In this session, we will present real, hands-on use cases showing how to standardize best practices in AL development for Microsoft Dynamics 365 Business Central using GitHub Copilot Skills within a custom agent framework. The goal is clear: to achieve consistent and predictable results, regardless of the developer or the project.

We will demonstrate how to move from a generic use of Copilot, often prone to interpretation and “hallucinations”, to a more deterministic model, where architectural rules, AL patterns, and quality standards are defined as executable skills. In this approach, Copilot stops improvising and starts applying concrete technical criteria while generating code.

Through examples covering integrations, functional extensions, and refactoring scenarios, we will show how this approach reduces variability, avoids recurring errors, and ensures that generated AL code consistently follows the same best practices, turning AI into a reliable ally for professional Business Central development.