# B2B

# B2B

# B2B

# AI

# AI

# AI

# PoC

# PoC

# PoC

AI-Driven Search Integration:
Optimizing Employee Self-Service

TL; DR

PROBLEM

Employees struggle to find relevant information within Rockworld due to scattered, uncategorized search results and overwhelming content. This leads to longer search times, increased dependency on HR, and frustration among remote new hires. Additionally, AI adoption in enterprises faces high implementation costs and ROI uncertainty, making large-scale investments risky.

SOLUTION

In a one-week UX sprint, we designed an AI-powered search PoC with categorized results, an AI chatbot, and structured escalation paths, reducing search time by 40% and boosting user satisfaction by 80%.

I led information architecture, wireframing, high-fidelity design, and prototyping to validate a scalable, user-centric solution.

TL; DR

High costs of AI implementation and uncertainty about ROI are often significant barriers to company-wide AI adoption. However, Rocket addresses this challenge with a "Pilot and Scale" approach: starting with small, manageable pilot projects to test and validate AI solutions.

Thus, out team's sprint week dedicated to UX exploration—from concept to testing—A PoC can effectively demonstrate the value of AI integration within the current product. During this sprint, I focused on everything from research, wireframing, and iteration to testing, all within one week, to assess whether this solution is truly valuable.

👩‍💻 My Role: Product Designer Intern | Rocket Companies

👩‍💻 My Role: Product Designer Intern | Rocket Companies

Information Architecture

Wireframing

UX Research

High-Fidelity Prototyping

Usability Testing

Information Architecture

Wireframing

UX Research

High-Fidelity Prototyping

Usability Testing

👥 Team

👥 Team

1 Product Owner

1 Product Manager

2 Product Designers (Me)

2 UX Researchers

2 Content Strategists

1 Product Owner

1 Product Manager

2 Product Designers (Me)

2 UX Researchers

2 Content Strategists

🕑 Duration

🕑 Duration

1-Week Sprint

1-Week Sprint

TL; DR

High costs of AI implementation and uncertainty about ROI are often significant barriers to company-wide AI adoption. However, Rocket addresses this challenge with a "Pilot and Scale" approach: starting with small, manageable pilot projects to test and validate AI solutions.

Thus, out team's sprint week dedicated to UX exploration—from concept to testing—A PoC can effectively demonstrate the value of AI integration within the current product. During this sprint, I focused on everything from research, wireframing, and iteration to testing, all within one week, to assess whether this solution is truly valuable.

TL; DR

High costs of AI implementation and uncertainty about ROI are often significant barriers to company-wide AI adoption. However, Rocket addresses this challenge with a "Pilot and Scale" approach: starting with small, manageable pilot projects to test and validate AI solutions.

Thus, out team's sprint week dedicated to UX exploration—from concept to testing—A PoC can effectively demonstrate the value of AI integration within the current product. During this sprint, I focused on everything from research, wireframing, and iteration to testing, all within one week, to assess whether this solution is truly valuable.

TABLE OF CONTENT

TABLE OF CONTENT

  1. Identifying the Problem

  • Usability challenges

  • Business goals

  1. Identifying the Problem

  • Usability challenges

  • Business goals

  1. Research & Insights

  • User interviews

  • Early insights

  • User persona

  1. Research & Insights

  • User interviews

  • Early insights

  • User persona

  1. Design Explorations

  • Design challenge & solution

  • Information architecture

  • Wireframes

  1. Design Explorations

  • Design challenge & solution

  • Information architecture

  • Wireframes

  1. Results & Validation

  • High-fidelity design & prototype

  • User scenario showcase (3 tasks)

  • User testing

  • Developer’s pushback

  • Impact

  • Key Learnings

  1. Results & Validation

  • High-fidelity design & prototype

  • User scenario showcase (3 tasks)

  • User testing

  • Developer’s pushback

  • Impact

  • Key Learnings

IMPACT

IMPACT

40%

40%

Reduction in time spent searching

Reduction in time spent searching

80%

80%

Increase in user satisfaction

Increase in user satisfaction

1

Identifying the Problem

2

Research & Insights

3

Design Explorations

3

Results & Validation

CONTEXT

CONTEXT

Rockworld: An Internal Enterprise Platform for 10,000 Employees

Rockworld: An Internal Enterprise Platform for 10,000 Employees

Rocket Companies, a leading mortgage lender in the U.S., operates a vast ecosystem of tools and platforms. As the company scaled, the need for an efficient, scalable, and user-friendly internal portal—Rockworld—became increasingly clear.

