jAI Financial Intelligence
Interpret fragmented money data and convert it into clarity and action. Reducing financial anxiety by lowering cognitive barriers.
Product
Neobank Super App
Role
Lead Designer
Users
Pan-India
Outcome
High-Trust AI
01. Problem Statement
Indian users struggle to plan and act on their finances not because they lack financial products, but because:
- × Money is spread across multiple bank accounts.
- × Transaction data is raw, fragmented, and hard to interpret.
- × Most fintech apps assume English proficiency.
- × Users set financial goals emotionally without checking feasibility.
The Constraint: Without Account Aggregator data, AI-driven insights and goal feasibility are unreliable.
02. Core Challenges
1. Inherited Product Direction
Core requirements were defined by BAs. Foundational decisions (e.g., mandatory AA) were locked. Design had to operate within predefined boundaries, not from first principles.
2. Mandatory Account Aggregator
AA linking was essential for intelligence but increased hesitation. Removing it wasn't an option, creating tension between business necessity and user trust.
3. Designing AI Without Over-Promising
Financial insights had to be explainable and conservative. No "AI certainty" or guarantees, to avoid liability and mistrust.
4. Multilingual Accessibility
AI responses needed to maintain tone and intent across 7 languages (English, Hindi, Bengali, Marathi, Tamil, Telugu, Kannada).
5. High-Trust Goal Feasibility
Labeling goals as "not realistic" carries emotional weight. Recommendations must feel supportive, not judgmental.
03. Approach & Solution
Requirement Deconstruction
Instead of designing screens immediately, we translated business logic into user-centric scenarios. Parallel collaboration with developers uncovered technical limitations early.
Designing Trust (Sample Data)
To reduce friction for the mandatory AA flow:
Sample Data Mode
Allowed users to explore jAI's value without linking accounts immediately.
Persistent CTA
Balanced exploration with business goals, accepting slower conversion for higher confidence.
Multilingual Conversational Design
We prioritized intent preservation over literal translation using simple financial language and a reassuring tone.
Core Intelligence (Phase 1)
Income Intelligence
- Auto-detection across banks
- Custom categories
- Income source insights
Expense Intelligence
- Editable categories
- Transaction re-categorisation
- User control > AI correctness
Predictive Budgeting
- Based on historical behavior
- Progress indicators
- Guidance, not enforcement
04. The Feasibility Engine
Goal Management extends jAI from insight to decision support. The objective: Help users understand if a goal is Realistic, Borderline, or Not Realistic.
How it Works
1. Input & Data
Goal details + User financial profile (Income/Expense prefilled from jAI). Always editable to preserve agency.
2. Backend AI Logic
Calculates SIP capacity, cost-of-living adjustments, and evaluates achievability using asset class ROI assumptions.
Optimization
For unachievable goals, AI recommends: Step-up SIP, Increased Amount, or Extended Duration.
Calculated
Each option is independently calculated. No feasibility logic runs on the client.
Continuity
Selecting an option redirects to investment flows with context passed forward.
05. Execution & Delivery
- Designed end-to-end user and conversational flows.
- Took explicit BA approvals before UI finalization.
- Delivered Phase 1 screens with detailed notes on AI assumptions and edge cases.
- Three major design iterations shaped by feasibility and compliance.
06. Learnings
Constraint-Led Design
Impact comes from shaping clarity within fixed decisions, not just removing constraints.
Trust = Control
Editability and explanation mattered more than advanced black-box prediction.
Multilingual Systems
Accessibility requires alignment across language, tone, data, and intent—not just copy.
Emotional Sensitivity
Labeling goals as "not realistic" must be paired with constructive paths forward.
Closing Note
"Designing AI-driven financial systems is not an idealized UX story. It’s a real product story about making intelligence accessible under strict constraints."