Product
B2B SaaS
My role
Design lead
Product Manager
Team
3 Engineer
1 Founder
1 AI Researcher
Timeline
3 months
Overview
Overview
About Holistic AI
Holistic AI is an end-to-end AI governance platform that helps enterprises manage AI risks, ensure compliance, improve transparency, and enhance the quality of their AI systems through automated assessments, risk mapping, and continuous monitoring.
.png){kind=small}
Request from client
"There are just too many questions—and most of them repeat what we've already documented internally. Can't this be easier?"
Old Onboarding Experience Audit
RESEARCH
RESEARCH
To uncover the true pain points, I initiated and led our first-ever user interviews—convincing the team of their value along the way.
01 User Interviews and Insights
02 User Personas
I created the team’s first set of user personas, giving everyone a clear, shared understanding of who our users are, what they need, and what they’re looking for—for the first time, we were all aligned around the people we were designing for.
Our primary users were admin-level accounts—typically AI leads, compliance managers, or technical project owners—who were responsible for onboarding AI projects and completing the Risk Mapping process.
Problem Statement
ideation
IDEAtion
I explored possible ideas and discussed with the team to check the feasbility and constraints
⬇️ ⬇️ Check out the comic below to follow the collaboration progress. Hover over the image and use the arrows to browse through all the pages.
FEATURES
features
Auto-onboard
-
Automatically onboard AI systems by scanning existing documentation, either through user-uploaded files and links or via auto-scanning connected third-party integrations like Confluence and Google Drive.Provide a template to help users create well-structured documents
-
Add a team collaboration feature for shared input and review
-
Require a final review step before submission to ensure answer accuracy
Manual-onboard
-
Remove non-essential project questions and allow users to complete them later, 20+ questions -> 3 questions
-
Retain all questions in the risk mapping questionnaire, as they are essential
-
Enable auto-save for incomplete responses to prevent data loss
-
Add a to-do list for imcompleted tasks
SOLUTION
Solution
Use case 1
Auto-onboard by AI via uploading documents
For users with existing documentation on their AI projects or systems, they can upload files or provide website links, allowing our AI system to scan and identify the AI projects. The system will automatically populate all project details and answer the Risk Mapping questions, requiring users to simply review, edit if needed, and submit to receive the Risk Mapping report.
Use case 2
Auto-onboard by AI via third-party integrations
The third-party integration feature allows users to connect tools like Confluence and Google Drive to our platform. Once connected, the system automatically scans linked folders on an hourly basis, detects AI systems, and pre-fills project details and risk assessment answers—reducing manual effort and keeping information up to date with minimal user input.
Use case 3
A better manual onboarding experience
For users who don’t have pre-existing documentation on their AI projects or prefer to use the questions as a guide to structure their projects, they can manually answer all questions and submit them to receive a Risk Mapping report.
USER FLOW
VISUAL DESIGN
VISual design
01 Design System
I created the first design system for Holistic AI, bringing consistency, scalability, and visual clarity across the platform—laying the foundation for a more unified and efficient product experience.
02 Screens
WELCOME PAGE
Add a welcome page for new customers that guides them directly into the project creation flow for a more streamlined experience
GET STARTED BUTTON
Add a "Get Started" button to the left navigation bar to let users quickly and easily start a new project
SELECTION
-
Allow users to choose between automatic onboarding and manual onboarding based on their preference
Auto-Onboard
upload files or use url to link docs
projects recognized by AI
Statuses for AI systems
-
There are four statuses for AI systems:
-
Pending: The system is still scanning documents and extracting information to complete the questionnaire.
-
Ready to Onboard: The AI system has been recognized and the questionnaire is complete. Users can now review the risk level and system details to decide whether to onboard it as a formal project.
-
Onboarded: The AI system has been officially onboarded as a project, and the assurance process has begun.
risk details
The risk level is preliminary and can be enhanced by implementing the suggested mitigation steps
We aim to provide clear explanations of the risk level to reduce user confusion and build trust in how it is determined.
We also break down the overall risk level into different categories to provide users with a more detailed explanation.
questionnaires
Manual-Onboard
SKip project details for manual onboard
-
When choosing manual onboarding, users can skip entering project details and go straight to answering the risk mapping questionnaire.
Dashboard
Project detail page
-
A detail page to display important information about an AI system
-
A to-do list provides quick navigation to the imcompleted tasks
Assurance Journey Map
-
Clearly highlight each step of the assurance process along with its completion rate.
-
The side panel offers quick access to any specific step.
edge case
EDGE CASE
No answers found from docs
-
If the AI can’t find relevant answers in the uploaded documents, it will leave those fields blank rather than making assumptions. We’ll also add visual cues to clearly remind users to review and complete any unanswered questions before submitting.
AI doesn't recognize AI systems
-
Since AI is not 100% accurate, there may be cases where it fails to detect an AI system even when one exists. In those cases, we’ll provide users with the option to manually onboard the project anyway, ensuring they maintain full control over the process.
impact
IMPACT