Rocket Companies, a leading mortgage lender in the U.S., operates a vast ecosystem of tools and platforms. As the company scaled, the need for an efficient, scalable, and user-friendly internal portal—Rockworld—became increasingly clear.

PROBLEM STATEMENT

PROBLEM STATEMENT

Identifying Recurring Issues in Rockworld

Identifying Recurring Issues in Rockworld

Since joining the team in May 2024, I have successfully launched 5+ Rockworld pages. Despite varying page performance, we identified consistent pain points affecting user engagement and conversion rates.

Since joining the team in May 2024, I have successfully launched 5+ Rockworld pages. Despite varying page performance, we identified consistent pain points affecting user engagement and conversion rates.

🕑

Lack of WIIFM
(Whats in it for me)

🫠

Lack of Actionable Solution

Low screen efficiency

🕑

Lack of WIIFM(Whats in it for me)

🫠

Lack of Actionable Solution

Low screen efficiency

BUSINESS GOAL

BUSINESS GOAL

Validating AI Integration with Minimal Resources

Validating AI Integration with Minimal Resources

Rocket Companies is exploring ServiceNow AI Virtual Agent to improve service delivery and create a faster, more intuitive user experience. Before committing significant resources, we aimed to validate AI’s impact by addressing key user pain points and testing its effectiveness in improving user satisfaction.

Rocket Companies is exploring ServiceNow AI Virtual Agent to improve service delivery and create a faster, more intuitive user experience. Before committing significant resources, we aimed to validate AI’s impact by addressing key user pain points and testing its effectiveness in improving user satisfaction.

Identifying the Problem

2

Research & Insights

3

Design Explorations

4

Results & Validation

USER INTERVIEW

USER INTERVIEW

Search is the Primary Discovery Method, But Users Are Frustrated

Search is the Primary Discovery Method, But Users Are Frustrated

We conducted internal testing of Rockworld with 10 participants at Rocket Companies. Findings show that 70% believe the search functionality needs improvement to maximize Rockworld's potential.

How We Validated This:

  • Conducted stakeholder interviews and user testing to gather real-world use cases.

  • Leveraged team expertise to validate assumptions and consolidate key issues.

  • Discovered additional pain points through research, shaping our AI-powered redesign approach.

We conducted internal testing of Rockworld with 10 participants at Rocket Companies. Findings show that 70% believe the search functionality needs improvement to maximize Rockworld's potential.

How We Validated This:

  • Conducted stakeholder interviews and user testing to gather real-world use cases.

  • Leveraged team expertise to validate assumptions and consolidate key issues.

  • Discovered additional pain points through research, shaping our AI-powered redesign approach.

EARLY INSIGHT FROM INTERVIEW

EARLY INSIGHT FROM INTERVIEW

Rockworld: The First Point of Reference for Employees, Especially For Remote New Hires.

Rockworld: The First Point of Reference for Employees, Especially For Remote New Hires.

🛠️

🛠️

Low Engagement with Integrated Tools

Low Engagement with Integrated Tools

Out of 12,300+ users, only 111 interacted with a specific tool last month.

Out of 12,300+ users, only 111 interacted with a specific tool last month.

👩‍💻

👩‍💻

High Dependence Among Remote Employees

High Dependence Among Remote Employees

With 30% fully remote and 70% hybrid employees, Rockworld is a critical resource.

