CASE STUDY
Recruiting Hub with AI assistent
Recruiting Hub with AI assistent
Designing a personalized, AI-powered workspace for recruiting
Designing a personalized, AI-powered workspace for recruiting
Recruiting Hub reimagines how recruiters, sourcers, and recruiting coordinators start and manage their day, bringing together critical events, tasks, recruiting activity, and AI-powered assistance in one contextual workspace.
I led the homepage experience from product definition through interaction design, research, validation, and delivery, and contributed to the broader AI experience strategy for how Recruiting Hub could evolve from organizing work to proactively helping users understand and act on it.
ROLE
Lead UX designer
Time
2026
SCOPE
Product definition · Interaction design · Prototyping · User research · Usability testing · AI experience integration
CUSTOMERS
Recruiters · Sourcers · Recruiting Coordinators · High-volume recruiting teams
1. The problem
1. The problem
Recruiting was powerful, but fragmented.
The existing experience centered around a complex recruiting pipeline. It was effective for managing candidates, but users still had to navigate across jobs, scheduling, tasks, talent pools, pipeline lists, and other surfaces to reconstruct a picture of their day.
We saw an opportunity to change the fundamental question the product answered:
Before
Pipeline
Jobs
Worklists
Scheduling
External trackers
QuickSight
…
Recruiter manually assembles context
Opportonity
What’s happening today?
What needs my attention?
Where should I act first?
2. The vision
2. The vision
From a system users navigate to a workspace that understands their work
The vision for Recruiting Hub extended beyond a traditional dashboard.
We wanted to progressively reduce the effort required for users to understand and manage their work:
Organize
Bring today’s events, tasks, jobs, pools, and pipeline activity into one personalized workspace.
Understand
Use recruiting context and AI to help users find information and identify what matters.
Act
Enable contextual assistance that can answer questions and help users complete recruiting tasks without navigating across workflows.
3. Designing for different ways of working
3. Designing for different ways of working
One workspace. Fundamentally different jobs.
Recruiters, sourcers, recruiting coordinators, and high-volume teams share the same platform, but their daily priorities are fundamentally different.
The challenge was therefore not simply deciding what belonged on the homepage. It was designing a system that could provide a useful starting point without assuming everyone worked the same way
Useful by default. Personal by choice.
The system provides role-aware defaults while allowing users to decide what appears, where it appears, and what deserves persistent attention
Role-based default view
Role-based default view
Sourcers & Full Life Cycle Recruiters
High-volume (Campus Hire) Recuiters
Recruiting Coordinators
And personalize further
Homepage configurator
Start with what’s relevant.
Flexible hierarchy
Reorder and reposition content.
Individual control
Show only what matters.
Homepage configurator
Start with what’s relevant.
Flexible hierarchy
Reorder and reposition content.
Individual control
Show only what matters.
4. From dashboard to workspace
4. From dashboard to workspace
Information should lead directly to action.
The homepage wasn’t designed as a collection of passive cards.
Each surface helps users move from awareness to understanding and, when appropriate, directly into action.
Events - See what’s happening and catch exceptions early.
Scan → Detect → Act
Example: 4/5 accepted → Grace pending → Resend invitation
Events - See what’s happening and catch exceptions early.
Scan → Detect → Act
Example: 4/5 accepted → Grace pending → Resend invitation
Tasks - Turn attention into prioritized work.
Prioritize → Understand → Complete
Tasks - Turn attention into prioritized work.
Prioritize → Understand → Complete
5. Research shaped the system
5. Research shaped the system
Testing what “my day” means across recruiting roles
I conducted moderated usability research across recruiting roles and organizations to evaluate more than usability. I wanted to understand which information deserved persistent space in a user’s daily workspace, how priorities differed by role, and where our product assumptions needed to change.
10 moderated sessions
5 organizations
12 Participants
9 hands-on prototype sessions
Key findings:
1 · Events created a shared anchor
All 9 in-scope participants kept Events, making it the strongest common element across otherwise very different workflows.
2 · A single default couldn’t serve every role
Participants assembled meaningfully different configurations based on role. The report explicitly concludes that “a single default set will not work” and recommends distinct presets for different recruiting contexts.
3 · Intentional relevance beat passive recency
Pinned content was highly valuable when users had persistent jobs, pools, or pipeline lists they actively cared about, while Recent Active content was much less consistently useful.
4 · Reminders pointed toward proactive assistance
Eight of nine participants kept Reminders, but multiple users described needs that should be generated automatically from recruiting state. Six participants specifically wanted system-generated reminders, and one independently described an agent-generated morning report.
Research → Product decisions
9/9 kept Events
→ Make today’s activity the anchor
Different roles built different homepages
→ Role-aware defaults + personalization
Pinned > inferred recency for focused work
→ Strengthen explicit user intent
6 users wanted system-generated reminders
→ Move toward proactive intelligence
6. From personalized to proactive - The next evolution: an AI recruiting assistant
6. From personalized to proactive - The next evolution: an AI recruiting assistant
The personalized homepage helps users organize their day around what matters. In the next phase, planned for Q4 2026, we are extending that foundation with a contextual AI assistant that can proactively surface priorities, answer questions, explain recruiting context, and help users complete manual tasks.
