Finnovó
Designing a Decision System for Financial Avoidance
The Core Problem~
Users didn’t abandon finance apps because of complexity — they avoided them because each interaction forced emotionally loaded decisions.
Finnovó explores how restructuring when and how decisions appear can reduce avoidance under cognitive load.
Impact & Outcomes
(Exploratory)
~
• Improved clarity around spending boundaries
• Increased user confidence in AI-assisted guidance
• Reduced perceived decision fatigue during budgeting
Role — Scope — Constraints
• Sole UX Researcher & Product Designer
• 4-week exploratory system design
• *conceptual, mobile-first
Focus
• Human–AI decision logic · Cognitive load
• Trust & failure states · System constraints
This is a UX-led system design artifact — not a UI showcase.
Finnovó explores how AI-assisted systems can reduce decision fatigue and guide financial behavior — while preserving user control and trust.
Why Most Finance Apps Fail at Decision-Making
Most personal finance tools are designed around a flawed assumption:
If users are given enough data, they will make better financial decisions.
In practice, the opposite consistently happens.
Modern finance apps require users to continuously manage:
• Constant balances, charts & categories
→ always-on financial awareness
• Multiple competing recommendations
→ conflicting guidance
• Manual rules, alerts, and thresholds
→ configuration responsibility
• Long-term goals — short-term trade-offs
→ unresolved tension
Instead of enabling clarity, these systems increase cognitive load and shift responsibility entirely onto the user.
A typical finance UX failure loop:
choice overload → emotional pressure → avoidance → abandonment
The result is not lack of discipline — but decision fatigue.
The Core Misalignment
Through secondary research and competitive analysis, a consistent pattern emerged:
Finance products tend to optimize for:
While users actually need:
• Visibility (more insights, more metrics)
→
Fewer, better-timed decisions
• Control (manual categorization, adjustable rules)
→
Clear boundaries instead of constant choice
• Engagement (frequent prompts and nudges)
→
Guidance that reduces mental effort, not adds to it
This misalignment leads users to disengage — not because they don’t care about money,
but because the system demands too much cognitive work.
Reframing the Problem
The problem is not budgeting.
The problem is decision design.
*The real question becomes:
" How might an AI-assisted system help users make fewer, better financial decisions
— without removing their sense of agency? "
This reframing shifts the design goal from:
Building a finance app
→
Designing a decision-support system
Design Implication
This insight set the foundation for Finnovó:
• The system should absorb complexity, not expose it
• Decisions should be timed and contextual, not constant
• AI should act as a supportive layer, not an authority
Only after redefining the problem this way did interface design become meaningful
Common False Assumptions in Finance Products
Assumption #1
"More data leads to better decisions"
Reality:
More data increases interpretation cost, not clarity.
Breaks when:
• Options multiply
• Context switching increases
• Uncertainty remains unresolved
📌
Failure mode:
Decision paralysis
Assumption #2
“Users want full control at all times”
Reality:
Users want guardrails, not micromanagement.
What users actually prefer:
• Defaults they can trust
• The option to intervene
• Not having to decide constantly
📌
Failure mode:
Control fatigue
Assumption #3
“Automation is either trusted or rejected”
Reality:
Trust is conditional and progressive.
Users trust automation when they understand:
• Why a suggestion exists
• What assumptions power it
• What happens if it’s wrong
📌
Failure mode:
Blind automation → distrust
ASSUMPTION → REALITY → FAILURE MODE
Blind automation reduces trust.
Transparent, reversible automation builds it.
System Constraints That Shaped Finnovó
Designing Guardrails, Not Just Features
1. Reduce Decisions, Not Information Accuracy
Constraint:
The system cannot surface information unless it removes a decision.
Design consequence:
Raw financial data is aggregated & delayed until it becomes decision-critical.
2. Preserve Agency Through Reversibility
Constraint:
No system action is final without user intent.
Design consequence:
All AI guidance is explainable, adjustable & reversible.
3. Default to Boundaries, Not Optimization
Constraint:
The system does not optimize for “more” or “less.”
Design consequence:
Users are shown safe, risky & forbidden ranges instead of goals.
4. Design for Cognitive Load, Not Engagement
Constraint:
Success is not measured by frequency of interaction.
Design consequence:
The system minimizes interruptions & surfaces only decision-critical moments.
Outcome:
Trust increased not through engagement, but through predictability and reversibility.
What the System Explicitly Refuses to Do
Defining constraints was equally important.
