Stealth Startup

a next-generation AI platform to bring the "auto-complete" experience to everyday business users and automate all mundane and repetitive tasks.

My Role
Team
Time

UI/UX  Designer(Contract)

UpStudio

Jan 2023 - Mar 2023

1 Overview

Context

Did you know that a staggering 60% of jobs spend at least 30% of the time on tedious and repetitive tasks, and workers long to get them automated?

AI automation can be applied to a wide range of tasks and industries, from manufacturing and logistics to healthcare and finance. The goal of AI automation is to make processes more efficient, accurate, and cost-effective by leveraging the capabilities of AI.

Challenge

A financial specialist in the Google finance team needs 60 mins to extract 40+ key terms from the contract to the Google sheet, with over 100K contracts/year.

Currently, a contract processing workflow needs the finance team to spend hours extracting key data from vendor contracts and comparing them to invoices before payment. This is a time-consuming and error-increasing process.

How might we help finance operation specialists review and process contracts more efficiently with AI automation?

Solution

An AI-powered contract processing automation product that aims to improve document handling productivity for enterprise users.

For this version, our client will work closely with Google (Business Process Outsourcing team), their primary customer, to discover automation opportunities to automate their contract processing workflows through AI suggestion and validation functionalities.

I will help Orby AI to navigate the user experience issues in AI suggestion and validation flows.

2 Research

Stakeholder Interview

I interviewed Finance operations specialists and Orby AI Engineers to gather in-depth qualitative data. I then created a user journey map to understand their contract reviewing process and the pain points they face.

  • Goal 1:
  • Understand User
  • Understand the contract processing workflow and current pain points from users.
  • Goal 2:
  • Understand AI Technology
  • Understand what can AI suggestions provide, and what is AI's strengths or limitations.

Pain Points:

  • Time-Consuming Reviews
  • Spend large amounts of time manually reviewing contracts and other documents.
  • Tedious Process
  • Tedious reviewing and keyword extracting tasks would prevent users from focusing on more strategic work.
  • High Review Errors
  • Manual review processes could result in errors and risks. Contract review values accuracy a lot.

Competitive Analysis

There is no other solution exactly like Orby AI. To help the Google finance team to develop its vision, I did a competitor analysis to understand the market landscape and product differentiators. I tried to use RPA and Google Doc AI together, which did not solve their problems.

Key Insights from the competitive analysis:
What Orby AI needs

  • Seamless Integration
  • Users need a more integrated and connected experience.
  • Actionable Suggestions
  • We should make the suggestion results more actionable to users.
  • Prioritization Focus
  • We should help users prioritize the items to review.
  • Guided Navigation
  • Navigation patterns between key terms and original document.

Define

Sharing the research insights, I organized a design sprint by partnering with the PM and engineer team, to define design focuses together.

How might we help finance specialists review and process contracts more efficiently with AI automation?
  • Efficiency
  • How do we optimize the suggestion validation workflow?
  • Trust
  • How might we build user trust AI suggestions?

3 Ideate

User Flow

Keeping the two goals in mind, I created the User Flow to design the overall workflow and how Orby AI interacts with the user.

Design Challenges Breakdown

Start a Task

Review Terms

Train Orby AI

Manually add a task

Guided experience

Engage users

Simplify understanding

Build confidence

Efficient process

Edge cases

4 Start a Task

Idea 1: Side panel

Idea 2: List inside Orby

Design Challenge 1:

How to manually add a task?

How do we help users to quickly identify different files with complex information? For unrecognized files, how can users manually add them?

Design Challenge 2:

How to simplify users understanding?

How do we determine the priority of the information presentation? How to layout and change the interaction of collapsing and expanding to help users understand the information?

Final Design

5 Review Terms

Wireframe

Consider all the filters and grouping
Make workflow as customized as possible
Chat robot to keep users engage

However, after the usability testing, I found that lots of users got confused and didn't know what to do first when facing all the options.

Design Challenge 1:

How do we optimize the key term suggestion and validation interaction flows through a Guided Experience?

Idea 1:

Idea 2:

  • Three cards give clear guidance for the user to choose what to do first.
  • Lack of context when user editing inside the panel.
  • Icon buttons are small and might easily miss press.
  • Able to see context when editing the file.
  • Takes too much space.
  • Low confidence button is small.

Design Challenge 2:

How do we build trust for AI suggestions during the review process?

Action 1: add confidence score

Action 2: add congratulation page

Design Challenge 3:

Design for edge case

Report a question

Add multiple inputs

Extract from table

First-time tutorial

Final Design

6 Train Orby AI

Design Challenge 1:

How to engage user during AI training?

Final Design

7 Takeaways