Back to selected work
BankingB2B SaaSLegacy redesignLead UX

VAM — Redesigning a legacy bank-agent application

Reorganized a fragmented desktop application around frontline jobs to be done, persistent system feedback, guided task flows, and reusable interaction patterns.

57%Fewer steps per task
4.3/5Agent confidence
63%Fewer escalations
90%Task success
Context
What it is

A B2B SaaS tool for bank agents to manage accounts, policies, transactions, and service requests.

Why redesign

Outdated UI patterns, inconsistent interactions, and module-first navigation no longer matched how agents actually worked.

My contribution

End-to-end redesign across discovery, journey mapping, IA, prototyping, design system, usability testing, and handoff.

Project goals
Reduce task time

Remove redundant steps and navigation dead ends.

Unify workflows

Bring account, policy, and transaction work into one coherent architecture.

Build for scale

Create reusable components and patterns engineering can extend safely.

Double Diamond
01Discover

Contextual interviews, workflow observation, and competitive review.

02Define

Personas, journey mapping, pain-point synthesis, and KPI definition.

03Develop

Information architecture, flows, wireframes, component library, and iterative prototypes.

04Deliver

High-fidelity design, usability validation, engineering collaboration, and QA.

Primary research
Contextual interviews

Explored task frequency, workarounds, friction, and unmet needs with frontline agents.

Workflow observation

Compared intended flows with real behavior to expose dead ends, redundant navigation, and missing system feedback.

Competitive review of enterprise banking tools pointed toward task-oriented navigation instead of feature-grouped menus.

Key findings
Navigation overload

Common tasks required roughly seven screens when two or three should have been sufficient.

Missing feedback

Actions lacked confirmation, loading, and useful error states, causing repeated actions and uncertainty.

Efficiency blockers

Inconsistent tables, date inputs, and unlabeled controls increased learning effort even for experienced agents.

Journey evidence

The highest friction appeared after account lookup.

LoginLow
Search customerMedium
View accountHigh
Update policyCritical
Confirm actionCritical
Success metrics
Task efficiency

Targeted a reduction from seven average steps to three using time-on-task and session analysis.

Agent confidence

Targeted improvement from 2.1/5 to above 4/5 using post-session feedback.

Escalation reduction

Targeted a 60% decrease in help-related supervisor escalations.

Design direction
Feature-grouped menus→ Task-oriented IA

Primary navigation reflects recurring jobs to be done.

Uncertain outcomes→ Persistent system status

Success, error, progress, and validation feedback appear at every consequential interaction.

Steep learning curve→ Guided disclosure

Labels, contextual help, and stepped flows reduce memorization and first-use burden.

Co-design & prototyping

Cross-functional structure workshops used card sorting and affinity mapping to test the proposed IA before high-fidelity design. Core flows were explored in low-fidelity first, then validated across three usability rounds with five bank agents per round.

Design system
Core components

Tables, filters, forms, status badges, action menus, confirmation dialogs, and notifications.

Design tokens

Typography, spacing, color, and elevation variables mapped to implementation-friendly rules.

Pattern library

Documented multi-step policy updates, bulk transactions, customer drill-downs, and exception handling.

Key screens
Agent dashboard

Pending work, recent interactions, daily metrics, and quick entry points for frequent tasks.

Customer account view

Account, policy, transaction, and open-case context in one consolidated view.

Policy update

Reduced a seven-screen process to a three-step guided flow with validation and review-before-submit.

Transaction management

Advanced filtering, sortable columns, status states, bulk actions, and safeguards around irreversible operations.

Outcomes
57%Reduction in average task steps
4.3/5Confidence, up from 2.1
63%Fewer supervisor escalations
90%Final task success rate
  • 40+ documented components delivered.
  • New-agent onboarding reduced from about three weeks to under one week.
  • No critical usability issues remained in the final QA round.
Email copied