• Manual categorization
• Competing recommendations
• Constant budget micromanagement
• Assumed financial literacy
These exclusions prevented the system from slipping back into UI-heavy, control-heavy patterns.
What We Explicitly Did Not Optimize For
Trade-Offs & Constraint
→
Why it mattered
→
Decision
Fully Autonomous Actions
AI never executes financial actions without user confirmation
→ Preserved trust & user agency
Exhaustive Financial Dashboards
No raw data walls by default — only decision-ready summaries
→ Reduced cognitive load
Punitive Alerts & Guilt Loops
Contextual, non-judgmental nudges only
→ Higher emotional safety & re-engagement
Engagement-Driven Design
System stays quiet unless intervention is meaningful
→ Quiet reliability over attention capture
These constraints enabled predictable AI behavior, clear financial boundaries, and trust built through restraint rather than intelligence.
Why a Decision Loop Was Necessary
In most finance apps, automation operates invisibly — recommendations appear without clear triggers, assumptions, or failure handling.
Finnovó treats AI not as a prediction engine, but as a decision mediator operating within strict system constraints.
A Constrained, Interpretable Decision Cycle
Every system action in Finnovó passes through the same loop.
The system cannot suggest, intervene, or alert unless all stages are satisfied.
Behavior
→
Signal
→
Suggestion
→
User Control
→
Feedback
Signal Detection
The system monitors financial signals passively without surfacing them.
Spending drift, Boundary proximity, Pattern deviation
Context Assembly
Signals are evaluated within user-defined boundaries, historical behavior, and risk tolerance.
No signal is evaluated in isolation.
Decision Eligibility Check
The system asks: Does acting now remove a meaningful decision for the user?
If no → system stays silent.
If yes → continue.
Explainable Recommendation
When a suggestion is surfaced, the system exposes:
• Why now
• What assumptions were used
• What outcome is expected
No black-box actions.
User Override & Reversibility
Users can adjust, delay, or reject any suggestion without penalty.
Rejection is treated as signal, not failure.
System Learning (Non-Authoritative)
The system adapts boundaries and timing — not user intent.
Finnovó does not “optimize behavior”; it optimizes decision timing.
Finnovó treats AI as a collaborator in decision-making — not an invisible operator.
What Happens When the System Is Wrong
Finnovó assumes uncertainty by default.
• No action is irreversible
• Ambiguous situations default to silence
• Users can inspect and rewind past system decisions
Why This Loop Matters
By constraining when AI can act — and exposing how it reasons — Finnovó reduces cognitive load without removing agency.
The system earns trust through predictability, not persuasion.
Trust is preserved not by correctness — but by recoverability.
Key Decision Moments
Intervention vs Silence
Before a Spending Decision
User is about to spend, but hasn’t crossed a limit yet
System Behavior:
• Detects proximity to a budget boundary
• Surfaces a subtle, dismissible nudge
• No blocking, no alerts, no urgency
→ Prevents regret without blocking intent
After an Overspend
User exceeds a budget boundary.
System Behavior:
• No alerts. No penalties.
• Post-action reflection shown later
• System recalibrates future guidance
→ Reduces shame and encourages recovery
Pattern-Level Risk (Not Single Events)
Repeated small deviations across time.
System Behavior:
• Detects behavioral drift
• Surfaces a pattern insight, not a warning
• User decides whether to act
→ Shifts focus from events to habits
When the User Disagrees with AI
User chooses to override a recommendation.
System Behavior:
• Override allowed instantly
• Rationale shown, not enforced
• No corrective pushback
→ Preserves user agency; AI remains an advisor
Principle Behind All Interventions
Intervene only when the system can reduce cognitive load —
stay silent when intervention adds none.
Failure Was the Design Constraint
Failure was not treated as an exception.
It was assumed, expected, and designed for.
Silence Tolerance
System remains unused but not abandoned
(no uninstall, no opt-out)
→ Quiet reliability
Finnovó does not attempt to prevent financial mistakes in real time.
It focuses on recoverability, reflection, and trust continuity.
Primary Failure: Budget Breach (Overspending)
Overspending is the most common and most emotionally charged failure.
System response
• No alerts
• No warnings
• No intervention mid-action
The system waits until after the decision, then reframes it.
Why
• Prevents shame-driven abandonment
• Preserves user autonomy
• Keeps trust intact even when goals are missed
Failure is absorbed — not punished.
Secondary Failures: When Users Disagree or Disengage
These are not treated as errors. They are treated as signals.