With 30% fully remote and 70% hybrid employees, Rockworld is a critical resource.

📄

📄

Overwhelming & Ineffective Content

Overwhelming & Ineffective Content

Long, complex articles lead to high bounce rates, increasing HR’s workload with repetitive tickets and emails.

Long, complex articles lead to high bounce rates, increasing HR’s workload with repetitive tickets and emails.

USER PERSONA

USER PERSONA

Focusing on Remote New Hires

Focusing on Remote New Hires

WHY? Remote new hires face the greatest onboarding challenges, making them the most demanding user group to design for. Ensuring clarity and accessibility for them benefits all employees.

WHY? Remote new hires face the greatest onboarding challenges, making them the most demanding user group to design for. Ensuring clarity and accessibility for them benefits all employees.

"“If remote new hires can understand it, we all can.”

"“If remote new hires can understand it, we all can.”

——Shannon, Interview Participant

——Shannon, Interview Participant

😫

😫

System Overload
System Overload

Navigating multiple disconnected systems is time-consuming.

Navigating multiple disconnected systems is time-consuming.

🕑

🕑

Lack of Timely Support
Lack of Timely Support

Finding accurate, relevant information quickly is difficult.

Finding accurate, relevant information quickly is difficult.

💬

💬

Confusing Internal Jargon
Confusing Internal Jargon

Terminology complicates navigation and understanding.

Terminology complicates navigation and understanding.

Identifying the Problem

Research & Insights

3

Design Explorations

4

Results & Validation

DESIGN CHALLENGE

DESIGN CHALLENGE

Create a simplified, AI-enhanced search experience that reduces friction, improves self-service, and minimizes HR dependency.

Create a simplified, AI-enhanced search experience that reduces friction, improves self-service, and minimizes HR dependency.

PROBLEM

PROBLEM

Research revealed that remote new hires struggle to navigate Rockworld due to fragmented tools and overwhelming content, leading to inefficiency—especially for foundational tasks like managing PTO.

Research revealed that remote new hires struggle to navigate Rockworld due to fragmented tools and overwhelming content, leading to inefficiency—especially for foundational tasks like managing PTO.

SOLUTION

SOLUTION

We streamlined the search experience by implementing a categorized, AI-powered system, ensuring the solution was both intuitive for users and scalable for future use cases.

We streamlined the search experience by implementing a categorized, AI-powered system, ensuring the solution was both intuitive for users and scalable for future use cases.

INFORMATION ARCHITECTURE

INFORMATION ARCHITECTURE

Balancing Simplicity & Scalability in AI Search Results

Balancing Simplicity & Scalability in AI Search Results

To achieve this, I proposed a multi-layered approach.

  1. AI Search Function: Provide relevant, categorized results to minimize cognitive load.

  2. AI Chatbot: Offer contextual guidance for deeper questions or clarifications.

  3. Submit Ticket: Enable escalation for unresolved queries through a ticketing system.

  4. Call HR Support (Emergency): Ensure direct contact for urgent or highly complex issues.

To achieve this, I proposed a multi-layered approach.

  1. AI Search Function: Provide relevant, categorized results to minimize cognitive load.

  2. AI Chatbot: Offer contextual guidance for deeper questions or clarifications.

  3. Submit Ticket: Enable escalation for unresolved queries through a ticketing system.

  4. Call HR Support (Emergency): Ensure direct contact for urgent or highly complex issues.

I developed an animated prototype to align the team on the information architecture, ensuring clarity and execution feasibility within Sprint Week.

I developed an animated prototype to align the team on the information architecture, ensuring clarity and execution feasibility within Sprint Week.

WIREFRAME

WIREFRAME

Validating Key Features Through Team Feedback

Validating Key Features Through Team Feedback

To quickly explore and prioritize design decisions, I created the wireframe and facilitated a dot voting session with the team. This helped us efficiently identify the most valuable elements while keeping iteration costs low.

To quickly explore and prioritize design decisions, I created the wireframe and facilitated a dot voting session with the team. This helped us efficiently identify the most valuable elements while keeping iteration costs low.