From organizing the work to helping users get the work done.
Personalized workspace
Organize what matters
Proactive intelligence
Surface what needs attention
Recruiting Assistant
Understand, guide, and act
7. Proactive guidance
7. Proactive guidance
Helping users decide what to do next
Recruiting Assistant doesn’t rely only on users knowing what to ask. It can use signals across recruiting workflows to identify important work and surface relevant next steps directly on the homepage.
Recommend
Suggest relevant next steps users can take immediately.
Prioritize
Surface tasks based on urgency, impact, and recruiting context.
Summarize
Highlight what needs attention today.
8. Contextual assistance
8. Contextual assistance
Helping users move from questions to workflows
Users can also initiate their own conversations with Recruiting Assistant. Instead of treating every question as an isolated prompt, the assistant can use recruiting context to understand intent, retrieve relevant information, and guide users toward the next step.
1. Start with intent
Make complex workflows approachable.
Suggested prompts help users discover what the assistant can do, while open-ended input lets them start naturally with what they are trying to accomplish.
1. Start with intent
Make complex workflows approachable.
Suggested prompts help users discover what the assistant can do, while open-ended input lets them start naturally with what they are trying to accomplish.
Guide with context
Turn conversation into action
The assistant can use available recruiting context to suggest relevant objects and actions, helping users move from a broad request into a specific workflow without manually navigating the system.
9. The interaction model
9. The interaction model
From guidance to action
Proactive
The system identifies what may need attention.
Conversational
The user expresses what they want to accomplish.
Contextual
The assistant connects intent with relevant recruiting context.
Actionable
The conversation leads into a real workflow.
The goal is not to add a chatbot to Recruiting Hub. It is to reduce the effort between recognizing what needs to be done and getting it done.
10. Outcome and what’s next
10. Outcome and what’s next
Validated and moving toward launch
The Q3 experience received strong positive feedback in user testing, validating the direction of the personalized homepage and its core workflows. The design was also reviewed with senior leadership and received strong support for the product direction.
The experience is now moving toward a planned launch in late September 2026. Building on this foundation, the Q4 experience is currently in design, extending Recruiting Hub with proactive recommendations and a contextual AI assistant.
Strong user validation
The Q3 experience received strong positive feedback across moderated user testing, validating the personalized homepage, role-aware configuration, and core daily workflows.
Senior leadership support
Reviewed and supported by senior leadership
Late September 2026
Personalized homepage planned for launch
Q4 2026
Recruiting Assistant in design
Strong user validation
The Q3 experience received strong positive feedback across moderated user testing, validating the personalized homepage, role-aware configuration, and core daily workflows.
Senior leadership support
Reviewed and supported by senior leadership
Late September 2026
Personalized homepage planned for launch
Q4 2026
Recruiting Assistant in design
Stakeholder feedback
“Yuhan is an exceptional UX designer and a fantastic partner to have on a product team. I’ve had the opportunity to work closely with her for our product at Amazon, and I’ve consistently been impressed by her diligence, customer obsession, and thoughtful approach to design.
Yuhan truly works backwards from the customer. Even while working on highly complex problems, she keeps the user at the center and grounds her design decisions in research and data. On one particularly challenging project, she developed an innovative user-testing approach that helped us gather meaningful customer feedback and turn those insights into better product decisions.
She also brings a strong curiosity and willingness to explore new technologies. We worked closely together on bringing native AI capabilities into the product, where she helped think through how we could make these capabilities genuinely useful and intuitive for our customers.
Yuhan is the kind of designer who combines strong design thinking, customer empathy, and excellent collaboration. I would highly recommend her to any team looking for a UX partner who is deeply committed to solving the right problems for customers.”
Vishal Soni
Product Leader at Amazon
“Yuhan is an exceptional UX designer and a fantastic partner to have on a product team. I’ve had the opportunity to work closely with her for our product at Amazon, and I’ve consistently been impressed by her diligence, customer obsession, and thoughtful approach to design.
Yuhan truly works backwards from the customer. Even while working on highly complex problems, she keeps the user at the center and grounds her design decisions in research and data. On one particularly challenging project, she developed an innovative user-testing approach that helped us gather meaningful customer feedback and turn those insights into better product decisions.
She also brings a strong curiosity and willingness to explore new technologies. We worked closely together on bringing native AI capabilities into the product, where she helped think through how we could make these capabilities genuinely useful and intuitive for our customers.
Yuhan is the kind of designer who combines strong design thinking, customer empathy, and excellent collaboration. I would highly recommend her to any team looking for a UX partner who is deeply committed to solving the right problems for customers.”
Vishal Soni
Product Leader at Amazon
Research participants are identified by role rather than name. Product screens are recreated with sample data.
Research participants are identified by role rather than name. Product screens are recreated with sample data.
