Marvel AI - Exam generation feature for educators which led to time saving of 5 hours per examination
Marvel AI - Exam generation feature for educators which led to time saving of 5 hours per examination
Marvel AI - Exam generation feature for educators which led to time saving of 5 hours per examination
User Research
User Research
Research Analysis
Research Analysis
Visual Design
Usability Testing
Interaction Design
Visual Design
Interaction Design
Usability Testing
Design System
Context
Reality AI was developing second iteration of an AI Chat Bot called Marvel AI to help educators in their day to day life. During the UX research, I found out that exam generation was one of the time consuming areas and Reality AI can use this opportunity to strengthen the Marvel AI features. Upon discussing this with the management, I got the opportunity to form and lead the team to design Exam Generation feature.
Context
Reality AI was developing second iteration of an AI Chat Bot called Marvel AI to help educators in their day to day life. During the UX research, I found out that exam generation was one of the time consuming areas and Reality AI can use this opportunity to strengthen the Marvel AI features. Upon discussing this with the management, I got the opportunity to form and lead the team to design Exam Generation feature.
Context
Reality AI was developing second iteration of an AI Chat Bot called Marvel AI to help educators in their day to day life. During the UX research, I found out that exam generation was one of the time consuming areas and Reality AI can use this opportunity to strengthen the Marvel AI features. Upon discussing this with the management, I got the opportunity to form and lead the team to design Exam Generation feature.
My Role
UX Researcher
Information Architect
Interaction Designer
Visual Designer
Duration
3 Months
Team members
2 Product designer
1 Design Manager
Industry
Ed-Tech
Users
Target users of Marvel AI were educators aged 24 to 50, actively involved in examinations, making up 70% of the total user base.
Users
Target users of Marvel AI were educators aged 24 to 50, actively involved in examinations, making up 70% of the total user base.
Users
Target users of Marvel AI were educators aged 24 to 50, actively involved in examinations, making up 70% of the total user base.
Insights
12 structured User Interviews and 50 User Survey responses were analyzed and following insights were identified.
Insights
12 structured User Interviews and 50 User Survey responses were analyzed and following insights were identified.
Insights
12 structured User Interviews and 50 User Survey responses were analyzed and following insights were identified.
Challenge
“The key challenge was to reduce the amount of time required to generate the examination”
Challenge
“The key challenge was to reduce the amount of time required to generate the examination”
Challenge
“The key challenge was to reduce the amount of time required to generate the examination”
Solution
The solution was in the form of a chat response by AI Agent. New Design system was also introduced including a dark and a light theme.
Solution
The solution was in the form of a chat response by AI Agent. New Design system was also introduced including a dark and a light theme.
Solution
The solution was in the form of a chat response by AI Agent. New Design system was also introduced including a dark and a light theme.
User research and its analysis led to a common framework followed by educators. This framework was used as a guiding point to strategize the exams
Editing tab
Generating exam required customization depending on situation hence the UX was designed with overall strategy while adding prompt and detailed strategy after the response. This method ensured minimal upfront friction.
Generating exam required customization depending on situation hence the UX was designed with overall strategy while adding prompt and detailed strategy after the response. This method ensured minimal upfront friction.
Generating exam required customization depending on situation hence the UX was designed with overall strategy while adding prompt and detailed strategy after the response. This method ensured minimal upfront friction.
86% Users expressed that the responses given by AI requires local editing and there is no need to change the entire thread
Selective Editing
Selective Editing
Making changes to the response using correction prompts created frustration amongst users because of time required to mention the location, articulating changes and hallucinations in the response. Hence editing features including Ask AI, Manual Edit, Lock and Delete were introduced.
Making changes to the response using correction prompts created frustration amongst users because of time required to mention the location, articulating changes and hallucinations in the response. Hence editing features including Ask AI, Manual Edit, Lock and Delete were introduced.
Making changes to the response using correction prompts created frustration amongst users because of time required to mention the location, articulating changes and hallucinations in the response. Hence editing features including Ask AI, Manual Edit, Lock and Delete were introduced.
Ask AI
Even though features like Manual Editing, Lock and Delete were useful, users expressed frustrations with corrected response provided by the agent, hence Ask AI was introduced to speed up local editing.
Even though features like Manual Editing, Lock and Delete were useful, users expressed frustrations with corrected response provided by the agent, hence Ask AI was introduced to speed up local editing.
Even though features like Manual Editing, Lock and Delete were useful, users expressed frustrations with corrected response provided by the agent, hence Ask AI was introduced to speed up local editing.
75% Users mentioned they will be able to use the full potential of AI if right assistance is being provided.
Dynamic Prompt Assistance
To make the most out of the prompts, dynamic prompt assistance was introduced which provided relevant suggestions based on the text entered by the users
To make the most out of the prompts, dynamic prompt assistance was introduced which provided relevant suggestions based on the text entered by the users
To make the most out of the prompts, dynamic prompt assistance was introduced which provided relevant suggestions based on the text entered by the users
Impact
Impact
Impact
Reflections
The impact was measured based on the metrics that were critical in driving the business for Simdaa Technologies and its client.
Reflections
The impact was measured based on the metrics that were critical in driving the business for Simdaa Technologies and its client.
Reflections
The impact was measured based on the metrics that were critical in driving the business for Simdaa Technologies and its client.
Explore case studies
Explore case studies
Shubham Gorule
Shubham Gorule
Shubham Gorule