This is a thoughtful question, and it’s good that you are separating design‑time generation from runtime behavior.
What is confirmed and safe to state:
The .blueprint file is a structured design artifact, not executable code
It captures case structure, data models, personas, and intent, not finished implementation
On import, Pega uses Blueprint to generate governed rule scaffolding that remains fully editable and reviewable
What is important from a Predictable AI perspective is that GenAI is used at design time, not as an autonomous code generator at runtime. The AI proposes structures; humans still review, adjust, and approve the resulting rules.
If others have insights from Blueprint‑driven delivery projects, especially around review and validation practices after import, that would add a lot of value to the discussion.
Blueprint is proprietary format. Import is a process of creating a rules basing on metadata information included in .blueprint file. As per me knowledge we don’t call LLMs during this process. GenAI is used by Blueprint itself.
Having a .blueprint file you can import it also on non GenAI enabled server.
It’s all about High Fidelity. You do the work on the Blueprint and we convert that into a working application - getting the details in the Blueprint means fare less development or refactoring later (e.g. Adding persona RBAC in Blueprint is commonly missed and far quicker to do in the Blueprint than manually in Dev Studio).
If you want more details on Importing, there are a series of videos from the Blueprint Expert Circle: