home screen.
The Challenge
Fitness apps make you work before you work out. Rigid databases, toxic streaks, and three screens just to start a session.
Team & Time
Dec 2025 to May 2026. Me as the first product designer, Lisa (CEO), Darren (creative director) and one developer.
What I Did
100+ screens, an atomic design system, a branding pivot, and features invented from real research, built up from ~20 inherited wireframes.
What Changed
Starting a workout went from 3 taps to 1. Sport Mode, which no app we reviewed had. No streak to break: a consistency score and badges instead.
How do you want to read it?
Fitness apps make you work before you work out.
Lisa McPhearson, Ambrisa’s CEO, is a competitive gym athlete. She had tried every major fitness app, and every one got in the way of the one thing people actually want to do: move.
Lisa had cycled through every major fitness app on the market. Her frustration was not about missing features. The name says what she wanted instead: Ambrisa comes from ambrosia, the drink of the Greek gods.
After guerrilla research with real gym-goers between sets, the same theme kept coming up across every conversation. Different bodies, different goals, same complaint.
FIELD NOTES · GYM FLOOR RESEARCH · N=12
“I do not want to think before entering the gym. I just want to start.”
P2 · regular gym-goer
“Let's say I played badminton today. It should just know what muscles I worked.”
P1 · mixed athlete
“Why do I have to go through three screens just to begin?”
P4 · casual lifter
The problem was not a missing feature. It was friction. Gymverse, Fitbod, Evolve You and MyFitnessPal all needed two to three screens before you could begin. For someone walking into a gym ready to move, that is three screens too many.
The brief was to design a fitness app. The job was to get out of the way.
Two sets of goals. Both had to be true.
We wrote down what success meant for the people using the app and for the business building it. Every feature idea had to pass both lists.
Try it · run an idea through both lists
If it failed either list, it didn’t ship.
For the user
- Start a workout in one tap
- Log any activity, not just gym exercises
- Track nutrition without manual data entry
- Feel supported, not judged or punished
- Use AI that feels like a coach, not a robot
For the business
- Ship a validated, investor-ready MVP fast
- Move quickly into user testing phase
- Build a design system that scales without chaos
- Differentiate clearly from existing fitness apps
- Drive organic user signups post-launch
Cut. A broken streak punishes people.
Every feature got filtered through both lists. If it failed either, it didn't ship.
Every section needed its own identity.
The original direction was dark everywhere: red and orange on black across Workout, Nutrition and Progress. It looked strong, but three sections with three different jobs all felt identical.
Flip it
One mood everywhere, or three moods with one brand.
After a collaborative review with Lisa and Creative Director Darren, we identified the core problem. All three sections serve completely different emotional purposes, but they felt identical.
Workout is about energy and focus. Nutrition is about nourishment and care. Progress is about reflection and data. One palette could not carry all three, so each section got its own, on a shared brand token layer.
Four decisions that shaped everything.
If you only read one section, read this one. Each decision has a short decision log underneath: the context, what I chose, what it cost, and how I’ll know it worked.
From 3 taps to 1.
THE CONSTRAINT
Lisa set one rule: after onboarding, a workout starts in one tap. No preview screen, no edit screen. Everything else had to fit around it.
Try it · start both workouts
Three screens before a single rep, or one tap.
Preview
WORKOUT
FULL THROTTLE PUSH
A sketch of the two flows from this case study. The taps and seconds are yours, not user data.
HOW I MADE ONE TAP WORK
The original flow was preview, then edit, then workout. I folded all three into one action-ready home: your plan, your options and your start button, all visible at once. The options that used to need their own screens moved into the home itself.
Tap the numbers · the real home screen
Everything the old preview and edit screens did, on one screen.
Decision log
Sport Mode, the feature nobody had.
WHERE IT CAME FROM
One quote in user research stopped everything.
Let’s say I played badminton today. It should just know what muscles I worked.
Try it · pick today’s sport
Log any activity. It knows what you worked.
Muscles worked
Muscle pairings are the ones on the Sport screen. 8 apps reviewed; none mapped a sport to muscles.
WHAT WE BUILT
Sport Mode lets you log any sport and maps it to the muscles worked, so casual activity becomes real training data. One conversation with a real user became a core product feature.
Not one of the eight apps we analysed offered this. Not Gymverse. Not Fitbod. Not MyFitnessPal. The four biggest are below.
8 competitors. 0 had it.
| COMPANY | SPORT-TO-MUSCLE MAPPING | ONE-TAP START | AI-NATIVE LOGGING |
|---|---|---|---|
| Gymverse | |||
| Fitbod | |||
| Evolve You | |||
| MyFitnessPal | |||
| Ambrisa |
COMPETITIVE SCAN · 8 APPS REVIEWED
Decision log
AI as infrastructure, not a feature.
