InsureBot — Redesigning a Humanized, Intelligent Insurance Assistant
Transforming an AI-powered virtual assistant into a human-centered, intuitive, and efficient tool for healthcare professionals
Overview
InsureBot is an AI-powered virtual assistant built on IBM Watson to help healthcare professionals retrieve insurance details. However, despite using robust technology, its potential remained largely untapped.
The experience was slowed down by:
Complex and rigid workflows
Robotic interactions with no empathy
Cluttered data presentation
Underutilized NLP and AI capabilities
Our goal: Transform InsureBot into a human-centered, intuitive, and efficient assistant by refining workflows, enhancing interaction design, improving personality, and elevating the overall interface.
Project Details
Collaboration Overview
Role
UX Designer
Duration
6 months
Team
1 Senior Designer, 2 UX Designers
Collaborators
1 Product Owner, IBM Development Team
Identifying Challenges
The Problem
1
Complex & Rigid Flows
The chatbot relied on long, linear conversations, making it difficult for users to:
  • Navigate or switch topics
  • Correct mistakes
  • Locate relevant information
→ Resulting in frustration and slow task completion.
2
Lack of Empathy & Engagement
The bot's tone felt robotic and impersonal. No warmth, personality, or human-like reactions — reducing trust and comfort during support conversations.
3
Underutilized AI Capabilities
Despite IBM Watson's powerful NLP engine, InsureBot failed to leverage:
  • Intent detection fully
  • Entity extraction
  • Autocorrect and error tolerance
  • Context-aware responses
This led to generic answers and missed opportunities for intelligent assistance.
Objective
Design an end-to-end experience that allows healthcare professionals to quickly and reliably access insurance details.
Mission
Transform InsureBot into a human-centered, high-utility product.
Research Phase
Understanding the Challenge
To address unclear requirements, we began by asking:
How can we understand the product and users better?
How might we make InsureBot more engaging and valuable?
How can users access critical information easily?
"With numerous questions and the need to revamp InsureBot, a strategic plan was devised."
Methodology
Process Overview
A structured approach ensured clarity and impact:
DISCOVERY
Understand current flows, identify gaps, analyze competitors.
DEFINE
Heuristic evaluation, CSAT review, and UX scope creation.
IDEATE
Issue prioritization, brainstorming, and early wireframes.
DESIGN
Refinements, high-fidelity prototypes, enhanced flows and visuals.
TECH SUPPORT
Developer collaboration + detailed UI Style Guide handoff.
Phase 1
DISCOVERY
Understanding InsureBot
We focused on simplifying three high-priority intents:
  • Eligibility & Benefits
  • Claims
  • Pre-Authorization
Activities:
  • Conducted stakeholder KT sessions
  • Streamlined conversation flows
  • Reviewed IBM Watson's tech stack and NLP algorithms
NLP Insights (Watson)
We studied different Algorithms
This allowed us to design conversations that align with Watson's interpretation of user inputs.
Competitive Analysis
Benchmarked chatbots across Aviation, Retail, and Fintech.
Insights
  • More human, empathetic tones improve engagement
  • Smart triggers & suggestive prompts help users act faster
  • Branding and persona create a memorable identity
  • Non-verbal cues (emojis, sounds, micro-interactions) boost satisfaction
  • Users expect flexible topic-switching
Key Takeaways for InsureBot
Heavy, complex data flow
Underutilized Watson features
Linear, robotic conversations
Lack of identity and brand personality
"After understanding InsureBot, the next step was to define issues and set clear UX goals."
Phase 2
DEFINE
Heuristic Evaluation
Mapped 20+ usability issues based on Nielsen's principles.
Key Issues Identified:
CSAT Review
User feedback highlighted: Frustration with long, rigid flows, Bot misunderstanding queries, Lack of empathy, Difficulty finding key information
This validated our usability findings and highlighted the need for flexibility and natural language support.
UX Scope
To transform InsureBot into a more intuitive and empathetic assistant, we focused on:
✔ Simplifying Flows
Shorter journeys, fewer steps, reduced cognitive load.
✔ Adding Empathy
Humanized tone to improve trust and comfort.
✔ Unifying Interaction Patterns
Consistent CTAs, layouts, and conversational blocks.
✔ Leveraging Watson Fully
Smarter, context-aware interactions powered by NLP.
Phase 3 & 4
IDEATE
Issue Prioritization & Brainstorming
We prioritized usability problems and explored multiple solution paths:
Examples:
Issue 1
Context switching required scrolling through long conversations.
Solution: Introduced "Menu" command + quick-access interaction patterns.
Issue 2
Robotic, rigid dialogue.
Solution: Custom conversational scripts with personality & empathy.
Impact vs Effort Metrics
Used to plan the roadmap and ensure quick wins without compromising long-term improvements.
DESIGN ITERATION
Progressive refinements—from wireframes → mid-fidelity → high-fidelity.
Focus areas:
  • Structure
  • Clarity
  • Accessibility
  • Humanized conversation patterns
Key Screens
Pre-Chat Form
Before: Too many mandatory fields → slow start
After: Minimal input → Faster, frictionless onboarding
Intent Switching
Before: Users navigated repeated, lengthy conversations to change topics.
After: Minimal input → Faster, frictionless onboarding
Context Banner
Before: No clarity on current topic. Users scrolled extensively to regain context.
After: Persistent context banner with simple switching options.
Auto-Suggestions
Before: Robotic conversation with no guidance.
After: Smart suggestions → Faster decisions and reduced typing.
Additional Enhancements
  • Customized natural dialog
  • Live Agent interaction design
  • Unified component library
  • Error-tolerant input handling
  • Consistent CTA hierarchy
  • Improved rating and feedback flow
Impact on Users
✔ 30–35% reduction in live agent escalations
More users completed tasks directly within the chatbot.
✔ Reduced cognitive load
Clear structure, simplified flows, and better hierarchy.
✔ Improved conversation satisfaction
Humanized tone + smarter auto-suggestions.
✔ Faster task completion
Intent switching and context awareness reduced navigation time.
Made with