Kenny Yukich

Most AI programs fail at the last mile. The model works, the demo lands, and then nobody uses the thing.

My work is that last mile. I run Microsoft Copilot adoption across 69 licensed users at a 200-person manufacturer where most of the workforce is on the floor, not at a desk — so every tool has to justify itself in minutes saved, not capability demonstrated. I build the agents and automations myself, in Power Platform, Copilot Studio, Python, and Azure, and I measure them by whether people still use them in week three.

I came to this from the parts department. I spent 2025 pulling orders, taught myself retrieval systems on nights and weekends starting in late 2022, and moved into this role in January 2026. That path is why I'm skeptical of AI that looks impressive and saves nobody any time.