AI Billing Audit Desk
Lead product designer
July 2024 - Jan 2025
Project Manager, Product Owner, On and off-shore development teams, 10+ stakeholders
Problem
Lessen’s billing team relied on 100% manual invoice reviews, causing severe bottlenecks. Valid invoices took up to 12 days to process, while rejected ones took over two months to resolve due to offline communication loops. Vendors lacked visibility into approval criteria, leading to high error rates and frustration.
Results
The implementation of an AI pre-screening system automatically routed 89% of submitted invoices directly to a "mass approval" bucket.
Intelligent hard and soft stop rules guided vendors during submission, dropping the invoice rejection rate from 6% down to just 1%.
Processing times were drastically reduced—valid invoices now clear in 1 day and errors are resolved in 3 days, enabling same-week vendor payouts.
83%
Reduced rejections by
89%
Automatic approvals
1%
Rejection rate
Process
Current State Analysis & Strategy
Collaborated with the Project Manager to audit existing manual review processes. We mapped out a new digital workflow and simplified billing requirements into actionable system rules.
Rule Definition & Logic Setup
Designed low-fidelity workflows integrating these new rules into the existing vendor invoice templates and mapped the journey to a brand-new internal Audit Desk for the billing team.
Wireframing & Prototyping
Collaborated with the Project Manager to audit existing manual review processes. We mapped out a new digital workflow and simplified billing requirements into actionable system rules.
Usability Testing
Conducted 30-minute qualitative testing sessions with experienced vendors and internal staff. Based on feedback, we refined the AI callout copy for clarity, removed unnecessary toggles, and adjusted layout elements to reduce the learning curve.
Design System Integration
Updated legacy interface elements and built out the new audit desk screens ensuring strict adherence to our modernized, cohesive design system for future scalability.
What I Learned
This product was a great space for me to explore the capabilities of AI both in my design process and in the product itself. A key part of my growth was determining the strategic 'hand-off' between AI-driven suggestions and automated backend rules, ensuring we didn't over-engineer solutions where simple logic sufficed.
Designing for this dual-sided ecosystem also required a delicate balance: I had to create a high-guidance submission experience for external vendors while simultaneously building a high-efficiency audit desk for our internal billing team. I was able to share my insight from usability testing with the internal team which was invaluable to them for this project and beyond.