviews with the co-founders, a literature review on what sustains peer communities, a concept test in Optimal Workshop, and a usability study in Maze.
Two carried into the product. Warm Welcome became the coffee chat that onboarding now ends on. The Streak became the monthly business update, slowed from weekly to monthly at the client’s request. The Leaderboard, Up Next, Circle Back, and The Bench told us what not to build.
The Streak: tests whether a small repeated action brings a CEO back to the platform Warm Welcome: tests whether one introduction at onboarding becomes a habit of reaching out
The founders’ facilitation is part of the product, so we designed around it
A stakeholder crit caught where our concepts were heading. CEO XB’s two founders select each cohort, make the introductions, and hold the room. Our early passes had quietly designed that job out, treating the app as the thing that connects people.
So we sorted the program by what a screen can actually carry. The platform owns the month between sessions: visible updates, scheduling, and the event calendar. It gives the founders somewhere to point members, and it stays out of the facilitation itself. The matching and the sessions remain the founders’ work, because that is what members joined for.
We built one journey end to end, then fixed what people stumbled on
A workshop decided every feature’s fate, and the cohort page became home
I designed the stakeholder workshop that put our surviving concepts next to the 2025 build and forced a call on every feature: kept, demoted, or retired, so nothing disappeared silently. The client’s number-one ask was that the cohort page become the home screen.
Narrow to one journey, then build it
We had inherited a working five-tab app from 2025 and were adding a cohort feed, a monthly update flow, and coffee chat scheduling on top of it. It would not all fit in one quarter, so we chose one path and finished it completely: onboarding into a cohort, the monthly business update, and the coffee chat it leads to. The home screen alone took three passes.
A moderated study found four problems, and each one changed the design
With the journey built and deployed, we ran a moderated usability study through Maze. We framed it around the client’s own test of a good update: “easy, intuitive, not a chore.” Participants booked their first coffee chat with no help, which met our first success criterion. Four things still came back:
- 38%entered the coffee chat flow through something else first“Honestly, that doesn’t feel like a button. It just feels like user info.”
An affordance failure, fixed at the component level.
- 63%called out a noisy home screen“There’s too many things that all need my attention. What do I do first?”
The layout’s hierarchy was fine. It just made the wrong thing most important.
- 63%did not understand the label “business move”Some read it as physically relocating the company.
The study’s lowest ease-of-use rating traced back to that one label, so “business move” became the monthly update.
- 75%wanted to write their own update and have AI improve it“I expected draft for me to gather data from across my account.”
I moved the AI from ghostwriter to editor, and taught it to read their history.
The study changed two screens: a quieter home screen, and an update that starts with the member’s own words
Drag the divider on either screen.
A design system CEO XB owns, and forty screens built from it
The 2025 build ran on GenSaas, an open-source library the product did not own, and 1,009 instances still pointed at it. I helped rebuild the system so CEO XB owns every token and component, and the forty final screens are built from it.
- 40Final-Screen FramesFour flows, built to match the deployed code
- 269Live Instances13 core components carry all of them
- 1,009Orphaned InstancesLeft by the retired 2025 library, all replaced
Every color is named for the job it does
The 2025 file called its colors Primary 700 and Neutral 300, so using one meant remembering which number was which. Now each name says what the color is for. Type is SF Pro, because the product is built from Apple’s components.
- Brand/Midnight#062C50tint · active tab · actions
- Brand/Cerulean#0C4C6Fkickers · links
- Brand/Cyan 200#9AE4F1kicker line on navy
- Text/Primary#252C33titles · body
- iOS Label#6B6B73sections · subtitles
- Text/Muted#929DA8timestamps
- Border/Default#BEC8D3locked outlines
- Surface/White#FFFFFFcards · glass
- Surface/Tint#E0E8F1chips · cells
- Grouped BG#F2F4F8screen ground
- Surface/Disabled#F5F5F5locked
Midnight navy and the 2025 logo were fixed. Everything around them is new.
One test decides every button’s color: would an accidental tap matter?
Midnight is for committing: agreeing, saving, moving through a required flow. Cerulean is for inviting: lower-stakes actions that do not demand anything. The test: if a member tapped this by accident, would it matter? If yes, Midnight. If no, Cerulean.
Moving to iOS 26 was cheaper than patching the old navigation
The tab bar is where a scope decision shows. We had scoped the work to features and left the platform alone. Then, in detailed design, we found that new features kept landing on the old 2025 navigation bars. Moving the app to iOS 26 and its Liquid Glass materials was the cheaper fix. Apple supplies the bars, the blur, and the hit areas, so the new features and the older screens finally read as one native product.
An agent helped build everything here, and checking its work became part of the design job
Claude Code ran through every stage of this project: six concept prototypes for research, the inherited design system audited and rebuilt, the forty final screens, and the deployed app. It read the design system straight out of Figma through MCP, a connection that lets an agent read a design file directly.
I shipped the six concepts as working software
Research needed six ideas in front of people, so each one shipped as a working, deployed prototype built from a structured prompt. I also wrote an eleven-page runbook that took a designer from a blank chat to a shareable URL in about ninety minutes.
Screens that looked right were linked to nothing, and the canvas never showed it
A teammate noticed that the newest frames carried the right components and asked what had changed. Neither of us knew, so I had Claude read the file through MCP again and count real instances against plain frames on every page.
- Nav / Tab Bar
- Cohort / Feed Post
- Member / Card
- P - Bordered Prominent
- Nav / Tab Bar
- Cohort / Feed Post
- Member / Card
- P - Bordered Prominent
A purple diamond means the layer is linked to the library. Grey means it is not. On the canvas the two screens are identical.
The fix was one sentence in the prompt, and a check on every screen
An agent will not use the design system unless you tell it to, and it will not tell you when it didn’t. So the library goes in every prompt now, and every screen gets checked afterward.
Build these screens.
The agent invented layers and labeled them with names it had seen. Where it had no reference, it made a name up, which is why the versions drifted apart.
Build these screens from our design system library.
The agent searched the library and placed real instances. This is the version that ships.
What building real software taught me about design
- 01
Working this way moved the job from making to specifying and checking.
The agent produces the artifact. Whether it is any good comes down to how precisely I brief it and how carefully I check what comes back. Both of those are design work.
- 02
The deployed code is the source of truth.
When the code and the design file disagreed, the code won, and I updated the file to match. A handoff needs exactly one source of truth.
- 03
An AI feature earns trust by showing its sources.
Participants assumed the AI read their real data and trusted it less when it clearly hadn’t. So I gave it their real history, made it an editor of their words, and added a line to every draft naming the month it drew from.
What I would do differently
- 01
Recruit the actual audience sooner.
Our participants were professionals standing in for European life-sciences CEOs. The navigation and affordance findings transfer. The findings about the premise itself do not, and I would trade two weeks of polish for more sessions with real CEOs.
- 02
Test whether a posted update actually leads to a coffee chat.
That is the bet the whole product rests on. We built both halves and tested each one separately, but the step between them, reading a peer’s update and booking time with them, never faced a participant.
- 03
Turn on analytics before launch.
A deployed product can watch how people actually use it, and ours mostly did not. So two of our three success criteria, a five-minute update and knowing what brings a CEO back, are still unproven outside a study session. From day one, I would track how long an update takes to post, how many members return between meetings, and how often reading a peer’s update leads to a booked coffee chat.