Don't overestimate domain expertise

Using LLMs for domain research in hospitality management revealed the same pitfalls as working with domain experts: information overload, jargon-heavy terminology rooted in legacy systems (like Oracle OPERA), and solutions presented instead of problems. LLMs compound these issues by blending vocabularies from competing systems, hallucinating coordination commands, and acting as yes-men that reinforce the prompter's existing beliefs. The author argues that ubiquitous language in DDD is a cognitive tool, not gospel, and that engineers should not try to become domain experts but rather collaborate with them. LLMs can help with initial orientation and organizing findings, but cannot replace the iterative feedback loop, critical thinking, and collaboration required for proper domain modeling.