AI chatbot for Internal knowledge
Healthcare
Employees across Optum frequently struggle to find accurate, up-to-date answers to commonly asked internal questions related to tools, programs, and operational workflows. Existing knowledge repositories like Confluence or static documents are fragmented, hard to navigate, and often require SME intervention for even basic queries.
Client
United Healthcare
Services
Visual Design
UI & UX Design
Conversational Design
Industries
Healthcare
Insurance
Date
Jan 2025
The problem
The problem
Employees across Optum frequently struggle to find accurate, up-to-date answers to commonly asked internal questions related to tools, programs, and operational workflows. Existing knowledge repositories like Confluence or static documents are fragmented, hard to navigate, and often require SME intervention for even basic queries.
Employees across Optum frequently struggle to find accurate, up-to-date answers to commonly asked internal questions related to tools, programs, and operational workflows. Existing knowledge repositories like Confluence or static documents are fragmented, hard to navigate, and often require SME intervention for even basic queries.
Increased dependency on SME
Slower onboarding for new joiners
Lesser operational efficency
Scattered knowledge
My role
My role
I was the Lead Designer on this project, which meant juggling a bunch of roles—from managing stakeholders to working closely with 2+ frontend developers.
I was the Lead Designer on this project, which meant juggling a bunch of roles—from managing stakeholders to working closely with 2+ frontend developers.
Who are the user's
Who are the user's
Optum employees across departments and roles who rely on the chatbot to quickly find accurate answers to internal tools, workflows, and process-related queries.
Optum employees across departments and roles who rely on the chatbot to quickly find accurate answers to internal tools, workflows, and process-related queries.
Setting the Stage
Setting the Stage
To design better experiences, it’s important to first understand the competitive landscape, user expectations, and the challenges they face.
To design better experiences, it’s important to first understand the competitive landscape, user expectations, and the challenges they face.
1
Why are employees currently struggling to find answers?
Why are employees currently struggling to find answers?
2
Why are existing tools(eg, Confulence, Copilot) not working well?
Why are existing tools(eg, Confulence, Copilot) not working well?
3
Why is it important to solve this problem now?
Why is it important to solve this problem now?
4
Who are the most frequent askers of internal questions?
Who are the most frequent askers of internal questions?
5
Who are the current information gatekeepers (SMEs. teamleads, support teams) ?
Who are the current information gatekeepers (SMEs. teamleads, support teams) ?
6
Who benefits most from instant answers (new joiners, Ops teams, etc)?
Who benefits most from instant answers (new joiners, Ops teams, etc)?
7
When are delays in finding answers most costly?
When are delays in finding answers most costly?
8
What types of questions will the chatbot answer(FAQ's , Links, SOPs. Systems)?
What types of questions will the chatbot answer(FAQ's , Links, SOPs. Systems)?
9
What content sources will it pull from ( Confluence, Sharepoint, Internal Wiki)?
What content sources will it pull from ( Confluence, Sharepoint, Internal Wiki)?
9
What level of personalization is needed (role-based, team-based, geo -specific)?
What level of personalization is needed (role-based, team-based, geo -specific)?
9
What success metrics will define its Value (time saved, SME load, Satisfaction)?
What success metrics will define its Value (time saved, SME load, Satisfaction)?
Asking the Experts
Optum employees across departments and roles who rely on the chatbot to quickly find accurate answers to internal tools, workflows, and process-related queries.
Optum employees across departments and roles who rely on the chatbot to quickly find accurate answers to internal tools, workflows, and process-related queries.
Question-1
Why do we need a chatbot instead of central repository of documents?
Question-2
What are the most common types of question employees struggle to get answers?
Question-3
Which team or department would you think benefit the most from this?
Question-4
What is the level of personalization you need like role based, team based etc?
Question-5
How do you wish to roll out this to the team?
Question-6
Are there any key KPI's you want to track?
Needs, Pain points & responsibilites
Needs, Pain points & responsibilites
Thinking through the Structure
Thinking through the Structure
Final design
Final design
UI were created in Figma using United Healthcare built-in design system features to ensure consistency and scalability. We incorporated ready-to-use Lottie animations to enhance micro-interactions and leveraged accessibility plugins to align with WCAG compliance standards. As we began presenting the designs, we iterated continuously to reflect stakeholder feedback and ensure alignment with business priorities.
UI were created in Figma using United Healthcare built-in design system features to ensure consistency and scalability. We incorporated ready-to-use Lottie animations to enhance micro-interactions and leveraged accessibility plugins to align with WCAG compliance standards. As we began presenting the designs, we iterated continuously to reflect stakeholder feedback and ensure alignment with business priorities.
Tone of Voice
Tone of Voice
One of the most overlooked aspects of chatbot design is tone of voice. It’s easy to focus on functionality—getting the chatbot to return the right answer, connect to the right source, or solve a ticket. But what often goes unnoticed is how the answer is delivered. Through this project, I learned that tone isn’t just “how it sounds”—it’s how the experience feels to the user.
One of the most overlooked aspects of chatbot design is tone of voice. It’s easy to focus on functionality—getting the chatbot to return the right answer, connect to the right source, or solve a ticket. But what often goes unnoticed is how the answer is delivered. Through this project, I learned that tone isn’t just “how it sounds”—it’s how the experience feels to the user.
Key Takeaways
Working on the Optum AI Chatbot taught me the nuances of designing for healthcare conversations at scale. Unlike consumer apps, the challenge wasn’t just about usability
Working on the Optum AI Chatbot taught me the nuances of designing for healthcare conversations at scale. Unlike consumer apps, the challenge wasn’t just about usability
Stakeholder alignment drivers direction
Stakeholder alignment drivers direction
with multiple teams involved sucess depends on balancing business needs with user experience rather than pushing one over other.
with multiple teams involved sucess depends on balancing business needs with user experience rather than pushing one over other.
Conversation design is system design
Conversation design is system design
I learned that chatbot design is less about clever prompts and more about structuring information so users always feel guided and safe
I learned that chatbot design is less about clever prompts and more about structuring information so users always feel guided and safe
Trust is non-negotiable in healthcare
Trust is non-negotiable in healthcare
Users needed clear, reliable, and compliant responses. even small errors could break confidence in the system
Users needed clear, reliable, and compliant responses. even small errors could break confidence in the system
Other Projects
Other Projects