


NOVA-AI
Product Design | UX Research
Timeline - 4 Days
2025
My Role - Product Designer
I designed NOVA-AI, a platform where businesses can build AI-driven support agents that act, learn, and improve over time. Created as part of a startup design challenge, it was a deep dive into blending technical complexity with human-centered design, even though the startup didn’t follow up. Here’s how I approached building smarter, more proactive AI for the future of customer care.
I designed NOVA-AI, a platform where businesses can build AI-driven support agents that act, learn, and improve over time. Created as part of a startup design challenge, it was a deep dive into blending technical complexity with human-centered design, even though the startup didn’t follow up. Here’s how I approached building smarter, more proactive AI for the future of customer care.
Problem Statement
AI Isn’t Enough: Why Customers Still Feel Abandoned After a Service Request
Despite the rise of AI in customer support, users still face slow replies, frustrating loops, and unresolved issues.
As a grad student relying on Uber Eats, I often dealt with late deliveries and refund problems, but the support system made it even harder. Scripted bots offered no real help, and escalation felt impossible.
That experience made one thing clear: today’s AI doesn’t solve problems, it dodges them.
NOVA-AI was my response, a system designed to act, escalate smartly, and help from the first interaction.
Despite the rise of AI in customer support, users still face slow replies, frustrating loops, and unresolved issues.
As a grad student relying on Uber Eats, I often dealt with late deliveries and refund problems, but the support system made it even harder. Scripted bots offered no real help, and escalation felt impossible.
That experience made one thing clear: today’s AI doesn’t solve problems, it dodges them.
NOVA-AI was my response, a system designed to act, escalate smartly, and help from the first interaction.
Possible Solution
I designed NOVA-AI to solve a problem we’ve all felt: customer support that feels slow, robotic, and unhelpful. Instead of giving users scripted replies or bouncing them between bots, NOVA-AI agents take action. They fix issues, escalate when needed, and get smarter with every interaction. It’s a no-code platform that helps businesses offer support that feels fast, human, and useful, the kind of help people expect but rarely get.
Possible Solution
I designed NOVA-AI to solve a problem we’ve all felt: customer support that feels slow, robotic, and unhelpful. Instead of giving users scripted replies or bouncing them between bots, NOVA-AI agents take action. They fix issues, escalate when needed, and get smarter with every interaction. It’s a no-code platform that helps businesses offer support that feels fast, human, and useful, the kind of help people expect but rarely get.
1 - Research
But first, I learnt how an “Agentic AI Works”
An Agentic AI is a smart assistant that can understand problems, take action, and learn from past experiences. Instead of just following commands, it actively finds solutions and resolves issues. When a user makes a request, the AI understands it, checks relevant information, and determines the best solution. If possible, it takes action automatically, such as processing refunds or updating orders. If the issue is too complex, it gathers all necessary details and smoothly hands it over to a human agent. Over time, it learns and improves, making customer support faster, easier, and available 24/7, ensuring a smooth and frustration-free experience for users.
An Agentic AI is a smart assistant that can understand problems, take action, and learn from past experiences. Instead of just following commands, it actively finds solutions and resolves issues. When a user makes a request, the AI understands it, checks relevant information, and determines the best solution. If possible, it takes action automatically, such as processing refunds or updating orders. If the issue is too complex, it gathers all necessary details and smoothly hands it over to a human agent. Over time, it learns and improves, making customer support faster, easier, and available 24/7, ensuring a smooth and frustration-free experience for users.
Technical User Research
To wrap my head around how Agentic AI works, I went down a research rabbit hole, analyzing existing products and relentlessly bugging my AI/ML friends until they either spilled their knowledge or blocked me.
Technical User Research
To wrap my head around how Agentic AI works, I went down a research rabbit hole, analyzing existing products and relentlessly bugging my AI/ML friends until they either spilled their knowledge or blocked me.



Ideation
Then, I slipped into "scribbling mode" to figure out how to turn this into something human-centric. I unleashed a flurry of flows and sketches - basically, my brain on paper just to wrap my head around all the requirements and ensure everything made sense.
Ideation
Then, I slipped into "scribbling mode" to figure out how to turn this into something human-centric. I unleashed a flurry of flows and sketches - basically, my brain on paper just to wrap my head around all the requirements and ensure everything made sense.
Agentic AI Flow - END TO END
Agentic AI Flow - END TO END



