MSN, recently known as Microsoft Start, is a news and content platform that provides users with news articles, entertainment, lifestyle content, and more.

organization

Microsoft AI



(Windows, Bing, Copilot, MSN)

role

Design, Development,

Research

collaborators

Developers, Project managers,

UX researcher

timeline

2 months

MSN, recently known as Microsoft Start, is a news and content platform that provides users with news articles, entertainment, lifestyle content, and more.

organization

Microsoft AI



(Windows, Bing, Copilot, MSN)

role

Design, Development,

Research

collaborators

Developers, Project managers,

UX researcher

timeline

2 months

MSN, recently known as Microsoft Start, is a news and content platform that provides users with news articles, entertainment, lifestyle content, and more.

organization

Microsoft AI



(Windows, Bing, Copilot, MSN)

role

Design, Development,

Research

collaborators

Developers, Project managers,

UX researcher

timeline

3 weeks

overview

AI-Powered Commenting Experience

An adaptive Smart Phrases feature that personalizes engagement in real time. Using poll interactions as context, it generates relevant comment suggestions or presents diverse viewpoints for users who skip the poll, making it easier for readers to join the conversation.

the problem

From Passive Reading to Active Engagement

The majority of users read articles but rarely engage in commenting. While introducing polls earlier in the year, the polls were quick and low-friction but didn’t drive community discussions by sharing perspectives, reacting to others, and sparking dialogue.

I needed to figure out how to lower the barrier to entry so more readers feel comfortable making that leap from passive reading to active contribution.

I needed to figure out how to lower the barrier to entry so more readers feel comfortable making that leap from passive reading to active contribution.

I needed to figure out how to lower the barrier to entry so more readers feel comfortable making that leap from passive reading to active contribution.

The constraints

Technical and Business Constraints

These constraints guided the design trade-offs that shaped our product vision.

There is currently no seamless experience for renting electric vehicles via an app.

the process

Defining Information Architecture

I defined the information architecture by organizing the interface into clear content and interaction blocks that guide users from context to participation.

the process

Defining User Flow

To visualize how interaction patterns change based on context, I mapped both the signed-in and not-signed-in user flows to show how interaction patterns adapt across contexts.

When discussions have fewer than five comments, AI surfaces full-statement starters to reduce friction; as engagement grows, opinion starters prompt users to complete their thoughts.

the solution

A Smart Commenting Feature That Adapts To Readers In Real Time

A smart commenting feature that adapts to readers in real time. Within the discussion panel, users see three suggestions tailored to how they responded in the poll. This approach makes it easier for people to jump into conversation while also encouraging more meaningful contributions.

Our first launched used full statements which was the fastest way to drive participation and collect baseline metrics. Product managers prioritized speed and wanted to validate demand before adding friction.

But from a design perspective, I knew an open field alone wasn’t sustainable. Without some guardrails, the panel risked being flooded with spam and low-quality posts.

I advocated for friction in the form of opinion starters - subtle prompts that encourages original input and discouraged bots.

The compromise was a hybrid system:

  • MVP (empty state): show full statements so the first users could engage quickly.


  • Active state (>5 comments): switch to opinion starters, guiding readers toward more thoughtful contributions.

Impact & Results

The updated commenting experience increased meaningful participation and improved perceived trust in discussions. Early data showed a higher click-through rate on AI-suggested comments and more sustained engagement across sessions.

Designing with AI surfaced nuanced questions around authorship, tone, and user agency. Collaborating with PMs and researchers taught me how to translate abstract AI capabilities into intuitive interaction patterns. It reminded me that good design doesn’t just make participation easier, it makes it feel authentic.

The updated commenting experience increased meaningful participation and improved perceived trust in discussions. Early data showed a higher click-through rate on AI-suggested comments and more sustained engagement across sessions.

Designing with AI surfaced nuanced questions around authorship, tone, and user agency. Collaborating with PMs and researchers taught me how to translate abstract AI capabilities into intuitive interaction patterns. It reminded me that good design doesn’t just make participation easier, it makes it feel authentic.

The updated commenting experience increased meaningful participation and improved perceived trust in discussions. Early data showed a higher click-through rate on AI-suggested comments and more sustained engagement across sessions.

Designing with AI surfaced nuanced questions around authorship, tone, and user agency. Collaborating with PMs and researchers taught me how to translate abstract AI capabilities into intuitive interaction patterns. It reminded me that good design doesn’t just make participation easier, it makes it feel authentic.