Yes, and it is the highest-converting AI use case measured — 26% of the panel uses AI to build the revenue system itself, and every one of them has it in production. The working pattern is that the model drafts and an engineer approves, not that the model deploys.
Keep a human on schema and business-logic decisions. The model is strong at inventory, cross-referencing dependencies and generating configuration; it is weak at knowing what a field means to a team and what will break downstream. Sandbox first, and keep the change reversible.
Every claim above comes from work LeanScale published. These are the sources.