THE PRINCIPLE
Most fitness apps bolt on an AI chat and call it smart. We designed AI into the fabric of the product, so logging food feels like talking to a friend, not filling in a form.
Watch it · the AI Coach
Say it the way you’d tell a friend.
A sketch of the Recipe Builder flow from the screens below. No macros are shown here on purpose.
Click through · the real screens
From three photos to one logged meal.
AI decisions
Where the AI acts, and where people stay in charge.
You confirm before anything is logged.
The AI names the ingredients from a photo or a sentence and asks if they make one meal. You set the portions, and only then are the macros calculated. “You can review before saving.”
Seen in · AI Coach · Recipe Builder · Custom MealYour own history before a guess.
The Food Library starts from meals you’ve already photographed. Tap to log one again, or use it as the base for a new entry.
Seen in · Food LibraryAutomation asks before it changes your plan.
Log a sport and Ambrisa asks whether to add it to today’s workout or replace it. It never rewrites your plan on its own.
Seen in · Sport ModeA coach, not a judge.
No streak counter to break and no guilt. Progress shows up as a consistency score and badges you keep.
Seen in · Progress & AchievementsClick through · log it without typing
Your own history comes before a guess.
The unhappy path · when search finds nothing
A dead end became a way forward.
The same fallback is in workout search, so an empty result never ends the flow.
THE FOOD LIBRARY
People already photograph their meals. Inside the AI Coach, the Food Library opens a personal archive of every meal photo the user has ever uploaded, each entry auto-tagged with macros. Instead of re-logging the same chicken and rice bowl for the tenth time, the user pulls it from their library and taps.
The insight: logging food is not hard because people forget what they ate. It is hard because the apps make them prove it from scratch every time.
THE RECIPE BUILDER
The way people talk about food with friends is nothing like how apps ask you to log it. Friends say ‘i made pasta with tomato sauce and some leftover chicken.’ Apps say ‘select item. enter serving size. add macro.’ The Recipe Builder lives inside the AI Coach. Users describe a meal in conversation. The AI builds the recipe, the user checks the portions, and the macros are calculated and saved.
THE FULL AI SURFACE ACROSS THE PRODUCT
- Workout search: understands natural language, not just keywords
- AI Coach chat: conversational interface for food logging
- Food Library: AI-powered discovery replacing static database search
- Recipe Builder: build and log meals through conversation
- Progress Insights: AI surfaces patterns the user has not noticed yet
No streak to break. Progress you keep.
WHY I SAID NO
Most fitness apps run on a daily streak. Miss one day, the count drops to zero, and the app makes you feel it. Our own user goals said the opposite: feel supported, not judged or punished.
WHAT I DESIGNED INSTEAD
Two things. A Consistency Tracker: one score out of 100, made of streak, volume and effort, so showing up still counts but one missed day is a dent in one part, not a reset to zero. And achievement badges that unlock at training milestones, the way games like Pokémon Go do it: do 10 squats for silver, 20 for gold. A badge you’ve earned stays earned.
Click through · the real screens
A score that bends, and badges that stay.
Try it · log some squats
Badges only go one way: up.
Decision log
Four decisions, each one a conversation that almost didn't happen.
Community, kept on purpose simple.
WHY IT EXISTS
A place where people can share their workouts and meals if they want to, and others can copy them, repost them or save them for later. A small community of fitness people.
WHAT YOU CAN DO
Post, share a workout or a meal, like, save, repost and get tagged. That’s it. Think of the very first Instagram. It’s still in the works, and keeping it this small is the point.
Click through · the real screens
Workouts and meals, shared like posts.
The hardest screen in fitness is the one you use mid-set.
Sixty seconds between rounds. Sweaty hands. A timer running. The workout screen has to be readable in a glance, logable in a tap, and never make you think.
Watch it · minute 14 of a 45-minute session
One flow that moves itself forward.
- 1Warm upMoves on by itself when the timer ends
- 2Main setLog your set with one tap
- 3RestCounts down so you don’t have to
- 4CompleteCelebrates the work you actually did
When most fitness apps design the “exercise screen,” they design it for the planning moment, the one where you're calm, looking at your phone on the couch, deciding what to do tomorrow. But that's not where the screen actually lives. It lives at minute 14 of a 45-minute session, between sets, when you're catching your breath and trying to remember whether you did 90 or 95 last time.
