Back to Home

Made complex financial decisions feel intuitive and trustworthy

Role Lead Product Designer
Team 2 Designers, 5 Engineers, 1 Data Scientist
Timeline 14 weeks
Tools Figma, Principle, Amplitude
AI Integration Conversational UX Fintech Trust Design Mobile
Aura AI financial assistant mobile interface

Bringing AI into finance without breaking trust

Aura is a conversational AI layer for a consumer fintech app serving 2M+ users. The company wanted to help users make smarter financial decisions — budgeting, investing, saving — through a natural language interface. The challenge wasn't the AI; it was building an experience that felt trustworthy with people's money.

I designed the conversational UX framework, trust patterns, and interaction model. The assistant achieved 73% weekly engagement and a trust score of 4.2/5 — unusually high for AI-driven financial tools.

Users had the data but not the confidence to act

The app already tracked spending, investments, and savings goals. But analytics showed 82% of users viewed their dashboards without taking action. Exit surveys revealed the core issue: "I can see my data, but I don't know what to do with it." Users wanted guidance, not just information.

  • 82% passive dashboard viewers (no actions taken)
  • Average user opened the app 4x/week but completed <1 financial action/month
  • 61% of users said they "wished someone would just tell me what to do"
  • Support tickets about "what should I do with X" grew 340% year-over-year
Financial data visualization showing user engagement patterns
User engagement analysis revealing the gap between data consumption and action

Three truths about trust and money

1. Trust is earned in explanations, not predictions

Users didn't care how accurate the AI was; they cared whether they could understand why it suggested something. Explainability > accuracy for adoption.

2. The uncanny valley of financial advice

Users were comfortable with AI for small decisions (round up savings, categorize expenses) but deeply uncomfortable with AI making larger calls (investment allocation, debt strategy). There's a clear threshold where AI goes from helpful to unsettling.

3. Conversation ≠ chat

Early prototypes used a chat interface. Users found it exhausting — too much typing for financial decisions. They wanted guided choices with the option to ask questions, not open-ended conversation.

User research session on financial trust patterns Team workshop analyzing user interview findings

Designing for trust at every interaction

Strategy: "Guided choice" over "open conversation"

Instead of a chat-first model, we designed a card-based interaction pattern where Aura presents structured recommendations with clear reasoning, and users can drill deeper through natural language only when they want to.

Dead end: the chatbot trap

First prototype was a full conversational chatbot. Users engaged initially out of curiosity but abandoned after 2-3 sessions. The effort-to-value ratio was too high — typing out financial questions felt laborious and uncertain.

Users told us they wanted a conversation. What they actually wanted was a confident recommendation they could question. That distinction changed everything.

The pivot: card-based recommendations with conversational depth

Aura proactively surfaces 2-3 actionable cards per day based on the user's financial data. Each card shows: what to do, why, expected impact. Tapping a card opens a conversational thread for questions.

Iteration: trust calibration

Every recommendation shows a confidence indicator and reasoning chain. For high-stakes suggestions (>$500 impact), Aura explicitly flags uncertainty and suggests consulting a professional. This "humble AI" pattern increased trust scores by 40%.

Design iterations of the card-based recommendation interface
Evolution from chatbot prototype to card-based guided choice interaction model

An AI that earns the right to advise

Smart cards

Proactive, contextual recommendations surfaced at the right moment (post-paycheck, unusual spending, goal progress). Each card includes a clear action, reasoning, and projected outcome.

Reasoning transparency

Every suggestion shows a simple breakdown of the logic: "Based on your spending pattern + your savings goal + current interest rates → here's what I'd suggest." Users can tap any factor to explore it.

Graduated autonomy

Users set their own comfort level from "always ask me" to "handle small decisions automatically." Aura remembers preferences and adapts, earning trust incrementally rather than demanding it upfront.

Smart cards recommendation interface Reasoning transparency breakdown screen
Graduated autonomy settings and trust calibration interface
The graduated autonomy settings allowing users to control their comfort level with AI decisions

Measurable outcomes across every metric

73%
Weekly engagement rate
4.2/5
User trust score
3.8x
Financial actions per user/month (up from <1)

Beyond the headline numbers, the redesign had a ripple effect across the business:

  • Users who engaged with Aura saved an average of 23% more than non-Aura users
  • App store rating increased from 3.8 to 4.5 after launch
  • The smart cards pattern was adopted as the company's standard pattern for all proactive features

For the first time, I feel like I actually understand what my money is doing. Aura doesn't just tell me what to do — it helps me understand why.

— Beta user, post-launch survey

What's next, and what I'd do differently

We're expanding Aura to support shared financial goals for couples and families — a complex design challenge around shared autonomy and joint trust.

If I could go back, I'd prototype with real financial data sooner. We spent too long with hypothetical scenarios that didn't surface the edge cases — like how to handle recommendations during market downturns. The emotional weight of real money changes how users perceive every interaction.