With the rise of tools like Lovable, v0, Replit and Bolt, the ability to spin up full-stack prototypes via natural language is becoming a commodity (IMO)

I’m curious if any engineering teams are building internal versions of these tools specifically for their Product Managers/Non-tech roles to self-serve?

We are currently debating whether to build a lightweight wrapper around an LLM that has context of our design system and internal APIs, allowing PMs to:

- Generate high-fidelity prototypes without bothering designers.

- Build simple internal dashboards (CRUD apps) without engineering tickets.

If you are doing this:

- How "production-ready" is the output?

- Are you using a generic model (Claude 4.5) or fine-tuning?

Do you let PMs deploy, or is it strictly for prototyping? And another tangent question: are software engineers moving from "coding an app" to "coding a platform that codes the app", lol