
We've Governed Citizen Devs for Years, Governing Ai-Built Software Is the Same Problem
Sharing a November conversation with Scott Durow, Rami Mounla and Tomas Prokop recorded after PPCC, on where Power Platform governance goes as coding agents take over authoring.
Outtakes of interest:
- Organizations still govern the people building things and the code they ship, same as the low-code years. The difference is that one person driving a coding agent session ships a lot more code in a lot less time. The biggest practical change since the recording is that AI harnesses (Copilot CLI, Claude Code, Codex) have gotten a lot better at holding a plan together and keeping focused context across a session.
- Day-to-day, the work that makes agents productive is the work that makes people productive. Clear ask, right context, a review before anything ships. Under-brief a junior or an agent, both stall.
- On Power Platform specifically, the governance principles from the low-code years line up cleanly with what you need to govern agent-authored work.
This content was made in support of the upcoming Update Days: Power Platform conference. https://power.updatedays.cz
Also, the discussion focuses primarily on governance, so it doesn't go deep on using your favorite AI harness for Power Platform development. That's a topic we're dedicating a full post to soon, and one we cover extensively at the conference.