Case Study 2025

FINSENSE: AI-Powered Financial Guidance

Designing a conversational AI experience that helps everyday people understand their spending, identify behavioral patterns, and make financial decisions without the stress.

Role Lead UX Designer, AI Interaction Designer
Scope Product discovery, experience design, AI interaction model, systems design
Platform Mobile application
Domain Personal finance, AI, behavioral design
Finsense App Interface
01 / WHY THIS PROJECT MATTERED

Most people know they should track their spending. Almost nobody does it consistently past the first two weeks.

The tools exist. The apps are free. The problem is not access.

The problem is that every budgeting app on the market was built around the assumption that users would behave like spreadsheets: consistent, disciplined, category-aware. Real people do not work that way.

shopping_cart They spend impulsively
event_busy They forget subscriptions
visibility_off They avoid their balance when behind
exit_to_app They quit apps that make them feel worse
"

FinSense was designed around a different assumption: that the right financial tool should adapt to how people actually behave, not punish them for failing to behave differently.

02 / WHAT I OWNED
psychology

Research and Strategy

  • record_voice_overUser interviews focused on budgeting anxiety
  • hubBehavioral spending pattern mapping
  • compareCompetitive analysis of existing tools
  • lightbulbAI opportunity identification
draw

Experience Design

  • chatChat-based decision flows
  • analyticsSpending analysis visualizations
  • subscriptionsSubscription tracking
  • psychology_altHabit breakdown screens
smart_toy

AI Interaction Model

  • edit_noteNatural language prompt design
  • targetContext-aware explanation patterns
  • recommendPersonalized recommendation logic
  • account_treeConversation flow architecture
widgets

Systems Design

  • schemaStructured logic for financial insights
  • dashboardHybrid chat and dashboard IA
  • library_booksConversation pattern library
03 / WHY EXISTING TOOLS FAIL

Before designing anything, I audited the tools people were already using and abandoning. The pattern was consistent across all of them.

category

Category-based budgeting assumes the wrong thing

Apps like Mint and YNAB organize spending into categories and show you totals. The assumption is that seeing the number will change the behavior. It does not.

Knowing you spent $340 on food delivery last month does not tell you which orders were worth it, which were stress purchases, or what a realistic target actually looks like for your life.

calendar_today

Rigid templates create immediate failure

Most apps ask users to set a budget before they understand their own patterns.

Users set unrealistic targets, fail them in week one, feel shame, and quit. The tool created the failure it was supposed to prevent.

show_chart

Dashboards require effort to interpret

Charts and trend lines are powerful for people who already understand their finances. For the users who need help most, they add cognitive load rather than reducing it.

You still have to do the work of figuring out what the data means and what to do about it.

gavel

None of them explain before they judge

Every existing tool surfaces a number and leaves the user to feel bad about it.

None of them contextualize the behavior, explain why it happened, or offer a realistic path forward. FinSense was built to do exactly that.

04 / RESEARCH: WHAT USERS ACTUALLY TOLD US

I conducted user interviews focused specifically on budgeting anxiety, not budgeting behavior. The distinction mattered.

I wanted to understand why people stopped using financial tools, not just how they used them.

I used AI to help synthesize interview notes and identify behavioral patterns across responses, which let me move from raw qualitative data to structured insights faster without losing the nuance in individual responses.

smart_toy
AI-assisted synthesis

Interview transcripts were processed using an LLM to cluster emotional themes and friction points by workflow stage. All findings were reviewed and validated by me before informing any design decisions.

Three themes became the foundation of the product strategy:

01

People want to understand their own normal, not hit an ideal

Most users did not want a perfect budget. They wanted to understand their own patterns well enough to make one realistic change at a time. The goal was self-knowledge, not optimization.

02

Behavior drives money problems, not categories

The real culprits were invisible: unused subscriptions running in the background, impulse purchases during stressful weeks, delivery spikes that crept up gradually. Users could not see these patterns in a category chart.

03

People only follow through when the next step feels achievable

Long-term financial plans overwhelmed users immediately. But a single specific suggestion consistently led to action. Small and concrete beat comprehensive and abstract every time.

"

I know I spend too much. I just don't know where to start and every app I've tried just makes me feel guilty without actually helping me do anything differently.

User interview participant
05 / PRODUCT STRATEGY: GUIDANCE OVER DASHBOARDS

The core strategic shift was reframing what a financial app is for. Not a tracking tool. Not a reporting tool. A guidance system.

Instead of

Show the user their data and let them figure out what it means

arrow_forward
FinSense does

Analyze the behavior, explain the pattern in plain language, and suggest one realistic next step

forum

AI Spending Advisor

The AI narrates spending trends in plain language, the way a financially literate friend would explain them.