Our Agentic AI initiates a conversation with the user based on their needs. If the query is valid, it answers the questions and resolves the issue using the knowledge base and its training.
If the AI is unable to resolve the query or returns a false response, it escalates the issue to a human agent or a higher-level agent in the escalation matrix. The AI then gathers the necessary information, responds to the user from start to finish, and trains itself to handle similar questions in the future.
Our Agentic AI initiates a conversation with the user based on their needs. If the query is valid, it answers the questions and resolves the issue using the knowledge base and its training.
If the AI is unable to resolve the query or returns a false response, it escalates the issue to a human agent or a higher-level agent in the escalation matrix. The AI then gathers the necessary information, responds to the user from start to finish, and trains itself to handle similar questions in the future.
Branding
As soon as I cracked the code on how Agentic AI works, I dove straight into Figma like a creative wizard, ready to cast some design spells. I pulled inspiration from the CozmoX website’s color palette, because why reinvent the wheel when you can just make it look magic with the right shades?
Branding
As soon as I cracked the code on how Agentic AI works, I dove straight into Figma like a creative wizard, ready to cast some design spells. I pulled inspiration from the CozmoX website’s color palette, because why reinvent the wheel when you can just make it look magic with the right shades?


The “Sketches”
I scribbled and scribbled, trying to bridge the gap between technical jargon and visual design.
Basically, turning Information Architecture into something the eyes would love.
The “Sketches”
I scribbled and scribbled, trying to bridge the gap between technical jargon and visual design.
Basically, turning Information Architecture into something the eyes would love.
Finally, “The Design”
FIGMA was used to its Full Potential.
Finally, “The Design”
FIGMA was used to its Full Potential.
Design Components
Design Components


High-Fidelity Screens
High-Fidelity Screens


Users will have access to the following components to help enhance their business:
Users will have access to the following components to help enhance their business:
Your Agents Section
Users can view the AI agents they have created for their business, categorized as ‘LIVE’ or ‘DRAFT.’ This section also includes action buttons to modify agents or resolve tickets that require human intervention.
Your Agents Section
Users can view the AI agents they have created for their business, categorized as ‘LIVE’ or ‘DRAFT.’ This section also includes action buttons to modify agents or resolve tickets that require human intervention.
Getting Started Section
This section provides step-by-step prompts to guide users in creating an AI agent from scratch.
Getting Started Section
This section provides step-by-step prompts to guide users in creating an AI agent from scratch.
Cost Savings Card
This card, displayed prominently, highlights the total cost savings generated by the AI agents. Users can also customize this view based on their preferences.
Cost Savings Card
This card, displayed prominently, highlights the total cost savings generated by the AI agents. Users can also customize this view based on their preferences.
Average CSAT Scores
This card presents the agent’s performance based on Net Promoter Score (NPS) ratings collected at the end of each call or chat.
Average CSAT Scores
This card presents the agent’s performance based on Net Promoter Score (NPS) ratings collected at the end of each call or chat.
Top User Intents
This card displays the most frequently asked questions by users, helping businesses identify key pain points and improve the overall user experience.
Top User Intents
This card displays the most frequently asked questions by users, helping businesses identify key pain points and improve the overall user experience.
Steps Explorer
This feature allows users to track how their AI agent learns from different scenarios and interacts with users through conversations.
Steps Explorer
This feature allows users to track how their AI agent learns from different scenarios and interacts with users through conversations.
Customize button
This helps users customize the dashboard according to their needs, with the theme, color, and addition of their business logos.
Customize button
This helps users customize the dashboard according to their needs, with the theme, color, and addition of their business logos.
Users can create and deploy their Agentic AI model in just five simple steps.
Users can create and deploy their Agentic AI model in just five simple steps.
Create a New Agent
Users can build an AI agent from scratch in just three easy steps.
Create a New Agent
Users can build an AI agent from scratch in just three easy steps.



Hire a New AI Agent
Users can purchase pre-trained AI agents from the marketplace. They simply need to upload a file to the knowledge base to enhance the agent’s expertise for their specific business needs.
Hire a New AI Agent
Users can purchase pre-trained AI agents from the marketplace. They simply need to upload a file to the knowledge base to enhance the agent’s expertise for their specific business needs.
Repurpose an AI Agent
Users can duplicate and modify an existing AI agent to create a new version, saving time and effort.
Repurpose an AI Agent
Users can duplicate and modify an existing AI agent to create a new version, saving time and effort.
Three simple steps to deploy their AI Agent to help users' businesses grow.
Three simple steps to deploy their AI Agent to help users' businesses grow.


Primary Settings
Primary Settings



Choose an Avatar
Select a visual representation for your AI agent. The options include both male and female avatars.
Choose an Avatar
Select a visual representation for your AI agent. The options include both male and female avatars.
Name Your Agent
Give your AI agent a unique name that aligns with its role.
Name Your Agent
Give your AI agent a unique name that aligns with its role.
Describe Your AI Agent
Provide a brief description outlining the agent’s purpose, responsibilities, and role. This step is crucial for defining how the agent will interact and assist users.
Describe Your AI Agent
Provide a brief description outlining the agent’s purpose, responsibilities, and role. This step is crucial for defining how the agent will interact and assist users.

