Automated Website Migration Using AI

Websites differ in their technical implementation, content structure and presentation. Simply extracting text and images is not enough—they must also be interpreted and converted into the appropriate structures of the target content management system.

Initially, we pursued a more predictable, deterministic approach, using AI for individual, precisely defined tasks. However, this approach was not flexible enough to handle the diversity of real-world websites.

As development continued, we therefore moved toward a more general agentic approach and also focused on evaluating and controlling migration quality.

I came up with the initial idea and designed the main technical solution for the project. Unlike when working independently, I focused primarily on planning the work, coordinating developers and making technical decisions.

Implementation was the work of the entire team. My role was to help maintain the technical direction, evaluate the chosen approach and adjust the way forward based on what we learned during implementation.

This case study is intentionally anonymized: it describes the general purpose of the solution and my role, rather than internal documentation, infrastructure architecture or business metrics.