Work

Work

Work

Xbox, Summer 2026

BerkeleyTime, 2026

Microsoft Fabric, 2025 - 2026

Dell Technologies, 2025

Rooh, 2025

More

More

More

Resume

Resume

Resume

Overview
Case Studies ->
Reflection
Overview
Case Studies ->
Reflection

Microsoft Fabric

Redefining onboarding and AI integration for Microsoftโ€™s data analytics ecosystem

Case Study

6 minute read

Macbook Pro

Role

Product Design Intern

Team

5 Consultants

2 Project Managers

1 Senior Product Designer

1 Creative Director

Timeline

Jan 2025 - May 2026

Skills

Product Design

Prototyping

AI UX

Data Systems

01
TLDR

Fabric is Microsoft's unified data analytics platform, used by 31,000+ organizations to move, engineer, and report on data in one place. It's powerful, but it shows you the data and nothing about where to go next or who is behind it: new users got lost across six overlapping workloads, and teams broke each other's reports without knowing who owned what.

In Spring 2025, I found the blocker was clarity, not complexity, which became three design principles and [your 2025 prototypes]. Microsoft brought us back in Spring 2026 to close the second gap: I designed three Copilot features, an agent change-approval flow, a dependency risk heat map, and workspace memories, that make ownership visible while keeping the human in control, validated in 7 interviews and 2 usability tests. [Result: what Microsoft took forward, or the strongest feedback you got.]

02
Context
What is Microsoft Fabric?

A unified data analytics platform that brings together data movement, data engineering, data science, and business intelligence all in one place

I worked in

these platforms!

Data Factory

Data Factory

Data Engineering

Data Engineering

Data Warehouse

Data Warehouse

Data Science

Data Science

Real-Time

Intelligence

Real-Time

Intelligence

Power BI

Power BI

Partner & Industry workloads

Partner & Industry workloads

Copilot in Fabric

Copilot in Fabric

OneLake

OneLake

Microsoft Purview

Microsoft Purview

Mapping every path a user could take

outlining user flows and feature journeys across Fabric's workloads

powerbi

warehouse

science

engineering

factory

real time intelligence

Making the human ownership & coordination behind data systems visible and navigable for enterprise teams.

SOME PARTS OF THIS CASE STUDY ARE PASSWORD PROTECTED

If you'd like to read more than what's presented, please reach out to me! karenhclai@gmail.com

Go to NDA Portion

02
The challenge

How can Microsoft Fabric integrate collaboration by making the human ownership and coordination behind data systems visible, automated, and navigable?

01

Collaboration

What does effective collaboration actually look like in the modern AI age? How does collaboration look in accordance to agents or humans?

02

Ownership & Coordination

Who is responsible for what when things break or change?

03

Copilot

Automation

Where's the balance between useful automation and eroding trust or clarity?

04
Ai Workflow

We sketched by hand, then used GitHub Copilot to turn sketches into working, code-backed prototypes inside the Fabric UI Kit. That let us explore far more directions and put them in front of users sooner.

03
What we designed

BEFORE a change happens

Dependency Risk

Dependency Visibility

See what a change will break before you make it.

Fabric MCP Integration

Dependency Visibility

Designing for human โ†’ agent (how much can your teammate see that your agent changed?) & human โ†’ human (how are changes you didn't make yourself surfaced for leaner collaboration?)

Workspace Handoff

Owner Departure

What walks out the door when a teammate leaves? Their context stays for whoever picks it up.

The comp analysis showed every tool loses context at the same two points: when an owner leaves, and when a change ripples downstream. We broke that into three moments to design for: before a change (see what it will break), as a change happens (see what an agent edited, and approve it), and after someone leaves (keep what they knew).

02
Finalizing the problem industry across competitors

How can Microsoft Fabric integrate collaboration by making the human ownership and coordination behind data systems visible, automated, and navigable?

Market-specific

Collaborative-focus

โ†’ Owner Departure

โ†’ Dependency Visibility

The work
Redefining onboarding and AI Integration for Microsoft's Data Analaystics Ecosystem

SOME PARTS OF THIS CASE STUDY ARE PASSWORD PROTECTED

If you'd like to read more than what's presented, please reach out to me! karenhclai@gmail.com

Go to NDA Portion

02
The challenge
Everything works,
but doesn't connect.

Overlapping Features

Fabric has multiple features that do similar tasks for slightly different use cases and users. How can we unify these features and provide one cohesive solution?

