Copilot Money
a money app that tells you what changed
Manifesto: tell me what changed
Copilot Money · a NUUKE concept · sample data
A bank statement tells you what happened. Four hundred lines, all the same size. Nobody opens a money app to reread the past. You open it to ask one thing: what changed? Dining crept up. A subscription appeared. Savings grew. Let the app do the sorting. You read the one line that matters.
01 · The brief
A money app that tells you what changed.
Copilot Money is a New York subscription app built by founder and CEO Andrés Ugarte, a software engineer who quit his job in 2018 to make it; it launched publicly around 2020. It brings bank, card and investment accounts from 10,000+ US institutions into one view, with ML categorisation, budgets with rollovers, recurring tracking and net worth. The model is subscription-only: “no ads, no gimmicks” and no data selling. The day Mint’s shutdown was announced, in November 2023, was Copilot’s biggest day ever.
- 100K+
- Subscribers, as of March 2024
- $6M
- Series A led by Adjacent, March 2024
- $10.5M
- Total venture capital raised
- ADA ’24
- Apple Design Awards finalist, Innovation
Design problems we chose
- 01
Trust at the connect step
Linking a bank is the scariest moment in any finance app. The copy, motion and security cues have to answer every fear before it becomes a bounce.
- 02
Charts that answer questions
Most money apps show what happened. People open them to learn what changed. Every chart has to earn its place by answering one question.
- 03
Review as a ritual
Checking new transactions should be a satisfying two-minute loop, and every correction should teach the model something.
Our goal
Every screen answers “what changed?” before “what happened?”, and the scariest step, linking a bank, feels like the safest one.
How the app gets made · our six-week sprint
- 01Week 1DiscoverPersonas, a month-long journey map and the questions people really ask a money app.Personas · journey
- 02Week 2DefineTen screens and one daily loop. The flow decides what the dashboard must answer first.User flow
- 03Week 2–3TrustThe scariest step, linking a bank, designed moment by moment with annotated copy.Onboarding
- 04Week 3VisualiseOne sample dataset, five chart types, and a written reason why each question gets its chart.Data-viz lab
- 05Week 4–5PrototypeLayers, wireframes and a tappable prototype we would put in front of ten users.Prototype
- 06Week 6Hand offTokens, specs and SwiftUI, so what ships matches what was designed.Tokens · code
02 · Research
Three people, one question: where did it go?
We built personas from Copilot’s public audience (design-minded, Apple-first US professionals, many of them former Mint users) and tested every screen against them.
Personas are NUUKE research assumptions built from the brand’s public audience, not real customers.
03 · Journey map
A month with money, and where it hurts.
Tap each moment to see what Priya is doing, how she feels, and the design opportunity it gave us. The two lows (mid-month and a surprise renewal) are both things that changed without anyone saying so.
Stage 1 · Payday
Doing
Paycheck lands; checks the balance.
Feeling
🙂 Relieved, briefly.
Opportunity
Show what’s already spoken for (rent, recurring) so “available” is honest.
Stage 2 · Week one
Doing
Taps a card twenty times; no idea where it’s going.
Feeling
😐 Vague, a little uneasy.
Opportunity
A daily two-minute review inbox: confirm, re-categorise, done.
Stage 3 · Mid-month
Doing
Finds out dining is running hot, by accident.
Feeling
😣 Anxious, a bit guilty.
Opportunity
A proactive “what changed” card: Dining is $90 ahead of last month, with 13 days left.
Stage 4 · A renewal
Doing
A free trial quietly turns into a subscription.
Feeling
😕 Annoyed.
Opportunity
Flag “new subscription spotted” after the first charge, before the second.
Stage 5 · Month end
Doing
Reviews budgets; rolls unspent grocery money forward.
Feeling
🙂 In control.
Opportunity
Rollovers make under-spending feel like progress, not a reset.
Stage 6 · Next month
Doing
Sees net worth tick up 1.8%.
Feeling
😄 Motivated.
Opportunity
Close every month with one line: what changed, and what to watch next.
04 · User flow
Ten screens. One loop: review, understand, adjust.
