◇ Could this help me?
Yes, in part. The structural idea worth taking is the prototype-versus-production split: letting non-engineers build throwaway prototypes with AI while production code stays behind engineering review is a sound way to get speed without shipping unreviewed work, and naming an explicit owner for agents and guardrails is sensible. Do not adopt the underlying premise that AI only makes the best people more productive, or the layoff framing around it. That is one executive's contested claim, presented here as reporting rather than as an established finding. This is a general practice.
I want to separate prototyping from production work in this project. Propose a workflow where anyone can build a prototype with AI in an isolated area of the repo, with clear rules on what a prototype may never do, and where promoting anything into production requires review against a checklist you draft. Include what that checklist should cover for correctness, security, and data access.