Christian Crumlish on Bringing Your Agents to Work

Christian Crumlish went from casual AI user to building Piper Morgan, an AI agent designed to collaborate with product teams. Along the way, he discovered something most of us are starting to feel: personal AI workflows are everywhere, but team AI workflows are a mess.

You’ve probably got your own AI habits by now. A prompt you reuse. A tool you swear by. A workflow that saves you hours. But what happens when your whole team starts doing the same thing? When everyone’s got their own agent, their own outputs, their own version of “done”?

That’s where things break.

In this conversation, we’ll dig into what changes when AI moves from single-player to multiplayer mode. What coordination mechanisms do teams actually need, shared vocabulary, agent protocols, documentation, governance? How do you make AI-generated output usable by humans who didn’t generate it? And what are teams outsourcing to AI that they really shouldn’t?

We’ll also talk about what this means for product managers specifically. If shipping is getting faster, what becomes the real job? What should PMs own that agents can’t? And when prioritization gets harder because everything is suddenly possible, what judgment needs to remain non-delegable?

Christian has been thinking about this longer than most of us. He’s building tools for it, writing about it, and living it with real teams. If you’re trying to figure out how your team should actually work with AI, not just use it, but operate well together, this one’s for you.