Insurebot
Enhancing Healthcare Insurance Information Access to Users
.png)
Overview
Insurebot assists healthcare professionals in retrieving insurance details but lacks key features. Despite IBM Watson usage, its potential is untapped. Complex flows and convoluted data presentation cause frustration.
The objective was to elevate InsureBot by optimizing workflows, refining flows and interactions, enhancing personality, and improving the user interface.
Starting with Questions...
There were many gaps in the requirements from the team about the InsureBot, so we started with a list of Questions.
-
How we can understand the product and its flow better ?
-
How we can understand our new and existing users ?
-
How might we provide an experience that is engaging and valuable to our users ?
-
How might we allow them to access their most critical information easily ?
“With numerous questions and the necessity to revamp the InsureBot, A strategic Plan was devised”
Redesigning a Better Experience
DISCOVERY
Understanding Insure bot
Competitive Analysis
DEFINE
Heuristic Evaluation
Issue Grouping
CSAT Reviews
Defining UX Scope
IDEATE
Issue Prioritisation
Solution & Brainstorming
Impact & Efforts Metrics
Initial Wireframes
DESIGN ITERATION
Wire Framing
Design Language
High Fidelity Designs
Prototyping
TECH SUPPORT
UI Style Guide to Dev
Team
Discovery
Understanding InsureBot
Understanding the InsureBot involved critical steps, including:
-
Understanding High Priority Intent & simplifying their Flow
-
Reviewing IBM Watson integration.
1.Understanding High Priority Intent & simplifying their Flow
We conducted KT sessions with stakeholders to gain a comprehensive understanding of InsureBot's operational flow. Among the various user intents, we identified three key areas of interest:
-
Eligibility & Benefits
-
Claims
-
Pre-Authorization
Upon a closer examination of these intents, we discovered that the chat flows and dialogues were needlessly complex. In order to enhance our understanding of these intents, we undertook an effort to streamline and simplify the flow.
Original Flow ( Claims)
Simplified Flow version ( Claims)
2. Reviewing IBM Watson integration.
In order to grasp the underlying structure of InsureBot, which is built on IBM Watson, we successfully completed the IBM Chatbot training available on their website. Throughout this training, we accomplished the following:
​
-
Gained familiarity with its framework, which encompassed essential chatbot concepts such as Utterances, Intents, Slots, and Regression.
-
Realized the critical role of NLP (Natural Language Processing) in translating human language into a format comprehensible to computers, a fundamental aspect of effective chatbot operation.
-
Uncovered that Watson Assistant relies on four primary algorithms to facilitate this intricate process: Intent Detection, Entity Detection, Irrelevance Detection, and Autocorrect
Discovery
Competitive Analysis
We examined chatbots in Aviation, Retail, and Fintech to assess their utility and performance in these industries. Our analysis yielded valuable insights for enhancing chatbot development and optimization. Here are some noteworthy key findings from our competitive analysis.
-
Humanising Conversation
-
Chatbot Branding
-
Creating external triggers to initiate the chatbot
-
Suggestive chats
-
Handling exceptions
-
Omni-channel experience
-
Redirecting the users for more information/external workflow
-
Non-verbal responses like sounds, emojis and feedback scale
-
The flexibility to change the subject in-between of a conversation.
Key Takeaways
-
InsureBot contains a substantial amount of data.
-
Data is presented in a complex manner, potentially confusing and frustrating users.
-
IBM Watson's capabilities are not fully utilised.
-
Conversations often feel robotic and follow a linear pattern.
-
Lack of branding in the InsureBot's interaction.
“After comprehending Insurebot, the imperative was to define issues and establish clear targets”
Define
Define
Define
Heuristic Evaluation
To uncover usability issues within Insurebot, we conducted a Heuristic evaluation which was aligned with Jakob Nielsen's 10 usability heuristics and revealed significant issues/concerns across different intents.
Discovered concerns
Heuristics
Efficiency and Flexibility (User Control and Freedom)
1.Changing context takes too long: It time-consuming to switch between conversation topics due to the need to scroll to the end of the conversation.
Efficiency and Flexibility (User Control and Freedom)
2.Disabled typing space hampers transitions.
4. Bot struggles to understand users
Recognition and Error Prevention
5. Information overload for users.
Aesthetics and Minimalist Design
6. Complex chat initiation process.
Efficiency and Flexibility (User Control and Freedom)
7. Bot doesn't address sub-intents directly: The bot fails to provide direct responses for sub-intents, leading to user frustration.