Onboarding happens once; the daily loop is dashboard → review → what changed. Tap any screen to jump to it in the live prototype.
Tap any screen to open it in the prototype ↓
05 · Trust-first onboarding
Six moments between “hello” and “here’s what changed”.
We designed onboarding around the moment people are most likely to leave: handing a bank login to an app. Step through it and tap the numbered pins to see what we show at each moment and why.
Welcome
Your money, beautifully organized.
Bring every account into one calm view.
- No ads
- No data selling
- Support from real people
1 month free, then $95/year
Step 1 of 3
Find your bank
US accounts only · also Venmo, Apple Card and Coinbase
Bank names are invented for this concept.
Hudson Federal
Here’s what happens next
- 1You sign in with your bank on a secure connection screen.
- 2Your balances and transactions come back to Copilot.
- 3You choose which accounts to include.
We never sell your data or show you ads.
Step 3 of 3
Choose what to include
You can change this later in Settings.
- Checking•• 4821 · $6,240
- Savings•• 0937 · $22,180
- Credit card•• 1180 · −$1,420
- Side-business checking•• 5520 · $3,905
Last four digits only. Sample balances.
Setting up
Sorting your history
1,284 of 2,314 transactions
You can leave. We’ll let you know when it’s ready.
You’re all set
Here’s what changed
Last 3 months
Dining grew fastest
Up 44% since June, mostly weekend dinners.
Moment 1 of 6 · The promise
Say the deal before asking for anything.
What we show
No ads, no data selling, support from real people, as the first thing you read.
Why
Copilot’s early users asked for exactly this. One request, quoted by Apple: “Please don’t sell my data or put in ads. I’m willing to pay a few bucks for this service.” We answer it before it’s asked.
What we show
The trial and the yearly price, before a single account is linked.
Why
A paywall after someone has handed over bank access feels like a trap. Price first makes linking a choice, not an ambush.
Moment 2 of 6 · Find your bank
Search first, with the real number.
What we show
A search field that says “10,000+ institutions”, with popular banks as text tiles below.
Why
People know their bank’s name; scanning a wall of logos is slower. The number quietly says “yours is almost certainly here”.
What we show
“US accounts only” in plain words, right under the search.
Why
A visitor outside the US should find out at step two, not after creating an account and getting their hopes up.
Moment 3 of 6 · Before you sign in
Preview the hand-off before it happens.
What we show
Three numbered steps: sign in with your bank on a secure connection screen, balances and transactions come back, you pick the accounts.
Why
The jump to a bank’s login is where people bail. Previewed, the redirect becomes a step they were told about rather than a surprise.
Exact wording depends on Copilot’s connection partner; we’d confirm it with their engineering team before shipping.
What we show
The promise again, repeated at the scariest moment.
Why
Reassurance works where the fear is. Saying it once on a welcome screen isn’t enough.
Moment 4 of 6 · Choose accounts
Every account is opt-in, visible and reversible.
What we show
A switch per account, defaulting on, with a line saying you can change this later.
Why
Control lowers anxiety. Couples like Sam & Alex may want the joint account in and the side-business account out.
“Change this later” assumes accounts can be hidden after linking; to confirm.
What we show
Only the last four digits of each account.
Why
Enough to recognise the account, nothing someone looking over your shoulder could use.
Moment 5 of 6 · First sync
Show the work, not a spinner.
What we show
A live count of transactions being read, with categories filling up as they’re sorted.
Why
Watching categories fill builds trust in auto-categorisation before you have to rely on it. A silent spinner here feels like a black box.
What we show
“You can leave. We’ll let you know when it’s ready.”
Why
Nobody should have to watch a sync. A notification brings them back for the payoff.
Moment 6 of 6 · First insight
Pay off the scary step within a minute.
What we show
A first “what changed” card instead of an empty dashboard.
Why
The reward for linking a bank is an insight. It also delivers on the whole concept: tell me what changed.
What we show
“12 transactions we weren’t sure about”, with a button to review them.
Why
Admitting low-confidence guesses is more trustworthy than pretending the model is perfect, and it starts the review habit on day one.
