LLMs suck at executing business logic, stick to using them for user-interface only. Avoid relying on LLMs for decision-making due to issues with performance, debugging, and other limitations. LLMs are better suited for tasks like transformation, categorization, and communication, not complex decision-making. While the capabilities of LLMs are constantly evolving, it’s best to use them for specific tasks that play to their strengths, while crucial logic should be handled by purpose-built systems. Stay cautious about pushing the boundaries of LLMs in applications and remember to leverage their strengths while being aware of their limitations.
https://sgnt.ai/p/hell-out-of-llms/