AI Ubiquity

How can Fabric implement Al in a ubiquitous way, how can we make it invisible/non- intrusive to the user?

01

AI Feature Uncertainty

Users were unaware of how AI was embedded in the platform and unaware of AI features.

02

Task Completion

Users struggled to complete certain tasks due to their uncertainty about what steps to take next or where to find necessary information.

03

Onboarding

Unintuitive navigation of tools/workspaces, leading to reduced productivity and frustration.

How might we simplify Fabricโ€™s onboarding and AI flows to help users navigate complex data systems with clarity?
03
What we found
Complexity wasn't the problem. Clarity was.

What did we take away from the Information Architecture map and user journey?

Overlapping features

We found that the challenge isn't complexity itself but inconsistent structure and terminology.

Data Engineering

Data Factory

Fragmented workloads

The 6 workloads function well individually, but feel disconnected.

complex users, unclear flows

Fabricโ€™s users are technically skilled

Lack of clarity (not data depth!) causes onboarding friction and feature drop-offs.

04
Real user feedback
16 users, from senior engineers to founders

job roles

data field

Senior Data Engineer

Analytics Engineer

Data Engineer

FIEld Solutions

Field Engineer

Lead Solutions Architect

non Technical

Business Analyst

Contractor

Founder

by the numbers

where I sourced interviews and and how many I completed

16

interviews

experience

usersโ€™ experience with Fabric and/or PowerBI

< 2 years of experience

5+ years of experience

2-5 years of experience

The work
Redefining onboarding and AI Integration for Microsoft's Data Analaystics Ecosystem

SOME PARTS OF THIS CASE STUDY ARE PASSWORD PROTECTED

If you'd like to read more than what's presented, please reach out to me! karenhclai@gmail.com

Go to NDA Portion

02
The challenge
Everything works,
but doesn't connect.

Overlapping Features

Fabric has multiple features that do similar tasks for slightly different use cases and users. How can we unify these features and provide one cohesive solution?

AI Ubiquity

How can Fabric implement Al in a ubiquitous way, how can we make it invisible/non- intrusive to the user?

01

AI Feature Uncertainty

Users were unaware of how AI was embedded in the platform and unaware of AI features.

02

Task Completion

Users struggled to complete certain tasks due to their uncertainty about what steps to take next or where to find necessary information.

03

Onboarding

Unintuitive navigation of tools/workspaces, leading to reduced productivity and frustration.

How might we simplify Fabricโ€™s onboarding and AI flows to help users navigate complex data systems with clarity?
03
What we found
Complexity wasn't the problem.
Clarity was.

What did we take away from the Information Architecture map and user journey?

Overlapping features

We found that the challenge isn't complexity itself but inconsistent structure and terminology.

Data Engineering

Data Factory

Fragmented workloads

The 6 workloads function well individually, but feel disconnected.

complex users, unclear flows

Fabricโ€™s users are technically skilled

Lack of clarity (not data depth!) causes onboarding friction and feature drop-offs.

04
Real user feedback
16 users, from senior engineers to founders

job roles

data field

Senior Data Engineer

Analytics Engineer

Data Engineer

FIEld Solutions

Field Engineer

Lead Solutions Architect

non Technical

Business Analyst

Contractor

Founder

by the numbers

where I sourced interviews and and how many I completed

16

interviews

experience

usersโ€™ experience with Fabric and/or PowerBI

< 2 years of experience

5+ years of experience

2-5 years of experience

The work
05
Reflection

In 2025 I learned users weren't overwhelmed by Fabric's power, just lost in its structure. In 2026 I saw they were also losing track of each other. Both years came down to one job: make the invisible legible, and leave people in charge.

Big thanks to the Microsoft Fabric team, especially Gresshaa Mehta and Brent Sandifer for their guidance, collaboration, and trust throughout this project!

Automate the work, not the judgment.

Designing for and with agents changed my question from โ€œwhat can AI do?โ€ to โ€œwhat should a person still decide?โ€ Routine edits can run in the background while you focus elsewhere. I worked the same way: Copilot helps me explore, and I made the calls.

Feasibility is a design material.

Every concept faced one question: is this buildable in Fabric's real architecture? Naming the constraints made the work more credible, not less ambitious

Based in San Francisco Bay Area

Open for hire

Back to home

Copy component

Copied

karenhclai@gmail.com

Open for hire

Back to home

Copy component

Copied

karenhclai@gmail.com