Redesigning a legacy reporting tool for 3x more efficiency
Product Design
AI Integration
Redesign
Overview
We were tasked to redesign the legacy custom reporting tool within Lending Insights, an enterprise BI platform used by credit union executives. The existing workflow was inefficient and required frequent reliance on support teams to generate monthly reports.
We redesigned the experience into an AI-enhanced reporting system that empowers executives to independently create, customize, and interpret reports more efficiently and with greater confidence.
Client
Allied Solutions
Team
7 UX Designers
Timeline
Aug - Dec 2025
Context & Legacy Workflow
Credit unions operate under strict regulatory oversight and must prepare detailed, audit-ready reports each month. These reports require highly specific, customized data views across multiple sources.
Lending Insights was built to support this need. However, its custom reporting tool was configuration-heavy and unintuitive.
As a result, executives relied on Allied’s business support teams to generate reports on their behalf. This created frequent back-and-forth communication, increased operational workload for support teams, and inserted an unnecessary middle layer into what should have been a direct, self-serve workflow.
Identifying the Core Problems
Audit
To identify exact friction points in the report generation workflow within Lending Insights, we conducted a detailed audit of the custom reporting tool using Nielsen’s Usability Heuristics.
**scroll through annotations below to see the most prominent heuristic violations**
Understanding the User & the Market
Primary Research
Comparative Analysis
Translating Insights into Solutions
Brainstorming
Using all the insights and takeaways gathered previously, we began sketching ideas and features for the new reporting experience. Below is one of my main sketches that shaped the prototype I made in the next sprint and some ideas the entire team gathered
AI Assisted Prototyping
To translate our sketches into testable concepts in a fast and effective manner, I suggested we leverage AI design tools. With encouragement from our clients, we utilized Figma Make to rapidly generate two distinct high-fidelity prototype flows for A/B testing. This allowed us to immediately visualize our ideas and gather high quality feedback.
Version A
Dashboard-First: Designed for quick, easy starts.
Pre-built Templates: Ready-to-use report layouts to jumpstart creation.
Opt-In AI: AI features must be manually enabled in settings.
Basic Widgets: Simple, guided customization
Version B
Tableau-Inspired Builder: A familiar workflow for experienced analysts.
Sticky Config Panel: Always-visible tools for precise data editing.
Bento-Style Grid: Freely drag, drop, and resize widgets.
Embedded AI: One-click forecasting and a full AI report builder panel.
Trust & Privacy: Explicit AI permission modal to ensure data transparency.
Validating Concepts and Merging Workflows
A/B Testing
To determine which design direction best solved the core problem, we conducted A/B testing sessions with 3 Subject Matter Experts (SMEs). The testing revealed a clear verdict: users needed a hybrid of both approaches.
Version A
Users strongly preferred the dashboard-first approach of Version A. They noted that having saved reports and pre-built templates immediately visible on the main screen drastically reduced friction compared to hiding them in a side menu
Version B
SMEs emphasized that the comprehensive widget configuration pop-ups and flexible bento-box layout from Version B were absolutely essential for the detailed data manipulation that advanced analysts perform
Iterations
After sketching the new combined version, we transformed it into another Figma Make prototype to test further. We conducted a rapid usability test with the same 3 SMEs to ensure our merged concepts hit the mark. This helped us refine details like system status notifications and AI data-source transparency.
Crafting the Final Solution
Centralized Custom Dashboard
Users no longer have to dig for reports; the new dashboard provides immediate entry points to build from a blank canvas, utilize pre-built templates, or quickly access visually organized saved reports
Live-Preview Widget Builder
Replacing clunky configuration lists, the new centralized modal allows users to configure data sources and metrics while instantly visualizing the chart in real-time before adding it to their canvas
Uniform Global Filtering
Users can apply global filters (like Date Range or Risk Tier) at the top of the report to seamlessly update all widgets at once, maintaining data consistency across the entire dashboard
Data Privacy & AI Permissions
Users can apply global filters (like Date Range or Risk Tier) at the top of the report to seamlessly update all widgets at once, maintaining data consistency across the entire dashboard
One-Click Predictive Forecasting
By clicking the AI icon on an individual widget, users can instantly project future data trends directly onto their existing graphs
Actionable AI Summaries
The Impact
To validate the success of our redesign, we conducted task-based usability testing comparing the legacy platform to our new high-fidelity "Greenfield" prototype. The results proved that streamlining the navigation, clarifying terminology, and introducing the live-preview widget builder drastically reduced user cognitive load.
By transforming an unintuitive interface into a direct, self-serve platform, we achieved significant workflow improvements:
Reflection
Looking back, integrating AI-assisted prototyping was undoubtedly the biggest catalyst for this project's success. I had experimented with Figma Make and other AI design tools in the past just for fun, but I am incredibly glad I could bring that experience into a professional setting and turn it into an actionable, highly impactful part of our workflow.
Through this pivot, I learned significantly more about effective AI prompting and how to utilize AI specifically for user testing. Most importantly, it made our A/B testing far more impactful. By rapidly generating high-fidelity screens, we didn't have to test abstract, low-fidelity wireframes; we were able to present Subject Matter Experts with realistic, interactive interfaces. This allowed us to gather deep, precise feedback much earlier in the process, ultimately leading to a highly validated final product.
