Building a Web Version in 10% of the Estimated Time Using AI

When your product already runs successfully on mobile, creating a web version often feels like starting over once again. That was the case for Dovkolo Club, a platform for Ukrainian relocants for easier adaptation abroad, created by a team from Sigma Software and supported by our venture wing Sigma Software Labs. Their mobile app had been smoothly operating since 2023, and the team wanted to bring the same experience to the web — but with minimal investment, because the goal was to produce an internal version for customer development.

Traditional development approaches would require a 400+ hour effort to design and build a web version — even with a simple UI stack. But instead of following the common path, the team decided to try something different.

The Challenge

The task seemed simple: create a web version that mirrors the mobile experience. The constraints?

• Minimal time and budget, because this was an MVP
• Fast iteration, with no room for lengthy design or development cycles
• No duplicated effort — business logic had to be reused, not rewritten

Although the original app was built with Flutter, reusing it for the web wasn’t viable. Flutter can build web apps but making them look native to the web would require more effort than we planned to invest.

So, the team asked themselves: Could AI help us deliver faster?

The answer turned out to be yes.

Our Approach: AI-Assisted Development with Cursor-Assisted Development

To speed up the process, one of our developers — surprisingly, a mobile engineer — proposed to try Cursor, an AI-powered coding tool. The team already had experience with it, making it a safe starting point for this experiment.

The mobile app served as the primary source of truth:

• Flutter as cross-platform UI framework
• Node.js backend
• Stable API endpoints

Initially, they considered using Swagger documentation as reference, but most business logic lived inside the mobile app code. Therefore, the developers pointed the AI directly to the actual codebase.

How the Workflow Looked

1. Expose the mobile codebase to Cursor
   The AI gained full context about existing flows, logic, and end points usage.
2. Generate each feature, page by page
   Cursor created initial Material UI web components.
3. Adapt authentication and core logic
   Since all logic was already implemented in the mobile app, AI mostly handled translation into a web environment (React + TypeScript).
4. Perform quick manual reviews
   Every generated feature went through a 15–30 minute code review: adjusting edge cases, improving structure, and fixing inconsistencies.
5. Refine design
   AI generated a clean but generic layout. As the designer put it:
   “We need to give it some love.” Manual UX polishing was still essential.

What We Used

• React + TypeScript for the front-end
• Existing Node.js backend via API
• AI-driven scaffolding for UI and logic
• Human reviews on each iteration

The Result: A Working Web Version in One Week

After just one week, a fully working demo version of Dovkolo Club website saw the world.

Not perfect, of course – some features were missing and bugs needed fixing – but for an MVP, it accomplished everything required:

• The core flow worked
• Focus group users could explore the platform
• The team had a sandbox to validate their hypotheses

Total time spent: ~40 hours to generate core functionality with AI, reviews, and adjustments instead of 400+

Making a production version would have required another ~40 hours for bug fixing, but still the result would have been dramatically faster than a traditional approach.

Key Takeaways

AI enables cross-technology flexibility: A mobile developer successfully shipped a web version. AI bridges gaps between frameworks and languages. technology flexibility

AI is ideal for MVPs and demos: When speed matters more than polish, AI-driven development drastically reduces effort.

Iterate fast, polish later: AI builds the raw version. Designers and developers finish the job.

AI doesn’t eliminate human input – but it reshapes it: Even with detailed prompts, manual reviews and UX adjustments were essential.

This wasn’t a flawless process. Bugs remained, UX needed refinement, and some features required manual intervention. But the goal was reached: a functional, testable web version delivered at a fraction of the cost and time.

For Dovkolo Club, this experiment proved that AI-assisted development can dramatically accelerate demo or MVP version creation. For the team, it became a blueprint for how AI can support — not replace — developers in delivering faster, smarter, and more efficiently.

• #AI/ML