video1.0<iframe src="https://www.loom.com/embed/d442996affe14bdea81014183f633988" frameborder="0" width="2128" height="1596" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>15962128Loomhttps://www.loom.com15962128https://cdn.loom.com/sessions/thumbnails/d442996affe14bdea81014183f633988-ab4c83feeea0cb8c.gif1701.846In this video, Wesley demonstrates how to use GitHub Copilot Agent in Agent Mode to build a complex business logic function using a test-driven development (TDD) workflow. The example focuses on calculating an invoice total with tax rules, discounts, and edge cases. Wesley starts by prompting the AI to create a TDD plan in a markdown checklist, which outlines one task per test and implementation step. He emphasizes the importance of small batch sizes and structured prompting to achieve better AI results. The Copilot Agent iterates through each checklist item, creating tests and implementation code step by step, with Wesley confirming commands for safety. As tasks are completed, the markdown file is updated in real-time. Once the full logic is implemented and passing all tests, Wesley prompts the AI to refactor the main function to reduce complexity. The AI successfully splits the logic into smaller functions, improving readability while maintaining test coverage. Wesley notes that while the result isn’t perfect, it’s a strong foundation and demonstrates how TDD supports clean iteration, validation, and safe refactoring with AI. He concludes by encouraging developers to explore this approach for business-heavy logic where clarity and accuracy are critical.