Software development is rapidly progressing from simple code completion tools to complex multi-agent systems, reflecting a broad adoption curve across the industry.
The developer's role is evolving from an individual code contributor to a manager of AI agents, emphasizing skills in validation, observability, and compliance.
Organizations can leverage fine-tuned AI models to enforce coding standards and best practices, providing effective guardrails that accelerate the growth of junior developers.
Strong organizational governance and a return to specification-led development are crucial for guiding AI tools to build valuable products and avoid wasted effort.
Microsoft Copilot uses a "skills" system, where documentation defines best-practice "golden pathways" to guide LLMs and ensure correct, efficient implementation.
Overcome model fatigue by using evaluation-led development and tools like Model Router, which automatically selects the most cost-effective model for a given prompt.
GitHub Copilot offers both a powerful command-line interface (CLI) for spec-driven work and a new desktop app that provides a user-friendly GUI for the same functionality.
New frameworks like Windows ML and Foundry Local allow developers to run hardware-optimized, compact AI models directly within applications on user devices for improved performance and security.
Jobs that call for the skills explored in this talk.