Claude Code Training by Modernizing a Real App

Interest in AI across the company was high. About 30 people attended the training, and the CEO stopped in to watch. The challenge was turning that interest into practical skill. The team supported 20 to 30 legacy applications, many built on older Microsoft technologies, and needed to see how an AI coding agent performs on the kind of aging, tightly coupled codebases they work with every day.

The company wanted its engineers to learn Claude Code on real work and leave with techniques they could apply right away. Leadership also wanted a clear view of where AI-assisted development was heading and how to plan for it. To make the training directly relevant, we chose one of their own applications as the teaching project: SalesForecast, an ASP.NET app on .NET Framework 4.x that hadn't been updated in 13 years.

The morning session was a hands-on tactics session. Shaun started with the team at 8 AM, walking through how to investigate an unfamiliar codebase and plan the work before writing any code. At 9 AM he kicked off the main Claude Code loop, and the team watched the work unfold until 1 PM, seeing how he guided the agent, reviewed its output, and kept a long-running task on track.

By the end of the morning, the team had seen SalesForecast converted from ASP.NET to React, 37 stored procedures moved into an Entity Framework ORM layer, the database migrated to PostgreSQL, and the app deployed to its own Azure resource group with cloud-based Microsoft Entra ID authentication. Because verification was part of the process, Claude Code also found and fixed 14 existing bugs, including several serious security holes. Throughout the session, Shaun shared practical tips and concepts that are easy to remember and apply, and answered the team's architectural questions about structuring and reviewing AI-assisted work.

To close the morning, Shaun logged into the old and new versions of SalesForecast side by side as the same user. The old app threw errors, and the new app worked cleanly, a concrete demonstration of what the team had just learned. The team said that watching an experienced engineer work through a real project was the most valuable part of the training.

In the afternoon, the same group returned for a strategy session covering what to expect from AI-assisted development and how to prepare for it, followed by an open Q&A. That evening, we met with senior technical management to discuss their plans in more depth.

This training shows the value of learning Claude Code from an experienced engineer on real work. By teaching through one of the company's own applications, Sembit gave the team practical techniques grounded in the code they support every day, along with a working example of what those techniques can accomplish. Leadership left with a clearer picture of what's coming and how to plan for it.