Being able to call custom assistants (which customization happens in OpenAI's side) besides the base models would be very helpful. The use case for all of us using AI-Forms is clear: developers would be enabled to do many more things (customizing the assitants with context, decision trees, etc) but keep the inputs asked from end users as simple as possible (the value of AI-Forms is this, end users not having to interact with a chatbot, just copy your inputs and submit).
As a bonus, being able to call different "assistants" based on the user logged in, would open many business opportunities. The same applies for AI-Forms, fine-tuned models, etc. Developers could charge for user-based customization and serve the different models/assistants/prompts to each user separately. That revenue potential is now not possible unless you get deep in custom code. @weak cedar, please think about it, it's not just custom context anymore, it can now be extended to assistants and client-level customized models. Solving this bottleneck, that requires coding otherwise, would be huge for AI-Engine.