So I designed the whole workout-in-progress experience as one continuous state machine: warm-up auto-advances into the main set, the main set into rest, rest into the next exercise. No menus. No back buttons hunting for what's next. The app holds the structure so you don't have to.
“You shouldn't have to think about what's next. You should only have to think about the rep.”
Designed for every body that's ever been to a gym.
Nine exercise variations. One design system. Because not every workout looks like a workout.
Try it · pick a movement
The logging screen takes the shape of the exercise.
Hold · time only
Plank
A plank isn’t reps. It’s a timer.
A pull-up is reps. A plank is time. A treadmill run is distance and pace. A dumbbell curl is reps × weight, but only on one side, then the other. A superset is two exercises stitched together with no rest. A bodyweight squat is just… a squat.
Most fitness apps pretend exercise is one shape, sets × reps × weight, and force every workout into that mold. Real training doesn't work that way. So I designed nine variations of the exercise logging UI, each one matching the actual structure of the movement it captures. Same design system. Same logging grammar. Nine different forms.
“The hardest part of a design system isn't the components. It's the edge cases you didn't want to design.”
100+ screens. One designer. One system.
I inherited approximately 20 wireframes from a previous UI designer. No documented system, no component library, no clear rationale. I built the entire design system from scratch, alone.
The system is atomic: tokens feed components, components feed templates, templates feed screens. Every spacing value, colour and variant is a Figma variable.
This project is also where I truly mastered auto-layout, building complex responsive components that hold their structure across every screen size and state.
the full component library, organized by atomic level.
SYSTEM HIGHLIGHTS
- Component library: buttons, cards, inputs, nav, modals, overlays, all variant-driven
- Color variables: per-section theming with a shared brand token layer
- Typography system: scale, weight, and line-height defined for every use case
- Weekly engineering handoffs: every component annotated and export-ready
The system wasn't asked for. I built it because the product needed it to survive.
Four layers per button. One developer.
Ambrisa is pre-launch and in active build, with a single developer. That made every visual decision a schedule decision too.
Darren, our creative director, cared about every detail of the UI. Our buttons weren’t one layer with a shadow. They were two: a convex top over a concave base, with a small gap between, and a set of shadows underneath. Our developer pushed back on the time it would take. Lisa and Darren chose quality over speed, knowing it would take longer to build.
Take it apart
One button, four layers.
- Label sits on the top layer
- Convex top catches the light
- A small gap, then a concave base
- A set of shadows under it all
It taught me to treat a visual decision as a build decision. Built once as a component, a four-layer button is built once by engineering. It’s a big part of why, next time, I’d set up the system before the first screen.
- Inherited ~20 wireframesNo system, no library, no rationale
- Branding pivotThree sections, three moods
- System + 100+ screensAtomic, variable-driven
- In buildWith one developer
DESIGNED · IN ACTIVE BUILD
What I designed
- Core workout flow: one-tap start, plan view, exercise cards
- Exercise logging with all 9 variation types
- Warm-up and cool-down auto-advance sequences
- Full atomic design system: tokens, components, templates
- Branding system with per-section color theming (Workout / Nutrition / Progress)
- Navigation architecture and information hierarchy
- Sport Mode: sport-to-muscle mapping engine
- AI Coach: conversational chat interface
- Recipe Builder: meal logging through conversation
- Food Library: photo-based meal archive with macro tagging
- Progress Insights: AI-surfaced training patterns
- Community features and social sharing
I can tell you exactly why each piece exists, and how I'd measure its impact.
What I shipped, what I learned.
The product is still in development. The foundation is solid, the system is scalable, and every decision has a reason behind it.
100+
SCREENS / DESIGNED
9
LOGGING / VARIATIONS
8
COMPETITORS / REVIEWED
6
MONTHS, / DEC → MAY
Process receipts — outputs, not measured outcomes
How I’ll know it worked
Ambrisa hasn’t launched, so there’s no usage data yet. This is the plan, set before launch.
Decision log
What I’d do differently
Build the design system before the first screen.
Ambrisa moved fast with little structure, so the system grew up alongside the UI. It works, but it isn’t where I’d want it, and changing it now ripples through every screen. Next time: tokens and components first, screens second.
This was the project where I stopped being a designer who executes and became one who shapes. I pushed back on the original branding direction with a reason. I invented a feature, Sport Mode, because I was listening closely enough to hear what was missing. I built a design system not because I was asked to, but because I knew the product needed one to survive.
The biggest technical growth was auto-layout. I went from knowing it to owning it, building components complex enough that engineers could hand them off directly without rework.
The foundation is solid. Every decision has a reason behind it.