I built Susan AI as an experiment to explore what a practical, personal AI assistant could look like when it is designed around real-world workflows rather than just conversation.
🔗 Live Demo: https://susan-ai.vercel.app/
💻 GitHub: https://github.com/susankarkarmakar-pixel
Why I Started Building Susan AI
The easiest way to use AI today is to open a chatbot, type a question, and get an answer.
But while using AI for different tasks, I kept asking myself a different question:
What if an AI assistant could become a practical workspace instead of simply being a chatbot?
I wanted something that could help me with:
- Learning and research
- Writing and editing
- Programming
- Understanding documents
- Summarizing information
- Brainstorming ideas
- Working with structured data
- Exploring the web
- Building AI-assisted workflows
That idea eventually became Susan AI.
Susan AI is my personal AI assistant project and an ongoing experiment in combining generative AI, web development, productivity tools and agent-style workflows.
What Is Susan AI?
Susan AI is a web-based AI assistant designed to provide a unified environment for interacting with AI.
The project is accessible directly from a browser:
👉 https://susan-ai.vercel.app/
Instead of thinking of it as another ChatGPT clone, I am developing Susan AI as a personal AI workspace.
The long-term goal is to make the assistant useful across multiple categories of work:
Susan AI
│
┌───────────────┼────────────────┐
│ │ │
Learning Creating Building
│ │ │
Research Writing Coding
Study Content Debugging
Analysis Reports Development
│ │ │
└───────────────┼────────────────┘
│
AI Workflows
│
Personal Productivity
The Core Idea
Most AI assistants begin with:
Question → Answer
I am interested in exploring something broader:
Goal → Context → AI reasoning → Tools → Output → Workflow
For example, instead of asking an AI only:
"What is climate change?"
a more useful assistant could help a user:
- Understand the topic
- Research relevant information
- Summarize documents
- Organize notes
- Generate a report
- Create study material
- Continue working with the result
That shift—from conversation to workflow—is one of the main ideas behind Susan AI.
Current Capabilities
Susan AI is being developed around several practical use cases.
1. AI Chat
The foundation is an AI conversation interface.
Users can ask questions, brainstorm ideas, explore concepts and receive AI-generated responses.
This provides the basic conversational layer on which the other capabilities can be built.
2. Learning and Research
One of the areas I care about most is using AI as a learning companion.
Susan AI can be used to:
- Explain difficult concepts
- Break complex topics into simpler sections
- Generate study material
- Brainstorm research ideas
- Summarize information
- Compare concepts
- Ask follow-up questions
The objective is not simply to provide an answer, but to help users understand why the answer makes sense.
3. Writing Assistance
Generative AI is extremely useful as a writing partner.
Susan AI can assist with:
- Articles
- Reports
- Emails
- Documentation
- Ideas
- Rewriting
- Summaries
- Content planning
One of my goals is to make AI-generated content more useful by keeping the human author in control.
AI should assist the writer—not replace the writer.
4. Coding Assistance
Another major area is software development.
I use AI extensively while exploring and building web applications, so coding assistance naturally became part of Susan AI.
Potential workflows include:
Problem
↓
Explain the problem
↓
Generate an approach
↓
Write code
↓
Review code
↓
Find errors
↓
Improve implementation
This makes the assistant useful not only for experienced developers but also for people who are learning programming.
5. Document and Knowledge Work
Real-world work rarely happens only inside a chat box.
Important information is often stored in:
- PDF files
- Text files
- CSV files
- JSON data
- Reports
- Notes
- Documents
A useful AI assistant therefore needs to work with information supplied by the user.
This is an important direction for Susan AI: moving from general AI knowledge toward context-aware assistance.
6. AI + Web Research
Another area I am exploring is combining AI with web-based research.
The basic concept is:
User Question
↓
Research
↓
Relevant Information
↓
AI Processing
↓
Structured Answer
The challenge is not simply retrieving information.
The real challenge is turning scattered information into something understandable, useful and actionable.
7. Agent-Style Workflows
One of the most interesting directions for Susan AI is AI agents.
A conventional chatbot waits for a user message and returns a response.