• Ignored recommendation → system adapts quietly
• User override → disagreement becomes learning input
• Repeated deviation → patterns surfaced, not mistakes highlighted
The system listens more than it corrects.
Recovery Over Prevention
Finnovó optimizes for:
• Reversibility without penalty
• Predictable behavior over enforcement
• Long-term trust over short-term compliance
Users are never forced back “on track.” They are invited to understand their own patterns.
Trust Model
Trust is not built by being right.
It is built by being predictable when things go wrong.
Finnovó remains consistent — even when users don’t.
Design Principle
Failure is the primary use case.
If the system works when users fail, it will work when they succeed.
Why This Matters
This approach reduces:
• Financial anxiety
• Drop-off after mistakes
• Defensive or avoidance behavior
And increases:
• Honest engagement
• Long-term retention
• Self-directed improvement
Failure is absorbed — not punished.
Outcome
Finnovó becomes:
• A system users return to after mistakes
• Not a system they avoid because of them
Interface as Consequence (UI)
The interface was treated as a consequence of system behavior — not a surface for features.
User Journey Map
Early sketches focused on placement of decision moments, not visual style.
From Decision Logic → Information Architecture
• Home
→
Decision awareness
• Expenses
→
Behavior visibility
• Budgeting
→
Boundary definition
• Statistics
→
Reflection, not control
• AI Modeling
→
Explanation & adjustment
Each surface exists to support a specific cognitive state — not a task list.
user flows
Each flow was designed to end cleanly — without looping the user back into the system.
Pen-and-paper sketches
Early sketches focused on placement of decision moments, not visual style.
Low-fi greyscale wireframes
Visual hierarchy was tested before color, motion, or branding.
Why the Interface Looks Quiet
Visual quietness was a functional decision, not an aesthetic one.
High-contrast interfaces create urgency.
Urgency increases decision pressure.
Decision pressure was the problem being designed against.
The interface intentionally:
• Limits color usage to boundary states
• Avoids high-saturation success or failure signals
• Maintains consistent visual hierarchy across screens
Even in dark mode, contrast was restrained to prevent visual dominance from becoming emotional pressure.
*Dark mode shown for consistency; system supports light surfaces in production
Screens as Decision Moments (Not Feature Tours)
The interface was treated as a consequence of system behavior — not a surface for features.
Home — Decision Awareness, Not Control
The home screen exists to prepare decisions — not demand them.
System context
• Detects behavioral drift
• Tracks proximity to boundaries
• Knows timing sensitivity
Intentionally hidden
• Transaction noise
• Predictive certainty
• “You should” language
Interface response
• Calm summary, no urgency
• Optional, dismissible nudge
• No raw data surfaces by default
Budget — Boundaries Over Goals
Budgeting was treated as boundary-setting, not self-discipline training.
System context
• Knows safe, risky, and forbidden ranges
• Detects slow drift, not single mistakes
Intentionally hidden
• Perfection metrics
• Optimization prompts
• Gamified streak pressure
Interface response
• Range-based budgets instead of targets
• Sliders emphasize reversibility
• No red states, no “failure” screens
Budget — Boundaries Over Goals
Budgeting was treated as boundary-setting, not self-discipline training.
System context
• Confidence-weighted signals
• Known uncertainty
Intentionally hidden
• Confidence theatrics
• Corrective language
• Escalation paths
Interface response
• Accept / Ignore / Adjust always visible
• Explanation available, not forced
• Overrides have no penalty
The system never escalates disagreement into friction.
Security & Trust
Protection without intimidation
Financial security shouldn’t interrupt flow or amplify anxiety. Finnovo’s security system is designed to protect users quietly, predictably, and only when risk meaningfully increases.
Progressive security
Security appears only when risk increases, not by default.
Friction proportional to risk
Stronger checks for higher-impact actions, not everyday usage.
Clarity over alarm
Alerts explain what happened and what to do — without fear-based language.
Biometric Login
Two-Factor Authentication (2FA)
Device Security + Password Manager
Data Encryption
01
Biometric Login
Enables instant, password-free login using Face ID or fingerprint authentication.
• Prevents unauthorized access without disrupting everyday usage
• Respects user flow while protecting critical moments
• Avoids constant security fatigue
Biometric authentication is device-dependent and supplemented by fallback security for sensitive actions.
This security system was designed and validated through simulated flows and UX reasoning, prioritizing emotional safety and trust — not performance metrics or financial outcomes.