Expected anxiety at each moment · our hypothesis
Peaks at the bank hand-off. That’s where we spent the most words.
06 · Data-visualisation lab
Same data, five charts, one right answer per question.
Before drawing a single dashboard, we put one sample month through every chart type and asked which one answers each question fastest. Pick a question, try each chart, and read our reasoning. The star marks our pick.
1 · Pick a question
2 · Try a chart
- Rent
- Groceries
- Dining
- Shopping
- Transport
- Other
- Saved / income
07 · 3D wireframe
Take the dashboard apart, layer by layer.
From grid to motion, every screen is built in five layers. Scroll and the exploded view assembles into the final UI. Drag to look around it.
Scroll to assemble · drag to orbit
- L1
Layout grid
4-column grid, 20pt margins and an 8pt rhythm. Safe areas for the Dynamic Island and the tab bar.
- L2
Wireframe
Hierarchy first: the month’s one number, a pace chart, then “what changed” before any list of transactions.
- L3
Content & data
Real copy and sample data: spent so far, the budget, and three plain-English changes since last month.
- L4
Colour & charts
Navy canvas, green for money in, red-orange for money out, blue for the line that matters. Category colours carry identity.
- L5
Motion & states
The pace line draws in when data syncs, the “now” dot breathes, and a change card nudges once, never twice.
08 · Prototype
Tap through it. Then switch to the wireframe.
This is the prototype we’d put in front of ten Copilot-style users. Follow the journey on the left and flip between Sketch, Wireframe and Hi-fi to see how each screen evolved. All amounts are sample data.
Interactive prototype
Copilot daily loop
Eight tasks, ten screens. Everything that glows is tappable.
Final UI, fully tappable.
Your journey 0/8
Get started
Connect your first account
Bank names here are invented for this concept.
October · spent so far
$3,182.40
of $4,600 budget · 13 days left
This month Last month
What changed
To review
Four new since yesterday. Tap ✓ to confirm, tap a category to change it.
Lucia’s Trattoria · $48.20
Choose a category
Future transactions only
Auto-categorisation
Why we chose 🍜 Dining
How sure we were
Very likely · 92%
What we looked at
You put Lucia’s in Dining 14 times before.
Listed by the bank as a restaurant.
Friday, 8:40 pm, $48: looks like dinner.
When you change it
We learn from your choice for this merchant. Past months stay as they are unless you ask.
Concept: the explanation pattern is ours. Signals shown are illustrative.
Spending · October
So far, against each budget.
Dining
$388 of $400 · 13 days leftWhat changed
$90 ahead of September by this date. Three more dinners out, all on Fridays.
MayJunJulAugSepOct
Top merchants
3 visits$142
4 visits$96
5 visits$71
Category colour
Budgets · October
$3,182 of $4,600 · on pace
Recurring
Next 14 days · $1,958.97
New subscription spotted
🎬 StreamBox · $15.99 / month
First charge Oct 16. Next expected Nov 16.
Oct 31$1,900.00
Oct 24$35.00
Oct 26$23.97
Nov 2$10.99
Net worth
$148,600
▲ $3,480 (2.4%) this month
- Stocks 60%
- Bonds 20%
- Crypto 9%
- Cash 11%
Sample data for this concept.
…
09 · Spatial sketch
Net worth you can walk around.
Copilot already runs on iPhone, iPad, Mac, the web and Apple Vision. We sketched what net worth could feel like in space: every account a stack of coins, debt hanging below the line. Drag to orbit, scrub or play the year, tap an account to focus it.
Net worth · Dec sample
$150k
▲ $44k since January
Accounts · tap to focus
10 · Design-to-dev handoff
From token to SwiftUI, with nothing lost in between.
Founder Andrés Ugarte has said a native Swift app “makes a difference the moment you start interacting with it.” So our handoff speaks Swift. Change a token and watch the component and its code update together; turn on the redline to see the spec.
Live component · BudgetRow
Over by $120 · $35 rolled over
Sample data. The same component in the prototype’s Budgets screen.
| Token | Value | Adjust |
|---|---|---|
color.accent | #FF4433 | |
radius.card | 18pt | |
space.inset | 16pt | |
space.gap | 12pt | |
type.amount.size | 20pt | |
type.amount.weight | bold |
// BudgetRow.swift · generated from design tokens
import SwiftUI
extension Color {
static let cmSurface = Color(hex: 0x001533)
static let cmAccent = Color(hex: 0xFF4433) // color.spend
}
enum Token {
static let radius: CGFloat = 18
static let inset: CGFloat = 16
static let gap: CGFloat = 12
static let amount: CGFloat = 20
}
struct BudgetRow: View {
let category: Category
var body: some View {
HStack(spacing: Token.gap) {
Text(category.emoji)
.font(.system(size: 22))
.frame(width: 44, height: 44)
.background(Color.cmAccent.opacity(0.18),
in: RoundedRectangle(cornerRadius: Token.radius * 0.6))
VStack(alignment: .leading, spacing: 6) {
HStack(alignment: .firstTextBaseline) {
Text(category.name).font(.headline)
Spacer()
Text(category.spent, format: .currency(code: "USD"))
.font(.system(size: Token.amount, weight: .bold, design: .rounded))
.monospacedDigit()
}
ProgressView(value: min(category.progress, 1))
.tint(.cmAccent)
Text(category.status)
.font(.footnote)
.foregroundStyle(.secondary)
}
}
.padding(Token.inset)
.background(Color.cmSurface,
in: RoundedRectangle(cornerRadius: Token.radius, style: .continuous))
}
}{
"color": {
"surface": { "value": "#001533" },
"accent": { "value": "#FF4433", "role": "spend" }
},
"radius": { "card": { "value": 18 } },
"space": {
"inset": { "value": 16 },
"gap": { "value": 12 }
},
"type": {
"amount": { "size": 20, "weight": "bold", "design": "rounded" }
}
}Illustrative SwiftUI for the concept. Color(hex:) is a small helper we’d ship with the tokens; Copilot’s real codebase isn’t public.
11 · Design system
Deep navy calm, with money in colour.
Colour tokens · click to copy
Type scale
- Display / numbers$3,182.40SF Pro Rounded 800 · 44/48 · tabular
- TitleWhat changedSF Pro 700 · 24/28
- BodyDining is up $90 vs. September.SF Pro 400 · 15/22
- LabelOCTOBER · SPENT SO FARSF Pro 600 · 11/14 · +0.12em
Components
vs. September+$90“What changed” card · one change, one number
Hex values are our working approximation, based on a third-party analysis of Copilot’s UI, not official brand values. Copilot doesn’t publish its typography; SF is our inference (Inter stands in on non-Apple screens). Chart colours for categories were validated for colour-blind separation on the navy surface.
12 · Measure
How we’d know it worked.
Connect completion
We expect: More people who start linking a bank finish it, because every scary moment has an answer on screen.
How we’d measure: Step-by-step funnel from “Connect” to first successful sync, A/B against today’s flow.
Time to first insight
We expect: New users see their first “what changed” card within three minutes of opening the app.
How we’d measure: Install → first insight card view, median and 90th percentile.
Review as a ritual
We expect: Most days, the review inbox is cleared in under two minutes.
How we’d measure: Review session length and clear rate, by week of use.
Chart comprehension
We expect: People answer “what changed this month?” correctly in one glance more often with our chosen chart than with the alternatives.
How we’d measure: Five-second tests per chart type, same sample dataset, 30 people each.
Learning categoriser
We expect: Manual re-categorisations per user fall over the first 30 days as the model learns from each fix.
How we’d measure: Re-categorisation rate by weekly cohort.
These are the success measures we would agree with the client before launch. As a concept project, no results are claimed.
Sources for brand facts
- Copilot Money: homepage ↗
- Copilot Money: FAQ ↗
- Copilot Money: Series A post ↗
- Copilot Money: web app launch ↗
- TechCrunch: $6M Series A ↗
- Crowdfund Insider: Series A ↗
- Apple Developer: Copilot spotlight ↗
- Apple Design Awards 2024 (finalists) ↗
- App Store listing ↗
- Copilot help centre: settings ↗
- Blake Crosley: third-party colour analysis ↗
