Six questions before you add an LLM

# Summary Cameron Palmer, an AI consultant, shares lessons from a failed attempt to use a single LLM agent for his entire lead discovery pipeline, which resulted in 10-20% error rates and was difficult to debug. He argues that companies should stop asking "where can we apply AI?" and instead identify specific problems first, then determine if an LLM is the appropriate tool—recognizing that LLMs trade determinism and repeatability for flexibility, making them unsuitable for processes requiring consistent, predictable results.