Align w/ User Mental Models

  • Structured search results to align with how users naturally seek information.

  • Users start with quick answers, move to actionable tasks, and delve deeper only if necessary.

Align w/ Hicks Law

  • Reduced cognitive load by grouping search results into logical categories, preventing decision fatigue.

  • Helped users process information faster by limiting overwhelming choices.

Align w/ Information Scent

  • Applied clear, intuitive labeling to help users predict where they would find relevant content.

  • Ensured distinct content groups to improve discoverability and navigation.

Identifying the Problem

Research & Insights

Design Explorations

4

Results & Validation

HIGH FIDEILITY & PROTOTYPE

HIGH FIDEILITY & PROTOTYPE

Use Case: Meet Alex, a Remote New Hire at Rocket Companies

Use Case: Meet Alex, a Remote New Hire at Rocket Companies

Alex is a fully remote new hire who needs to request Paid Time Off (PTO). Before the redesign, this process was confusing and time-consuming. With the new AI-powered search & virtual agent, Alex can now complete their request in minutes.

Alex is a fully remote new hire who needs to request Paid Time Off (PTO). Before the redesign, this process was confusing and time-consuming. With the new AI-powered search & virtual agent, Alex can now complete their request in minutes.

Task 01: Searching for PTO Information

Task 01: Searching for PTO Information

Before:

Alex searches “PTO request” and encounters a long list of uncategorized results filled with scattered, repetitive information, making it difficult to find the right answer.

Before:

Alex searches “PTO request” and encounters a long list of uncategorized results filled with scattered, repetitive information, making it difficult to find the right answer.

After:

Alex toggled AI Search, use natural language" i need 10 days off next month, what options available" .

After:

Alex toggled AI Search, use natural language" i need 10 days off next month, what options available" .

Task 02: Finding Clear Search Results

Task 02: Finding Clear Search Results

Before:

Alex clicks through multiple long articles just to figure out how to submit a request.

Before:

Alex clicks through multiple long articles just to figure out how to submit a request.

After:

Search results are now categorized, making them clear and actionable:

After:

Search results are now categorized, making them clear and actionable:

Task 03: Getting Immediate Help with AI Assistance

Task 03: Getting Immediate Help with AI Assistance

Before:

If Alex is still confused, their only option is to email HR and wait for a response.

Before:

If Alex is still confused, their only option is to email HR and wait for a response.

After:

AI Chatbot offers real-time assistance, answering common questions or escalating to HR support if needed.

After:

AI Chatbot offers real-time assistance, answering common questions or escalating to HR support if needed.

USER TESTING & MY ITERATION & CONCLUSION

USER TESTING & MY ITERATION & CONCLUSION

AI PoC Validates 100% User Satisfaction and Highlights Scalable Potential for Investment.

AI PoC Validates 100% User Satisfaction and Highlights Scalable Potential for Investment.

  • Quick Answer Performance:

    • 100% of participants successfully understood the PTO policy and approximate accrual within 10 seconds.

  • Categorized Search Results:

    • 90% found the categories intuitive and helpful for narrowing their search.

    • Participants appreciated the separation of "Quick Answer" and "To Do," reducing cognitive load.

  • Escalation Paths:

    • 70% preferred Submit Ticket for unresolved issues, citing convenience.

    • 30% opted for Call HR but noted they would only use it for urgent needs.

  • Overall Feedback:

    • 100% rated the AI search experience as faster and more intuitive compared to their usual workflow.

  • Quick Answer Performance:

    • 100% of participants successfully understood the PTO policy and approximate accrual within 10 seconds.

  • Categorized Search Results:

    • 90% found the categories intuitive and helpful for narrowing their search.

    • Participants appreciated the separation of "Quick Answer" and "To Do," reducing cognitive load.

  • Escalation Paths:

    • 70% preferred Submit Ticket for unresolved issues, citing convenience.

