Where AI actually belongs in a .NET-product

This page is the post template. Everything below is scaffolding: keep the structure, replace the words with what you actually learned on a real project.

The problem

Open with the situation as it was before: what people did by hand, how long it took, and why it mattered. One concrete example beats three general statements.

What we tried

Describe the first approach, even the one that failed. Mention the stack in plain terms — .NET service, existing database, the model you called — without turning the post into a tutorial.

What actually shipped

Explain the version that stayed in production: where the model sits in the flow, what it is not allowed to do, and what a person still checks afterwards.

What I would do differently

• One lesson you would tell a colleague over coffee.
• One thing you measured, or wish you had measured.
• One warning for anyone copying this approach.

When you are happy with the text, remove the placeholder badges from the blog list and this page.