"The sheer volume of contracts and invoices we process manually is overwhelming. It feels like we're constantly drowning in paperwork, leaving little to no time for strategic analysis of financial data."

"The sheer volume of contracts and invoices we process manually is overwhelming. It feels like we're constantly drowning in paperwork, leaving little to no time for strategic analysis of financial data."

- Yinkei, A Google Finance Specialist

Orby AI - Say Goodbye to Repetitive Tasks

Streamlines those repetitive tasks like auto-completing your entire workday. Contracts, invoices, data entry – all handled with the magic of AI.

My Role

UX Design Lead

Team

UXDs: Sabrina Zhu, Wenbo Guo, Chenxin Sun; SWEs: Will Lu, Tianjun Fu

Duration

Jan 2023 - Mar 2023

Release Date

May 2023

Challenge

The Google BPO team was overwhelmed by the manual processing of over 500,000 documents annually. Their existing workflow was plagued by inefficiencies—slow, error-prone, and unsustainable.

Results

I led the design of Orby AI, revolutionizing these cumbersome workflows into a streamlined, intuitive experience. By engaging stakeholders throughout the process and refining the solution through continuous feedback and user testing,


we achieved remarkable results. Orby AI delivered an 🚀 85% productivity boost, a 4X ROI in just days, and transformed the entire user experience.

Highlights

Effortless Learning

We designed a learning experience that was as intuitive as it was effective. By incorporating quick tips and a step-by-step tutorial, we minimized the effort required to get users up to speed, ensuring they could start benefiting from Orby AI almost immediately.

Smart Task Management & Prioritization

To tackle the issue of task overload, we introduced a Smart Suggestion Panel that not only notified users of new tasks but also suggested priorities based on urgency and impact. Automatic file syncing streamlined their workflow, and the ability to manually add tasks allowed users to maintain control over their workload.

Streamlined Validation & Reduced Errors

We knew that accuracy was critical. Therefore, we reimagined the data review process with simplified categories and tabbed filters, optimizing the layout to enhance readability. This change alone significantly reduced the cognitive load on users, leading to more accurate and efficient data processing.

Meet Amy

Armed with these insights, we created a user persona to illustrate the need to address these core user anxieties.

BPO TEAM’S Journey

With a tight three-month deadline looming, we faced a critical challenge: prioritize functionalities that would deliver the most significant impact for Google's BPO team processing contracts. To tackle this head-on, I crafted a user journey map which visually highlighted the most pressing areas for our initial focus.

Uncovering issues
Uncovering issues

We met with Google's BPO team to understand their day-to-day operations.

On average, each team member handles around 15 contracts and invoices daily, all of which entail the daunting task of manual processing in their workflow. Therefore, developing a solution to streamline these tasks became a key focus.

Inefficiencies from switching between tools

BPO specialists wasted valuable time navigating between emails, Drive documents, and spreadsheets. This fragmented data ecosystem caused inefficiencies and hampered their ability to complete tasks quickly.

Errors Haunt Manual Extraction

Manually extracting data points from contracts resulted in inaccurate data impacting downstream tasks and financial planning, causing significant downstream issues.

Time Crunch & Missed Deadlines

The BPO team, felt overwhelmed by the sheer volume of tasks and the time-consuming nature of manual data extraction. This often led to missed deadlines and increased pressure.

Key Takeaways

Our user research revealed three insights:

Seamless Contract Processing

Users want a smooth, uninterrupted workflow that eliminates the need to switch between multiple tools or platforms during contract processing.

Reduced Error Rates

Accuracy is crucial. Users seek tools that minimize manual errors, ensuring that data is captured and processed correctly the first time.

Minimal Learning Curve

Users prefer a system that is easy to learn and use, with intuitive interfaces that don't require extensive training or prior technical expertise.

The Backbone of the project

Armed with key insights from our user research, I spearheaded a design sprint in close collaboration with our product manager and engineering team. We tapped into the development team's deep expertise, delving into the nuances of AI's capabilities and constraints.

Together, we asked the critical "How Might We" questions at every stage of the user journey, carefully crafting a seamless experience that harmonizes user needs with the full potential of AI.

Taming the Workflow Beast

The design sprint exposed the unconventional nature of the existing workflows. Instead of jumping to solutions, I prioritized understanding user experience. Through in-depth daily workflow observations, I pinpointed areas of friction and the "chaos" users faced within the process.

Eliminating Context Switching

Our research revealed a major hurdle for users: the constant need to switch between different tools to complete tasks. This context switching significantly slowed them down and caused frustration.

To visualize the extent of this problem, we created a user workflow that mapped out the disruptive nature of context switching. This helped us focus on a key solution - redesigning the workflow for a centralized experience, minimizing tool switching and streamlining the entire process.

Layout Patterns

While designing the suggestion panel, a crucial challenge emerged:

Information Overload at First Glance: As the starting point of a Chrome extension and a frequently used function, the panel needed to present a large volume of information clearly and concisely to avoid overwhelming users.

Eye-tracking Patterns

Optimizing the review layout for eye-tracking patterns is crucial for ensuring a smooth and efficient user experience, especially considering this is the core functionality of the Orby AI platform. When reviewing extracted terms and comparing them with the contract document, users typically follow a specific visual path:

Refining Onboarding Journey

The initial focus for Orby AI was clear: Streamlines BPO specialists' workflows. However, during design sprints with stakeholders, a crucial element emerged – the onboarding experience.

As we meticulously mapped the user flow, a potential roadblock emerged: the traditional 100-200 contract training process. Recognizing this could hinder user adoption, we prioritized exploring alternative approaches.

Unveiling Edge Cases

While streamlining the core review process was essential, we also focused on mitigating potential data accuracy issues arising from edge cases. These situations include:

Long Paragraph
“Not in Doc”
Discrete Content
Table Structure
Reflection
1. Iterate, Iterate, Iterate!

This project underscored the value of rapid iteration. Quickly refining designs and seeing immediate results was a game-changer. Moving forward, I’m inspired to think bigger and iterate boldly.

The Impact & A Look Forward:

Orby AI has revolutionized document processing by automating workflows of any complexity and driving significant efficiency gains. Looking ahead, there’s immense potential to scale these solutions, further enhance automation, and continue pushing the boundaries of what's possible. The journey doesn’t end here—it's just the beginning of redefining how complex workflows are managed with simplicity and precision.