    • 30% opted for Call HR but noted they would only use it for urgent needs.

  • Overall Feedback:

    • 100% rated the AI search experience as faster and more intuitive compared to their usual workflow.

👨‍💻 Developer's push back

👨‍💻 Developer's push back

Data Privacy Concern

Data Privacy Concern

"Integrating federated search across platforms like Workday and Slack sounds great, but the metadata bridge might be difficult to implement given our current resources and time constraints."

— Developer, The Pulse

"Integrating federated search across platforms like Workday and Slack sounds great, but the metadata bridge might be difficult to implement given our current resources and time constraints."

— Developer, The Pulse

"Integrating federated search across platforms like Workday and Slack sounds great, but the metadata bridge might be difficult to implement given our current resources and time constraints."

— Developer, The Pulse

"Integrating federated search across platforms like Workday and Slack sounds great, but the metadata bridge might be difficult to implement given our current resources and time constraints."

— Developer, The Pulse

"Integrating federated search across platforms like Workday and Slack sounds great, but the metadata bridge might be difficult to implement given our current resources and time constraints."

— Developer, The Pulse

"Integrating federated search across platforms like Workday and Slack sounds great, but the metadata bridge might be difficult to implement given our current resources and time constraints."

— Developer, The Pulse

My iteration based on feasibility after sprint:

My iteration based on feasibility after sprint:

To address feasibility concerns, I revised the Quick Answer feature to use static data summaries instead of real-time data fetching. This decision reduced technical complexity, mitigated data privacy risks. By transparently communicating that responses are policy-based and providing clear next steps for accessing personalized information, the design balances user clarity and technical constraints.

To address feasibility concerns, I revised the Quick Answer feature to use static data summaries instead of real-time data fetching. This decision reduced technical complexity, mitigated data privacy risks. By transparently communicating that responses are policy-based and providing clear next steps for accessing personalized information, the design balances user clarity and technical constraints.

IMPACT

IMPACT

This project improved user satisfaction, reduced HR workload, and validated the feasibility of AI-powered solutions for internal platforms.

This project improved user satisfaction, reduced HR workload, and validated the feasibility of AI-powered solutions for internal platforms.

40%

40%

Reduction in time spent searching

Reduction in time spent searching

80%

80%

Increase in user satisfaction

Increase in user satisfaction

WHAT I LEARNT

WHAT I LEARNT

AI Best Practice and Work under Pressure.

AI Best Practice and Work under Pressure.

  1. Principles of AI Integration:

Working on this project gave me the opportunity to dive deep into AI design resources, with Google’s People + AI Guidebook and IBM’s AI Design System being particularly influential. These frameworks inspired me to approach AI projects with a user-centered mindset while adhering to structured and ethical design practices. They provided invaluable guidance on how to create AI-powered solutions that are both intuitive and scalable.

  1. Delivering Under Tight Timelines:

The 5-day sprint week was a transformative experience that challenged me to prioritize speed and collaboration over perfection. It pushed me to make quick, confident decisions, communicate effectively with the team, and adapt to constraints without losing focus on user needs. As a junior designer, I learned to navigate the balance between compromise and advocacy, proving that meaningful outcomes can be achieved even in fast-paced and high-pressure environments.

  1. Principles of AI Integration:

Working on this project gave me the opportunity to dive deep into AI design resources, with Google’s People + AI Guidebook and IBM’s AI Design System being particularly influential. These frameworks inspired me to approach AI projects with a user-centered mindset while adhering to structured and ethical design practices. They provided invaluable guidance on how to create AI-powered solutions that are both intuitive and scalable.

  1. Delivering Under Tight Timelines:

The 5-day sprint week was a transformative experience that challenged me to prioritize speed and collaboration over perfection. It pushed me to make quick, confident decisions, communicate effectively with the team, and adapt to constraints without losing focus on user needs. As a junior designer, I learned to navigate the balance between compromise and advocacy, proving that meaningful outcomes can be achieved even in fast-paced and high-pressure environments.