Research AI
Generative AI for blogs
Role
Product Design Intern at IBM
Timeline
May - August 2024
Team
VP of Design, Product Managers, Senior Product Designers, and Engineers
Responsibility
Interaction Design, Visual Design, Prototyping, User Testing, and Mockups
Mission
Problem
“
“
I spend a lot of time scanning the research blog just to understand the main idea.”
“
“
Some links lead to long documents with hundreds of pages.”
“
“
I have to leave the blog to learn more about large language models.”
Research
We discovered that generative AI helps people find information faster by condensing articles into a summary at the top.
Google Search
Bing
HMW
How might we use generative AI to help readers discover content they are interested in within blog posts?
Layout exploration
Full page view
Sidebar view
Modal
Full page view
Requires dedicated full page
Sidebar view
Enables readers to ask questions about the blog
Modal
Modal comes and goes as lightly as needed
Requires an extra click to open and close the chat
Challenge
Problem #1
Readers spent 10-20 minutes searching for relevant content within blog posts.
“I spend a lot of time scanning the research blog just to understand the main idea.”
Solution
Takeaways provide readers with a quick understanding of the blog post, saving them time finding content.
Outcome
Stakeholders were aligned and the simplest form of this feature is now launching across all 500 blog posts.
Problem #2
Readers cannot find content they are interested in within blog links.
“Some links lead to long documents with hundreds of pages, which is frustrating.”
Solution
Link previews allow readers to understand content inside links without leaving the page.
Initially, we faced pushback on AI integration due to computational cost.
My proposed MVP provides the same user value without computational cost.
Problem #3
Readers are unable to learn more about specific content.
“I have to leave the blog to learn more about large language models.”
Solution
This solution provides content specific explanations for readers to learn more.
Modal spotlights the blog content
Least eye movement for the reader
Generated content does not save automatically
Difficult to discover at first
Impact
After conducting over 20 in-person interviews with users testing high fidelity prototypes we found that...
of readers found the modal helpful in diving deeper into content they found interesting.
Outcome
All three proposed solutions are now launching across all 500 blog posts.
Takeaways
Link preview
Content explanations
At the end of my internship, I had the priviledge to present my launched features at the MIT-IBM Watson AI Lab.
What I've learned
The importance of proposing MVPs to stakeholders to showcase product impact.
If I had more time