I have provided a "Let AI Create" button that allows users to provide brief pointers or an overview of what the agent should do. The AI will then work its magic to transform your input into a well-crafted role description.
I have provided a "Let AI Create" button that allows users to provide brief pointers or an overview of what the agent should do. The AI will then work its magic to transform your input into a well-crafted role description.
Customize Style & Add Special Instructions
Customize Style & Add Special Instructions



Describe your conversational style
Set your preferred conversational style to customize how the AI interacts with users. Choose the tone, formality, and response style that suits your needs.
Describe your conversational style
Set your preferred conversational style to customize how the AI interacts with users. Choose the tone, formality, and response style that suits your needs.
Special Instructions
Special instructions allow users to guide the AI’s actions within specific boundaries. These instructions help the AI understand your unique needs while staying within the defined guidelines, ensuring personalized and accurate results.
Special Instructions
Special instructions allow users to guide the AI’s actions within specific boundaries. These instructions help the AI understand your unique needs while staying within the defined guidelines, ensuring personalized and accurate results.
AI Agents Expertise
Users can select relevant information from the knowledge base you've uploaded, including FAQs. The AI will then use this content using RAG to deliver accurate and context-aware responses based on your selections.
AI Agents Expertise
Users can select relevant information from the knowledge base you've uploaded, including FAQs. The AI will then use this content using RAG to deliver accurate and context-aware responses based on your selections.
Data Privacy & Security
Data Privacy & Security



Data Privacy
Users/businesses can specify how much an AI Agent can utilize contact profile information to tailor the experience based on rules and regulations set by the respective government.
Data Privacy
Users/businesses can specify how much an AI Agent can utilize contact profile information to tailor the experience based on rules and regulations set by the respective government.
Security
Users/ businesses can set up risk protection measures for the AI Agent to follow by a toggle button. These instructions will guide content generation, chats, Q&A, rewriting, summarizing, and more. The chosen settings will be included in the prompt to ensure safety, which may increase token usage.
Security
Users/ businesses can set up risk protection measures for the AI Agent to follow by a toggle button. These instructions will guide content generation, chats, Q&A, rewriting, summarizing, and more. The chosen settings will be included in the prompt to ensure safety, which may increase token usage.
Test the Agent
Test the Agent



Users/ business can test their agent in a no-code canvas.
Users/ business can test their agent in a no-code canvas.
Users can choose to make a call, send a text, or access a ready-to-use code for integrating the agent into an external platform. A transcript will also be shown to the user/ business.
Users can choose to make a call, send a text, or access a ready-to-use code for integrating the agent into an external platform. A transcript will also be shown to the user/ business.
Canvas Components
Canvas Components



1
Users can choose whether the agent can respond to customers via call or text. Additionally, users can record their voice by selecting an option from the functions or connectors module, allowing our AI to clone the voice and match the user's conversational style.
1
Users can choose whether the agent can respond to customers via call or text. Additionally, users can record their voice by selecting an option from the functions or connectors module, allowing our AI to clone the voice and match the user's conversational style.
2
Users can review the flow and assign roles based on the settings and escalation matrix they configured during onboarding, which can be edited later.
The AI agent will route the query to a human agent if the response is false.
2
Users can review the flow and assign roles based on the settings and escalation matrix they configured during onboarding, which can be edited later.
The AI agent will route the query to a human agent if the response is false.





3
The AI gathers the query from the user and responds back to the customer.
3
The AI gathers the query from the user and responds back to the customer.
Prototype
Prototype
Mockups
Mockups






Future Innovation & Scope
Future Innovation & Scope
NOVA-AI started as a design challenge response, but its potential extends far beyond.
Next, I plan to conduct user testing with real users and business teams to validate the experience, uncover friction, and fine-tune the product for practical use.
Looking ahead, the goal is to evolve NOVA-AI into a smarter, more adaptive system, with collaborative learning, privacy-safe personalization, and a simplified setup that makes it accessible even for small businesses.
What began as a fix for frustrating support flows is now a blueprint for intelligent, action-first customer care.
NOVA-AI started as a design challenge response, but its potential extends far beyond.
Next, I plan to conduct user testing with real users and business teams to validate the experience, uncover friction, and fine-tune the product for practical use.
Looking ahead, the goal is to evolve NOVA-AI into a smarter, more adaptive system, with collaborative learning, privacy-safe personalization, and a simplified setup that makes it accessible even for small businesses.
What began as a fix for frustrating support flows is now a blueprint for intelligent, action-first customer care.