Receipt review
Helped the operations team review receipts 40% faster, saving 100s of hours, or ~$250k in overhead.
Problem
Level’s vision was to deliver dental, vision and any employee benefit all via 1 card. So in 2022, engineering rebuilt the core transaction engine to support lifestyle spending accounts like Vanguard’s Flexfund. But, this new backend introduced massive processing bottlenecks and headaches for our Transactions Ops team.
- 40% of all system transactions (~180k/year) failed automated processing and required human intervention.
- ~120k of those manual transactions required explicit receipt review.
- 5 Ops associates spent 65 hours combined every week manually reviewing receipts, costing ~$250k/year.
- The legacy workflow was slow, fragmented, and unpredictable. Hiring more staff to scale was financially unsustainable.
Solution
Deliver a fast, short-term UI overhaul while leadership evaluated long-term OCR/ML automation.
- Split-View Modal Pattern: Designed a new component featuring a high-powered receipt preview on the left, paired with an optimized data entry form on the right.
- Smart Categorization & Focus: Placed suggested merchant categories above input fields and allowed users to skip or queue complex receipts to maintain momentum.
- Safety & Fallbacks: Built-in duplicate transaction alerts, invalid receipt flags, and an “eject” option back to legacy UX for edge cases.
- Execution & Rollout: Handed off via interactive Figma prototypes and video walkthroughs; validated via a 1-week A/B test monitored in FullStory before rolling out to all 12 Ops associates.
Outcomes
- Ops associates reviewed receipts 40% faster on average.
- $250k+ Annualized Savings: Reduced thousands of manual human hours, saving hundreds of thousands of dollars.
- Auto-Processing Lift: Contributed to moving automated transaction processing from 60% toward a 70% target by late 2023.
- Design System Reuse: Established reusable split-view and file-preview UI patterns for future internal admin workflows.
Reflection
- Automation is Hard: Training OCR/ML models requires thousands of clean, labeled receipts—making immediate UX interventions far more cost-effective in the short term.
- Inclusion Drives Adoption: Involving Ops teammates deeply throughout research, prototyping, and testing built the trust necessary for seamless product adoption.
- Designing for Efficiency with Empathy: Streamlining repetitive internal tasks frees teams from low-value work, but structural tech shifts require thoughtful change management.