2024 · IBM

Generative AI for the IBM Research blog

Designed and shipped three generative AI features, helping over 90% of readers increase reading comprehension.

My Role

I was the UX Design Intern leading the experience for 14 weeks from discovery and implementation, iterating closely with Design, PM, and Engineering.

I led 20+ interviews to identify key pain points, facilitated design sprints by iterating on prototypes based on user testing to present to 12 stakeholders, and launched three user-centered genAI features.

Goal

IBM Research blog is the main destination with over 100,000 visits per month and 500+ blog posts. Our goal is to leverage the platform to enhance the readers’ experience through generative AI.

A collection of articles on the IBM Research blog

Challenge

After interviewing 20 users:

  • 18/20 readers had difficulty finding relevant blog posts.
  • Researchers struggled to find relevant content inside blog links.
  • 50% of readers dropped off halfway, losing engagement.

Opportunity

How might we help readers engage with blog content they are interested in?

Problem #1

Readers spent 10–20 minutes searching for relevant content within blog posts.

“I spend a lot of time scanning research blogs to understand the main idea.”

Solution

Takeaways provide readers with a quick understanding of the blog post, saving them time finding content.

IBM Research blog with an AI-generated takeaways panel
Using the IBM Carbon AI design system.

Design vision

Allows clients to explore blogs they are interested in and increases retention by showing relevant blogs at the summary level.

IBM Research blog with an AI-generated takeaways panel
Using the IBM Carbon AI design system.

Problem #2

Researchers struggled to find relevant content inside blog links.

“Some links lead to long documents with hundreds of pages, which is frustrating.”

Iteration #1

Provides a summary of the link to help readers understand the link without leaving the page. However, the position of the preview required too much eye movement.

IBM Research blog with the first takeaways-panel iteration

Iteration #2

Summary was too long. Readers wanted a quick glance to understand the blog link.

IBM Research blog with the second takeaways-panel iteration

Solution

The link preview highlights the key part of the link summary to help users quickly understand.

IBM Research blog with the final takeaways-panel solution

North star

Gives users flexibility to ask any questions about the blog, but requires more token costs.

IBM Research blog with the first takeaways-panel iteration

Problem #3

Clients and researchers struggled to discover new blog posts.

IBM Research article with newly published blog cards highlighted at the bottom
Only 10% of readers end up scrolling to the bottom where there are new blogs to discover.

Iteration

Side position was distracting to the readers when reading.

IBM Research blog page with a related blogs panel

Solution

By revealing relevant blogs in between sections, we enhanced discoverability and retention by 90%.

IBM Research blog article with a related blogs card revealed between sections
IBM Research blog page with a related blogs panel

Impact

What I’ve learned

Framing the MVP as a starting point drove team buy-in and defined a plan to execute the design vision.

If I had more time

I would explore the C-suite’s problem space to design solutions tailored to their priorities.

Next Project: Pixora