Document Processing

Background

Document management is core to Rabbet, which uses AI to split large "Draw Summaries" into individual documents and parse relevant data. Despite good accuracy, errors in splitting and parsing were common, making review and correction cumbersome.

Objectives

Improve the document processing interface to reduce errors, enhance user experience, and streamline the review process.

Role

Lead Designer, overseeing one other designer.

  • Conducted research to understand existing interface issues.

  • Collaborated with stakeholders to develop solutions.

  • Scoped features accurately.

Methods

We interviewed users, the CEO, and internal teams (customer success and sales). Through an affinity mapping exercise, we identified and prioritized pain points, focusing on the top five issues to solve iteratively.

Problem-Solving and Solutions

Problem 1: Errors in Document Splitting

Issue: The AI often made errors in splitting documents, leading to costly mistakes and inefficient corrections.

Solution: We pre-empted the split by asking a simple question that provided more clarity about the uploaded documents, which improved accuracy and avoided unnecessary splits. This resulted in more precise splits and reduced the need for costly corrections.

Problem 2: Loss of Context in Document Review

Issue: Once documents were split, they lost their context within the draw, making it difficult to review them effectively.

Solution: We updated the document table to sort by the original order and added an "uploads viewer" to allow users to see the original uploaded document, the splits, and make adjustments as needed. This maintained the necessary context and improved review efficiency.

Problem 3: Difficulty in Tracking Reviewed Documents

Issue: Users had no way to track which documents had been reviewed, complicating the review process.

Solution: We added a "reviewed" label and button for tracking, and enabled document assignment to specific users, with sortable and filterable options to manage review status. This provided clear visibility into the review progress and responsibilities.

Problem 4: Cumbersome Data Entry Interface

Issue: The data entry interface was inconsistent and difficult to navigate, hindering efficient data processing.

Solution: We expanded the modal to full screen, standardized form elements for consistency, and added a sidebar for easier document navigation and context. These changes streamlined data entry and made the interface more user-friendly.

Problem 5: Understanding the Effect of Individual Documents on the Draw

Issue: Users were unclear on how their data entries affected the overall draw, leading to confusion and inefficiency.

Solution: We added a draw total bar to the document modal to show progress and implemented toast messages to display the effects of individual document entries on the draw. This provided immediate feedback and improved users' understanding of their actions' impact.

Outcomes and Impact

  • Efficiency Gains: Reduced document processing time by 60%.

  • Customer Satisfaction: Contributed to an overall NPS score of 50.

  • Positive Feedback: Received positive feedback from stakeholders and users, validating the redesign's success.