Efficiency and Flexibility (User Control and Freedom)
8. Button statuses are unclear
Visibility of System Status
9. Table presentation inconsistent.
Consistency and Standards
10. No clear editable area indicators.
Visibility of System Status
3.Intent changes are confusing: Changing conversation topics confuses users.
Consistency and Standards
11. Inconsistent CTA buttons: The Call to Action (CTA) buttons vary in design and function throughout the conversation, leading to inconsistency.
Consistency and Standards
12. Word hierarchy issues, "LIVE AGENT" isolation.
Aesthetics and Minimalist Design
13. Dialogue lacks personality and empathy: The bot's responses come across as too straightforward and lack a human touch in terms of personality and empathy.
Aesthetics and Minimalist Design
14. CSAT/NPS display needs improvement.
Aesthetics and Minimalist Design
15. Different exit flows for closing chats needed.
Consistency and Standards
16. Lack of spacing between dialogues.
Aesthetics and Minimalist Design
17. Bot struggles with multiple inputs.
Recognition and Error Prevention
18. Limited language comprehension.
Recognition and Error Prevention
19. Action buttons need clearer labels: Users would benefit from more descriptive labels on action buttons for better clarity.
Visibility of System Status
20. Uniform response for ratings, regardless of sentiment: Ratings receive uniform responses, irrespective of sentiment.
Recognition and Error Prevention
Issue Grouping : Heuristic Evaluation
Define
CSAT Review
Evaluating the identified issues with user feedback was crucial, and thus, a Customer Satisfaction (CSAT) review was conducted to gauge our target audience's sentiments regarding our service or product. Here are some insights gathered from existing users:
UX Scope:Key Takeaways
Despite IBM Watson's capabilities, InsureBot's conversation structure is rigid, often requiring users to navigate through lengthy dialogues when changing context, leading to an unnatural conversation flow. After our research, we've identified UX improvement areas for "InsureBot":
“The Next step is to craft innovative solutions and design strategies that closely align with users needs and objective”
Ideate
Issue Prioritisation and Brainstorming & Solutions
Issue prioritisation is essential to focus on critical problems first, while solutions brainstorming fosters creativity and helps generate innovative ideas for effective problem-solving.
Some issues and their solution
Issue 1: Context change is too far
​
Solution: Change of context requires the user to traverse a lengthy conversation
Issue 2: Bot Dialog - Humanizing Natural Conversation
​
Solution: Ensuring the users are able to have a spontaneous conversations with the chatbot.
Ideate
Impact & Efforts Metrics
Impact and Efforts Metric are used to evaluate Design changes and improvement , where “Impact” measures the positive effects on user experience, while “efforts” gauge the resources required for a design change, helping priortize improvement effectively.
“Now, it’s essential to continuously refine and improve the design based on the previous findings, feedback and testing results to enhance the ongoing user experience.”
Design Iteration
Low-Fidelity wireframes
In the initial design phase, we quickly generated design concepts, iterated on them, and collected early feedback to incorporate into the final design layout.
Design Iteration
High-Fidelity wireframes
Following numerous design iterations, we arrived at a high-fidelity final design that aligns seamlessly with the company's established design language.
Screen 1 : Pre chat form
Before
Excessive fields are required to initiate a conversation with the bot.
After
Minimal Input: Reducing the required fields to make the entry process more efficient and less time-consuming.
Screen 2 : Intent Change
Before
Change of Intent requires the users to traverse a lengthy conversation
After
Simple Intent Switching: Users can easily change context by typing "Menu."
Screen 3 : Context Banner
Before
No reference on current chat and requires traverse a lengthy conversation to change the context.
After
Context banner at the top provides conversation reference and enables users to switch context easily.
Screen 4 : Auto- Suggestion
Before
Robotic Conversation where no auto suggestions is provided to the users
After
Auto Suggestion for helping users to take decision Quickly and easily.
Some other features added to “Insurebot”
Customised natural dialog
Live Agent Interaction Design
Learnings
Building a chatbot involves various components and technologies. Some of the essential elements and considerations include:
-
NLP ( Natural language processing ) for language understanding.
-
Dialogue management for conversation flow. It determines how the chatbot should respond based on the current state of the conversation, user history, and context.
-
Speech recognition (voice chatbots).
-
AI for learning and improving.
-
API integration for external actions. Chatbots can be integrated with external systems and services through APIs, enabling them to perform actions like making reservations, checking the weather, or accessing user-specific data.
-
User Authentication and Security: Implementing user authentication and ensuring data security are important for protecting user information and maintaining trust.