Despite the Microsoft AI promises, we still struggle with generic outputs, confused Copilot responses, and the constant need to double-check everything the algorithms spit out. This post explores the issues and suggests a few fixes.