Clayton Rodgers
← All case studies File No. 03

HR Buildathon

For: Dentsu International Built with: Microsoft Copilot

The problem

Most AI-in-HR programming follows the same shape: a vendor demo, a training deck, maybe a pilot license handed to a few people. None of it asks anyone to actually build something themselves, and that gap between watching AI get used and using it yourself is exactly where adoption stalls. I’d seen this firsthand: after running a live, two-hour build session with my own team where I built something real in front of them instead of presenting about AI, people started asking for access unprompted, the same week, instead of waiting for a formal training program. The barrier to adoption was never really about the technology or even about information; it was about permission, the unspoken assumption that “building” is a job for specialists, not for HR generalists sitting in the room.

That single session doesn’t scale as a habit for a team of one, which raised the real question: what happens if you turn that same premise into an event, put the tools directly into HR practitioners’ hands, and have them build, instead of sitting through another AI-in-HR presentation.

What I built

I pulled together my own team of HRBPs for a three-hour session. We started with a walkthrough of Copilot itself, what it can actually do and how to prompt it well, then spent the rest of the time each building something that solved a real problem in our own day-to-day work.

What happened

The event increased AI utilization across the HR function by 60% and directly produced the AI Coaching Agent documented elsewhere on this site. Just as important as anything that got shipped, it built real confidence in the team: people walked out having actually built something with AI themselves, not just watched a demo of it.

What I’d do differently

I’d make it recurring instead of a one-off. A single three-hour session is a good jolt of confidence, but the habit of building fades if it isn’t reinforced on some regular cadence.