Transforming Document Automation Through Human-Centered Design - A Case Study by the Founding Product Designer.
Project Overview
As founding Lead Product Designer at Orby AI, I designed a Generative AI Automation Platform that revolutionizes enterprise document processing. Launched with Google in early 2023, Orby streamlines workflows through a seamless observe-learn-adapt-automate cycle, transforming how organizations handle invoice processing.
My Role
I joined Orby AI as the company’s Founding Product Designer. I designed the experience of the review and validation process. I hold a cross-functional workshop, collaborated with stakeholders, engineers and project managers.
The Challenge
Financial teams faced overwhelming document volumes, with specialists spending 45 minutes per invoice, struggling with accuracy concerns, file accessibility issues, and skepticism about AI's reliability. Any solution needed seamless integration with existing tools like Google Drive, SAP, and Oracle.
Research & Design Approach
I developed personas (primarily finance specialist Amy), conducted contextual inquiries, and collaborated with stakeholders to create a strategy centered on transparency, progressive learning, and seamless integration with existing workflows.
Design Strategy
Based on my research findings, I developed a design strategy centered around three key principles: Building user confidence by making AI operations visible and allowing users to verify and override suggestions when needed; creating interfaces that help Orby learn from user actions and improve over time; and designing experiences that fit naturally into existing workflows and tools.
Solution Design
My comprehensive solution included: a training mode that provided visual feedback as Orby learned; a goal-oriented suggestion list with clear file differentiation; and a confidence-building review interface that displayed AI confidence scores and enabled one-click validation while maintaining user control.
First, I broke down the user needs into actionable steps:
Deliverable - Suggestion List
I designed intuitive interface with visual differentiation between file types, hierarchical information display, and AI-powered suggestion panel to help users quickly identify and prioritize important documents.
Deliverable - Accessible Review Page
I created transparent interface showing confidence scores, highlighted source text, and side-by-side comparison of original and extracted data, enabling efficient validation while building user trust in AI accuracy.
Iteration Process
Through close collaboration with stakeholders and users, I refined the interface over multiple iterations. The training mode made learning visible, file management enhanced metadata display, and the review process implemented intuitive keyboard shortcuts that transformed the user experience.
Measurable Impact
End-user testing showed processing time dropped from 45 to 9 minutes per document (80% reduction) with 90% improvement in extraction accuracy. Client success stories included a tech giant achieving 100% audit rate (saving $4M) and 64% invoice automation for a top real estate firm.
100% audit rate achieved for a tech giant, saving $4M
95% reduction in processing time for a leading IoT solutions provider
64% invoice automation for a top real estate firm
65% reduction in reconciliation time for an AI and data platform
60% reduction in manual data entry for a leading streaming platform
Reflection
This project taught me to prioritize user needs, communicate openly with engineers, embrace iterative design, and balance automation with user control. The Google Project for Automating Invoice Processing with Orby AI stands as a testament to how thoughtful design can transform complex AI into intuitive tools that solve real problems.