How I Built a Real Automation Framework from Scratch Using Vibe Coding (Part 1 )

Automation is changing fast and tools powered by AI are making it easier than ever to build frameworks from scratch. In this blog, I’ll walk you through how I created a live automation framework using Generative AI tools like Cursor, Perplexity, and Aqua.

Whether you’re a beginner, manual tester, or just starting out in automation, you’ll find this step-by-step guide helpful. Let’s get started.

Why This Blog Matters for QA Professionals

If you’re still doing only manual testing and haven’t started learning automation yet—now is the time. Not only is automation growing, but AI is also evolving fast, and if we don’t keep up, we risk being left behind.

Recently, many of my students asked, “Can we build an automation framework from scratch using AI?”

So I decided to do it myself starting from zero and here’s exactly what I did.

Step 1: Understanding What We’re Building

I picked a simple example: a login page with a dashboard.

The app flow is:

• If login is successful, the user is taken to the dashboard
• If login is unsuccessful, an error message appears

This is a basic, real-world use case that many websites follow.

Step 2: Creating a Product Requirement Document (PRD) with AI

The first thing we need in any automation project is a Product Requirement Document (PRD).

Here’s how I created it:

1. I took a screenshot of the login and error pages.
2. I used a tool like Perplexity AI and gave it a detailed prompt: “Act like a Business Analyst with 20 years of experience. Based on the screenshots provided, create a PRD for this login application.”
3. The tool generated a clean, structured PRD, which I used as the base for the rest of the project.

Step 3: Writing the Test Plan Using AI

Next, I needed a test plan.

Using the PRD generated above, I gave the AI another prompt:

“Based on this PRD, create a test plan in a professional format.”

The AI created a well-organized test plan, including:

• Test objectives
• Scope of testing
• Test environment
• Schedule and deliverables

And yes—you can use this same approach for your projects too.

Step 4: Generating Test Cases Automatically

With the test plan in place, it was time to write the test cases.

Again, I used AI (like ChatGPT or Perplexity) and asked:

“Create 8 detailed test cases in tabular format for the login page, including both valid and invalid scenarios.”

In a few seconds, I had a clean list of test cases, including:

• Positive test (valid login)
• Negative test (invalid password, blank fields, etc.)

You now have Step 1 to Step 3 of a full testing cycle:
✅ PRD
✅ Test Plan
✅ Test Cases

Bonus: Test Strategy & Bug Report Templates

Beyond the basics, I also asked the AI to generate:

• Test Strategy Document
• Bug Reporting Format
• Test Matrix

With the right prompts, the AI gave me everything I needed to manage a complete testing cycle from start to finish.

If you want these prompts or templates, just comment “Resources” and I’ll share them.

Step 5: Tools That Made It Easy

Here are the tools I used:

• Cursor: A modern AI-powered IDE (we switched from IntelliJ to Cursor recently)
• Perplexity: Great for generating documents and prompts
• Aqua: Converts audio to text, very useful during demo sessions

I’m offering a 1-month free trial of Cursor Pro for those who want to try it. Just type “Cursor” in the comments and I’ll share the link.

What’s Next: Building the Automation Framework (Part 2)

This blog was all about preparing the documentation—something many testers skip, but it’s crucial.

In Part 2, I’ll show you:

• How to take the test cases and build a real working Selenium automation framework using Cursor
• How to set up Page Object Model and run tests
• How to integrate reporting, logging, and more

Final Thoughts

This is not the future this is happening right now. AI is already helping us as testers do better, faster, and more efficient work.

Here’s a quick recap of what we did:

1. Took screenshots of the app
2. Used AI to generate the PRD
3. Created a test plan using prompts
4. Built test cases (positive + negative)
5. Prepared strategy and bug reporting formats

Everything was built from scratch using AI—no need for hours of manual typing or guesswork.

If you want the templates, prompts, or Cursor trial—just comment “Documents” or “Cursor” below.

And if you liked this blog, do let me know! I’ll keep sharing more real-life QA automation journeys like this.

Thanks for reading 🙏
— Pramod Dutta
SDET | QA Mentor | 14+ Years Experience