The author's initial experience with Large Language Model AIs (LLMs) was disappointing, as they expected perfect accuracy but received nonsensical responses. However, the author came to understand that LLMs are not reasoning tools and should be used to manage cognitive load rather than replace human judgment. To effectively leverage LLMs, a systematic approach is necessary, involving categorizing tasks based on complexity, predictability, and human judgment required. A framework for task automation can help identify tasks suitable for LLM automation, such as repetitive or solved problems. The author also emphasizes the importance of designing workflows that maximize efficiency while ensuring human oversight and control. By striking a balance between human and machine intelligence, individuals can harness LLMs' power to automate mundane tasks and augment their intelligence, ultimately creating a future where humans and machines work together.