DECISION PRESSURE REDUCTION LOOP
User-Level Outcomes (NOT metrics-first)
Observed shifts in user behavior:
• Users delayed fewer purchases impulsively after boundary nudges
• Overspending moments did not lead to disengagement or shame-driven churn
• Users overrode AI suggestions without abandoning trust in the system
What mattered:
The system reduced decision pressure, not financial mistakes.
System-Level Outcomes
What changed in the system itself:
• Fewer alerts fired, but higher action relevance
• AI recommendations became more explainable, not more accurate
• Overrides improved future guidance instead of triggering corrective loops
The system learned restraint before it learned optimization.
MORE SIGNALS → FEWER SURFACED DECISIONS → HIGHER TRUST
Business & Technical Realities
As an exploratory system design, Finnovó operated under real constraints:
• No access to real banking APIs
• No historical user trust or training data
• AI outputs could be uncertain or wrong
These constraints defined the system. The interface is a consequence of them.
Validation & What Would Be Tested in Production
What This System Was Not Validated On
This was an exploratory system — not a live fintech product.
• No real banking APIs
• No long-term user data
• No financial outcome claims
These aspects were validated through design reasoning, simulated flows, and behavioral modeling:
What Could Be Validated Pre-Production
System Predictability
• The system behaves consistently across similar failure scenarios
• No hidden escalation paths or surprise automation
• Overrides & breaches guarantee traceable failures
→ Goal: Make system stress states transparent & understandable.
Emotional Safety Under Failure
• No punitive alerts after overspending
• No guilt-based copy or corrective pressure
• Recovery language remains neutral and forward-looking
→ Goal: Prevent avoidance, shame-driven churn, or disengagement
Override & Disagreement Handling
• Overrides never trigger penalties or friction
• Ignored recommendations quietly inform future guidance
• Disagreement strengthens, rather than weakens, trust
→ Goal: Trust through reversibility, not compliance
Decision Pressure Reduction
• Fewer surfaced decisions during high-noise moments
• Suggestions framed as optional, deferrable & reversible spent
• Silence preserved when intervention adds no value
→ Goal: Reduce cognitive load without reducing agency
The system remains stable when users aren’t.
What Would Only Be Validated in Production
If deployed with real users, validation would focus on:
• Long-term trust retention after repeated failure
• Calibration of when to intervene vs remain silent
• User understanding of AI uncertainty
• Behavioral drift over time (habits, not events)
These cannot be responsibly simulated
Success Criteria (If Deployed)
Finnovó would be considered successful if:
• Users return after mistakes
• Overrides don’t cause disengagement
• Fewer alerts with higher relevance
• Stable trust despite imperfect predictions
The system succeeds if users stay — even when it’s wrong.
Why This Validation Approach Matters
Most fintech products validate control. Finnovó was designed to validate resilience — separating what can be reasoned about safely from what must be earned through time and trust.
This avoids over-claiming intelligence and instead earns credibility through restraint.
What we chose to optimize for
Primary Success Signals
Reduced Decision Friction
Fewer moments where users hesitate or abandon budgeting actions
→ Signals cognitive safety, not faster decision-making
Budget Boundary Adherence
Spending remains within suggested ranges over time, allowing natural deviations without triggering punitive feedback.
→ Healthy behavior, not discipline theater
Healthy override rate (not zero)
Users occasionally reject system suggestions while maintaining long-term engagement.
→ Trust without obedience
Recovery Speed After Failure
Time taken for users to re-engage after overspending or boundary breaches.
→ Emotional safety indicator
These signals were chosen over traditional ‘performance’ metrics.
Why This Project Matters
Finnovó is not about budgeting.
Good UX is less about adding features — and more about deciding what not to surface, when, and why
Key Design Learnings
• Behavior-first design outperforms feature-first design Especially in financial contexts where fear and overload dominate.
• Motion is not decoration — it sets emotional pacing Micro-feedback reduced hesitation more effectively than static clarity.
• AI is most useful when it reduces choice, not when it explains everything Contextual hints outperformed predictive recommendations.
What I Would Improve With Real Users
• Validate GenIndic language clarity with real financial stress scenarios
• Test adaptive states triggered by spending behavior over time
• Measure recovery speed after financial mistakes, not task completion
• Personalize onboarding to reduce early cognitive load
This project will evolve beyond a 2D interface into a system that explores how motion, spatial computing, and AI can make financial information feel less intimidating and more humane.