Engly Chang
Product Designer

What

Enabled every xDesigner ship production code

"The late AI adopter's cheat sheet to catch up." Recognizing the choice paralysis caused by the rapid explosion of AI tools across Meta, I published an internal prioritization playbook. This framework gave teams a clear baseline to evaluate new technology, helping them align on which AI tools to adopt and master.

AI tool framework

Authored a guide aimed to help non-technical people ship code. I also ran a 3 hour workshop that helped people use the guide, which resulted in 100% of "designers" (Product Designer, Content Designer and User Researchers) shipping production code.

Impact

100%

xDesigners shipped within the team

What

External webinar prototype

Leveraging an internal AI tool, built a tightly scripted product prototype for an external webinar illustrating the direction the product team is heading.

Impact

14 days → 3 days

Turn around time in comparison to typical 2 week turnaround

Problem to solve for

Automation scaling

Risk review tends to be complex and with terms that the average person might not understand. As the platform moves towards automation, the hypothesis of integrating a trained AI modal that understand risks, could help improve people's understanding overall.

What

Help agent integration

Foundations in the Risk Org owns the tool that every team at Meta has to use for risk assessment in order to launch any feature. Led design and on-time launch of Help AI Agent, advocating for platform retention to better understand user needs. Identified need for "omni-state" through comprehensive flow exploration.

Impact

100% launch

Design to ship in ~3 weeks

What

Agentic "help" integration

Prompt construction–three use cases covered for a comprehensive end to end experience with three types of entry points.

Insight

Omni state

Discovery and design of initial "omni state" made subsequent pivot more efficient

Problem to solve for

Misunderstanding of AI

Since AI recommendations were introduced, we were seeing a problem where people assumed all factors have were scanned by AI but they're not. Risks were missed with a higher rate with recommendations—we set out to solve this with design while the AI model improves.

What

Multi-prong approach improvement

Update design to help risk factor understanding, expanded explanation and dynamic categorization and organization. Optimize design for AI run latency and communicate to users where they are in the process.

Impact

~17% ⟶ ~5%

Failure rate reduced reduction across shipped features

Problem to solve for

Redesign of risk factor page

Update design to help risk factor understanding, expanded explanation and dynamic categorization and organisation. A range of explorations are done from less friction to most engagement needed from the users.

Impact

11%

Decrease in failure rate from risk factors page redesign

Problem to solve for

Risk factor optimization for AI

The improvements are aimed to help users understand AI automation better, and complete the complex review they need to go through with more accuracy. While it's a suite of end to end features, the main surface is on the page product owners has to review if a risk applies to their project.

Impact

21%

Reduction in audit fail rate through AI latency improvements

What

Horizon Worlds & Quest OS merge

In 2024, Horizon's social systems began merging with Quest OS. Horizon Worlds gained priority access to core social surfaces, and design work was needed to bridge the gap while helping people transition smoothly.

Problem to solve for

Duplication of settings

Across VR and Horizon Worlds, many product surfaces — avatar, friends, privacy — existed in duplicate. Part of the effort involved migrating services already built for Facebook and Instagram.

Impact

70%

VR OS integration for the unified friends flow

Neutral

Privacy migration launch landed with neutral engagement impact

What

Facebook Stories: Interfaces

FB Stories is used daily by 1/3 of the world. Interfaces is FB Stories's multi-surface expansion strategy. I led design on FB Lite, web, Messenger and emerging wearable platforms. FB Lite is a lightweight interface on feature phones on Androids. It's optimized to run on phones with less processing, slower connection, and the app takes up little storage. It's so popular with the rest of the world outside of U.S., in fact, it's the second largest interface for Facebook.

FB Lite: Top of Home

A project where we tested to bring a collection of content to top of home instead of just the Story tray.

Insight

7+ experiments

Each led to mitigation experiments that further gave us learnings on changes

What

Concept and two card experiments

We first focused on bringing a Reels entry point to top of Home. Replacing the Story tray with a two card system, adding Reels to the mix. Stories tile shows options from full Stories features to most simplified UI.

Insight

The changes are too drastic and need to be dialed back

What

Adding Reels to Stories only

The next set of experiments dial back on replacing the entire Stories tray. They focus on understanding the implications of adding a Reels consumption entry point to the eco system while tweaking the prominence of Story creation.

Insight

When Reels replaces the first Story production tile, it reduces Stories production

What

Mobile & FB Lite: Feature development

Over the years, we iterated on features like:

• "Peek", interactive stickers, and bulk upload stories.
• We shipped a Story only surface on Lite. Design iteration on how we can continue to develop this dedicated surface.
• Sprint on FB iOS/Android to explore audience control and scheduled sharing

... and many more

Impact

+347MM

+1.365M

+347MM incremental views (vs. 180MM goal / 250MM stretch)

+1.365M incremental Original Exclusive Stories Production Daily (vs. 1M goal)