UX research and design sponsored project rethinking how knowledge and tools get discovered across a large organization.
Role: Researcher + Designer
Impact:
This was a sponsored capstone project where I worked as a researcher and designer on Amazon's IT team to investigate how knowledge and tools get discovered across a large organization, in partnership with Amazon and under the mentorship of Amazon UX stakeholders.
Across the project I contributed to the full research lifecycle: survey design, interview facilitation, usability testing, and synthesis and led design on the platform's Slack integration, which I built solo and deployed as a working prototype using a live AI API.
The work was well-received by our Amazon stakeholders, and the project was greenlit for implementation, with an IP transfer meeting proposed with VPs and Directors to move it forward.
This project gave me hands-on experience leading research inside a large, unfamiliar organization, translating findings into a concrete design direction under real stakeholder pressure, and building and shipping a working technical prototype on my own.
Course: HCDE 593: Capstone │ Winter/Spring 2026
Team: Seamus McPeake, Amy La, Bri Tran, Trushaa Ramanan, Zainab Tanveer
Timeline: 6 months
Methods: Usability Testing, Surveys, Interviews, Synthesis, Prototyping
Tools: Figma/Figma-Make, Qualtrics, Slack API, Anthropic API, Railway, Claude
The problem
Amazon employees are constantly discovering new tools and workflows at work, but that knowledge rarely travels beyond their immediate team.
The research (survey)
What drives tool discovery across the organization, and what gets in the way?
Can we leverage Gen-Z employees to spread new tool discovery across the organization?
WE WERE APPROACHED WITH A PROBLEM:
What leadership wanted to understand:
Gen-Z employees will be more motivated to share new discoveries with others than employees from other generations.
Stakeholders came to us with a hypothesis:
WHAT WE WANTED TO UNDERSTAND
Role distribution and how common sharing behavior is overall
Whether generation predicts sharing behavior
General sharing frequency
94% had shared a tool or workflow at least once
Generation wasn't shaping up to be a meaningful differentiator in sharing behavior or motivation
What it showed:
With the question reframed, we set out to understand:
Designed to capture role distribution, sharing frequency, and what prompts someone to look for a new tool, alongside age, so behavior could be compared across generations.
Survey
300+ Amazonians
AN ONLINE PLATFORM
Why people share, or don't, in their own words
How trust recommendation actually gets built
What is motivating individuals who do share to do it
What sharing looks like in practice, day to day
Where sharing is happening
Amazonians are already motivated to share with each other, they just don't have a platform to do it
Certain people like to share, certain people don't
Most individuals share with people they're familiar with
Sharing is happening in Slack
An early pivot
We brought this data to our stakeholders who presented the hypothesis, and they wanted to narrow the lens further, from generation to technical vs non-technical employees.
Pushing back
In addition to the survey, I ran some secondary research about Gen-Z and their link between social media and motivation to share.
Like the survey, my findings weren't shaping up to be a meaningful differentiator in sharing behavior or motivation.
What we found
I pushed for the team to step back instead of reframing the same assumption. If generation wasn't the differentiator, it might be better to focus on finding out what is motivating them, if any of them are motivated.
The reframe
The research (interviews)
What we set out to learn
7 interviews consisting of a mix of ICs and managers across multiple organizations and roles. I facilitated or took notes on 3 of the 7.
Who we talked to
What we heard
7 interviews
The research (key findings)
Amazonians are already motivated to share
Other important findings:
Coding, affinity mapping, and thematic analysis across the survey and interview data pointed to one consistent thread, regardless of role, org, or generation.
The barrier wasn't motivation, it was absence of infrastructure to support them doing it.
Synthesis
Homepage/Recommending
Trust lives in the recommender, not the tool
Most individuals share with people they're familiar with
Sharing is happening in Slack
We created 4 behavioral profiles, grounded in patterns that came directly from what we heard and in observed actions.
We found the Nudge Contributor behavioral profile to be the most important and untapped group.
Behavioral Profiles
What we built (Platform)
A homepage centered helping visitors find tools that will help them with their role, each tagged with category and org-approval status. There's also a "based on your recent tool usage" section suggesting relevant workflows, sometimes including embedded video walkthroughs.
We built an online platform where users can discover new tools, leave feedback about tools, give recommendations to others, and see how others interact with their recommendations across organizations.
Tool & Workflows
Users can discover and download tools, see who is recommending each tool, and what recommenders are using the tool for.
Workflows go a layer deeper: step-by-step guides tied to a real contributor, with tips, supporting files, and a Q&A thread where people ask how others are actually using it in practice.
Analytics
A personal dashboard showing contributors how their recommendations are performing: views, "found useful" counts, and how many teams and orgs a shared workflow reached.
Workshops
Contributors can host live sessions tied to their most requested workflows, with others registering directly from the workflow page turning a written guide into an optional real time walkthrough when people want more than documentation.
Usability Testing
What we tested:
We conducted testing on Amazonians in a mix of our behavioral profiles, organizations, roles, managers vs IC's, and ages.
Who we tested
Navigability of the dashboard
Discoverability of key features
Trust signals and recommender visibility
Sharing flow and friction points
12 Amazonians
What changes as a result
Removed an early manager specific analytics view after participants flagged it as surveillance-adjacent
Adjusted discoverability of function for users looking to contribute to platform
FOLLOWING UP USABILITY TESTING
Why a Slack-Bot?
What we built (Slack-Bot)
Our research showed Slack was already where sharing and discovery happened informally. Rather than asking people to adopt a new destination, we designed the platform's entry point to meet them where that behavior already existed.
Our usability testing sessions were showing that users were having a difficult time discovering and engaging with features on our webpage which encouraged them to contribute recommendations of tools they use.
So a Slack bot became the place where our "Nudge" or encouragement to contribute would live.
How I built it
I self assigned myself to build the bot and built it over the course of 2 days. I used Claude to code the bot and deployed the code through Railway, running some of the AI features of the bot bot on Anthropic API and linking the bot into Slack's API.
I created a private Slack channel to test the functionality of the bot with my teammates, so the bot ended up being fully functioning on its own.
"Slack is where sharing already lives."
Nudge feature
The first feature is a "Nudge" that informs users a recommendation of a tool they frequently use could be helpful to others and allows them to post a recommendation through the bot.
Recommendation detection feature
The bot has the ability to detect when a recommendation has been made between users in Slack. The bot interjects and asks users who have made recommendations if they would like to post a recommendation for a tool to the dashboard.
Users can post these recommendations through the Slackbot and it will appear on the dashboard.
Tool recommendation requests
Users can request recommendations for new tools through the bot utilizing the AI's API. Users describe what they're looking for and the bot will give recommendations based off of what tool library the dashboard currently has.
This utilized Anthropics API but during integration, Amazon would preferably implement their own.
What came out of this?
By the numbers
Outcome
The project was greenlit for implementation at Amazon, with an IP transfer meeting proposed with VPs and Directors to move it forward. That meeting was ultimately not held due to scheduling conflicts during a busy period at Amazon, and the project's status beyond that point is unknown to me.
Regardless, as a team we received overwhelmingly positive feedback on the project from stakeholders on the Amazon IT team. See some testimonials below:
"The HCDE team reframed our problem space in ways that will directly influence our product strategy. Their research uncovered real user insights, and their design concepts pushed our thinking well beyond incremental improvements."
— Yuwei Li, UX Designer, Amazon IT Services
300+ survey responses
7 interviews
12 usability testing sessions
"V Studio has armed us with both data and direction to develop a system that will change how Amazonians discover software, and in turn increase productivity. The impact of this project is hard to measure but very significant."
— Mike Berg, UX Manager, Amazon IT Services
Reflecting on the project
I'm very grateful I got the chance to work with Amazon on this project and would like to thank them for sponsoring us. I'd like to thank the Amazon IT team specifically for their time and encouragement to me.
Additionally I would like to thank everybody on my team (V-Studio). Everybody on my team was great and had amazing contributions to this project.