"We look at articles of AI not being as helpful as executives might think it is, and also now agentic AI can get around safety protocols."