
Vibe Coding and Human in the Loop: Insights from Directions EMEA 2025
This year, I attended the most important Microsoft event for the first time — Directions EMEA 2025 in Poznań. Over three days (November 2–4), I dove into the world of innovative technologies alongside my supervisor, Dr. Maciej Suchomski. We attended a variety of sessions: while Dr. Suchomski focused primarily on the latest developments in the Microsoft ecosystem and talks with high operational relevance, I concentrated mainly on topics related to artificial intelligence.
The event featured over 300 sessions centered on AI agents, Copilot, and their integration with Business Central. A key highlight for me was Vibe Coding — an approach that understands programming as a dialogical process between the developer and AI. The transformation of the development process was especially evident in the session “Microsoft Presents: Vibe Coding” by Steffen Balslev and Vincent Nicolas. They demonstrated how AI agents can autonomously work in real time using linguistically formulated tasks: developers state requirements in natural language, Copilot interprets the context, generates runnable code, suggests optimizations, and manages repetitive workflows. Technologies like the Model Context Protocol (MCP) ensure agents receive precise instructions and operate reliably within the system. This creates an ecosystem of agents, Copilot features, and context protocols that reshape the way developers work. Live demos made clear how seamless the interaction between human and machine has become and how much it simplifies daily tasks. Microsoft expects that within the next two to three years, a large portion of coding could be automated, as new agents will not only generate code but also handle planning, validation, and execution processes.

Personally, this new Microsoft ecosystem signifies a palpable shift in the developer’s role: moving away from pure coding toward orchestrating intelligent systems that combine creativity, architectural insight, and strategic problem-solving.
At the same time, new opportunities arise: increased efficiency, time savings on routine tasks, and more room for conceptual and strategic work.
Beyond AI topics, I also attended practical best-practice sessions. One particularly interesting session was “Advanced Coding for OGs: Best Practices Post-Update” by Arend-Jan Kauffmann and Eric “Waldo” Wauters. They shared 43 concise tips for AL developers in Business Central, focusing on performance, AI compatibility (e.g., for Copilot agents), and avoiding bottlenecks in real-time applications like logistics integrations. Each tip was accompanied by a short AL code snippet, an explanation, and a concrete use case — an extremely valuable toolbox for developers and ISVs.
Conclusion
Maciej Krzan, PhD candidate and developer at XTRAS forward thinking






