Most generative AI projects don't fail because the technology doesn't work — they fail because the content fed into the system isn't structured, scoped, or quality-controlled. In this panel, Lief Erickson, Fabrice Lacroix, and I discuss the patterns we've seen across real projects, including why "almost right" answers are more dangerous than obvious hallucinations. If you're a content professional trying to figure out where you fit in your organization's AI initiative, this conversation is for you.