An agent-style system can potentially:
Understand Goal
↓
Plan
↓
Use Tools
↓
Process Information
↓
Take Actions
↓
Return Result
I am experimenting with this direction because I believe the next generation of AI applications will increasingly focus on task completion rather than text generation alone.
Technology
Susan AI is a modern web application deployed on Vercel.
The project uses a web-based architecture designed to keep the interface accessible directly from the browser.
The AI layer currently uses Google Gemini Flash-Lite.
The overall architecture is evolving as I experiment with new capabilities.
A simplified conceptual architecture looks like this:
┌─────────────────┐
│ User │
└────────┬────────┘
│
▼
┌─────────────────┐
│ Susan AI UI │
└────────┬────────┘
│
▼
┌─────────────────┐
│ AI Application │
│ Layer │
└────────┬────────┘
│
┌────────────┼────────────┐
▼ ▼ ▼
AI Model Tools User Data
│ │ │
└────────────┼────────────┘
▼
┌─────────────────┐
│ AI Output │
└─────────────────┘
The architecture is intentionally being kept flexible because I expect the project to evolve considerably.
Why I Chose to Build It Myself
There are already many excellent AI assistants.
So why build another one?
Because building something yourself teaches you things that simply using a product cannot.
While developing Susan AI, I get to explore:
- AI API integration
- Prompt engineering
- AI UX
- Web application architecture
- Authentication concepts
- State management
- API design
- Error handling
- AI response streaming
- File processing
- Agent workflows
- Deployment
- Performance
- Product design
The project is therefore both a product experiment and a learning laboratory.
What I Have Learned
One of the biggest lessons has been that building an AI application is not simply about connecting an LLM API.
The model is only one component.
A useful AI product also needs:
1. Good UX
Users should understand what the AI can do without reading a manual.
2. Good Context
The AI needs the right information at the right time.
3. Good Prompts
Poor instructions can produce poor results even with a powerful model.
4. Good Error Handling
AI systems can fail in unexpected ways.
Applications therefore need graceful handling of:
- API failures
- Invalid input
- Rate limits
- Network problems
- Unexpected responses
5. Trust
AI applications should clearly communicate limitations.
A confident answer is not necessarily a correct answer.
That is especially important when AI is used for research, government work, education or decision support.
The Bigger Vision
Susan AI is not finished.
In fact, I consider the current version the beginning.
My long-term vision is to explore a personal AI environment where different capabilities can work together.
For example:
Susan AI
│
┌───────────┼───────────┐
│ │ │
Research Writing Coding
│ │ │
└───────────┼───────────┘
│
Documents
│
▼
AI Agents
│
▼
Automated Workflows
The ultimate question I want to explore is:
Can a personal AI assistant become a useful digital workspace for an individual's everyday intellectual work?
I don't know the final answer yet.
That's exactly why I am building it.
What's Next?
Some of the areas I want to explore further include:
- Better AI memory and context
- More reliable document understanding
- Advanced AI agents
- Tool calling
- Multi-step workflows
- Better web research
- Improved coding assistance
- Personal knowledge management
- More structured outputs
- Voice interaction
- Better mobile experience
- AI-powered productivity workflows
The roadmap will evolve as I continue experimenting.
Building in Public
I am sharing Susan AI because I believe the process of building AI applications is as interesting as the final product.
There will be experiments that work.
There will also be experiments that fail.
Both are useful.
I plan to continue documenting the development process, technical experiments, design decisions and lessons learned.
If you are a developer, student, researcher, AI enthusiast or someone experimenting with AI applications, I would genuinely appreciate your feedback.
Try Susan AI
🌐 Live Demo:
https://susan-ai.vercel.app/
💻 GitHub:
https://github.com/susankarkarmakar-pixel
If you try it, I'd love to know:
What would you want your personal AI assistant to do that current AI chatbots don't do well?
That question may shape the next version of Susan AI.
Final Thought
We are moving from an era where software simply waits for instructions toward an era where software can increasingly understand goals, reason about tasks and assist with execution.
Susan AI is my small experiment in that direction.
It started with a simple question:
What would I build if I wanted my own AI assistant?
I am still exploring the answer.
And that's the most interesting part.
Built by Susankar Karmakar
Top comments (0)