Not this: "Food: 23% over budget" But this: "You spent significantly more on delivery this month, mostly on weekday evenings. That's usually a stress or tiredness pattern. Want to look at it together?"
radar

Behavior Detection First

FinSense surfaces invisible leaks: recurring subscriptions the user forgot about, impulse spending clusters, delivery spikes tied to specific times of week.

These are the patterns that actually drain budgets and they are invisible in a standard category chart.

lightbulb

Explanations Before Recommendations

Every insight follows the same structure: here is what I noticed, here is why it matters, here is one thing you could do.

The explanation comes first. The suggestion comes second. This was the single design decision that had the biggest impact on user trust.

chat

Chat as Primary Interface

Instead of navigating tabs and dashboards, users ask questions in natural language.

"Why did I spend so much last month?" "Am I on track for my trip in July?" "What subscriptions am I not using?"

The AI answers in context, with access to the user's actual transaction data.

06 / KEY DESIGN DECISIONS AND THE REASONING
01

Hybrid layout: dashboard for visibility, chat for decisions

A pure chat interface loses the at-a-glance awareness that makes financial tools useful day to day. A pure dashboard interface buries the guidance users actually need.

FinSense uses both: the dashboard gives users a current state view, the chat is where they go when they want to understand something or make a decision.

balance

Tradeoff: Two surfaces create a navigation challenge. I resolved this by making the chat persistently accessible from the dashboard rather than treating them as separate sections. The dashboard surfaces anomalies. The chat explains them.

02

Behavior-based insights instead of category totals

Rather than showing "dining: $340," FinSense shows "delivery orders up 60% this month, mostly weekday evenings."

The same data, reframed around the behavior rather than the category.

engineering

This required a different data model and close collaboration with the engineering team to define what patterns the AI would be trained to detect.

03

Realistic budget modeling based on actual behavior

Budgets in FinSense are generated from the user's real transaction history, not from generic templates.

The AI proposes a budget that reflects what the user actually spends, then identifies one or two areas where a realistic reduction is possible. Users do not start from zero. They start from their own baseline.

04

AI conversation flow design

Designing what the AI says was as important as designing the screens it lives in.

I built a conversation pattern library that covered: how the AI introduces an insight, how it handles a user who pushes back, how it responds when it does not have enough data, and how it maintains a tone that is supportive without being condescending.

person

The AI was designed to feel like a financially literate friend, not a bank chatbot.

07 / RESULTS FROM USER TESTING

These results come from usability testing sessions conducted with participants across different financial situations and comfort levels with money management tools.

35-50% Reduction in budgeting stress

Self-reported after using FinSense for two weeks compared to their previous tool

psychology Higher comprehension

Users understood where their money went without needing to interpret charts

speed Faster decisions

Minutes instead of hours, because the AI surfaced the relevant context directly

task_alt Increased follow-through

Users attributed this to recommendations feeling personalized and achievable rather than generic

verified

Trust finding

Users trusted the AI significantly more when it explained the pattern before offering a suggestion. This validated the explanation-first design principle and shaped the entire conversation flow architecture.

favorite

Most appreciated feature

The behavior detection that surfaced invisible spending patterns users had not noticed themselves. This was consistently described as the moment the app felt genuinely useful rather than just another tracking tool.

08 / WHAT I LEARNED

AI in product design is not primarily a technical question. It is a trust question.

Users do not resist AI because they do not understand how it works. They resist it because they do not trust that it understands them.

The explanation-first pattern was the most significant design decision of the project and it came directly from that insight. When the AI explained what it noticed before telling a user what to do, trust scores went up measurably in every testing round.

lightbulb

The AI was not more capable in those flows. It was more legible. Legibility and capability are not the same thing and in AI product design, legibility matters more.

Designing conversation flows is technically demanding in unexpected ways

It required thinking about edge cases the way an engineer thinks about them:

  • What does the AI say when it does not have enough data?
  • What does it say when a user disagrees with its assessment?
  • What does it say when the same pattern repeats for the third month in a row?

Every one of those cases needed a designed response, not a default one.

The feeling of being analyzed vs. being understood

FinSense reinforced something I now bring into every AI project: the best AI interfaces reduce the feeling of being analyzed and increase the feeling of being understood.

Those are opposite emotional experiences and the design is what creates the difference between them.

08 / FINAL SCREENS

The culmination of research, iteration, and user-centered design—bringing FinSense to life.

FinSense Final Screen 1
FinSense Final Screen 2

Like what you see?

Let's build something together.

Get in touch arrow_forward View more work