I think the "safer" areas are definitely in applying the Gen AI to very specific use cases, where you can build a powerful, deep experience for users in ways that ChatGPT alone cannot.