Project Type
Mobile App
Role
Lead Product Designer
Discipline
UX Research, AI Copilot Design, Product Strategy
Year
2026
FinPilot
A personal project built around one question: why do people who track their finances still feel anxious about them?

The Problem
Most finance apps solve the wrong problem. They help you track. Nobody needs more tracking.
I had used four personal finance apps over two years. All of them did one thing well: they showed me where my money went. Charts, categories, weekly summaries, the occasional guilt trip about coffee spending. Then they left me alone with that information and a mild sense of anxiety. None of them answered the question I actually had. Given all of this — what should I do next? The gap between financial awareness and financial action is exactly where anxiety lives. And none of the existing tools were designed to close it. They were designed to inform. Informing isn't the same as helping.

Research — 11 Conversations, 3 Patterns that wouldn't go away
I spoke with 11 people across different life stages and financial situations: early-career, mid-career, recently made redundant, freelance, dual-income. I deliberately included people who described themselves as 'financially anxious' and people who described themselves as 'financially confident' — because I wanted to understand what the gap between those two states actually looked like, not assume I already knew. The conversations weren't about budgeting tools. They were about trust, control, and the specific feeling that precedes a financial decision. Three patterns appeared consistently enough that I treated them as design constraints rather than just insights. The Avoider Mid-20s, irregular income. Opens the app, feels overwhelmed by the number of things to attend to, closes it. Says: 'Looking at it somehow makes it feel worse than not knowing.' The app's comprehensiveness is the problem — it surfaces too much and guides too little. The Tracker Late 20s, stable salary, logs every transaction meticulously. Has a complete picture of their finances. Still doesn't act on it. Tracking has become a substitute for deciding — the ritual of logging feels like doing something, even when nothing changes. The Reactive Early 30s, two income streams. Makes financial decisions impulsively when prompted by a specific event — a bill, a purchase, an unexpected expense — then regrets them and repeats the cycle. Knows they need a system. Hasn't found one they trust. Every person I spoke with had a version of the same sentence: 'I know I should do something. I just don't know what.' Tracking apps had been answering a question nobody was asking. The real question was always: given my situation right now, what's the right move?

The reframe that changed the shape of the product
The insight that restructured everything wasn't in the data — it was in a pattern across the data. Every person who described themselves as 'financially anxious' had accurate information about their situation. They knew their numbers. What they lacked was confidence in what the numbers meant and certainty about what to do next. The reframe: from retrospective reporting to forward-looking guidance. From 'here's what happened' to 'here's what to do.' That shift — from dashboard to advisor — became the core design principle. Everything else followed from it.
The goal wasn't to build a smarter dashboard. It was to eliminate the need for one.
The Solution - Replace the dashboard, Surface one decision at a time
Before — Anticipate The AI briefing. What changed in your finances since you last opened the app, what's coming up this week, and one recommended action — before the user has had to ask any questions.

During — Act One action card per session. Tap to see the data behind the recommendation — transparent, readable, not buried. Create goals, automate, see goal projections. Three options: act on it now, schedule it for later, or dismiss with a reason. Every dismissal trains the model.

After — Track progress, not transactions Weekly AI summary: Financial Health Score, Spending Breakdown, what's coming next. No ledger. No category pie chart. Just the three things worth knowing.



What I tried & rejected
I initially designed a conversational AI interface — a chat-style copilot where users could ask questions about their finances. The design was solid. The concept was logical. The user testing told a different story. When people are anxious about money, the last thing they want to do is articulate their anxiety to a machine. The blank chat prompt felt like a test they might fail. Several participants said some version of 'I don't even know where to start.' That sentence — 'I don't know where to start' — is precisely the problem a finance app should solve. A chat interface handed it back to the user. The pivot to action cards came directly from that failure. Instead of asking users to initiate, the product initiates. That shift — from reactive to proactive, from question-based to recommendation-based — is the design decision that everything else built on. The conversational interface wasn't a bad idea. It was an idea for a different user in a different emotional state.
Impact

Metrics projected from published fintech behavioural benchmarks and usability research patterns — not live product data.

Reflection
FinPilot solidified something I now apply to every product I design: the measure of good UX isn't whether people understand the interface. It's whether they do something differently because of it. Understanding is table stakes. Behaviour change is the point. The other thing it taught me: guidance that users don't trust won't change behaviour — no matter how accurate it is. Trust is earned at the interaction level, not the feature level. The action card isn't trustworthy because it's technically correct. It's trustworthy because it speaks plainly, shows its reasoning, and never asks more of the user than the moment can bear.
Good design reduces thinking, not just steps. Clarity without a call to action is incomplete. And guidance users don't trust won't change behaviour — no matter how accurate it is.
Check out the next project
