The article says they could have asked an AI themselves, but that is missing that I may have an agentic context with exactly the right information to answer the question well, which they do not have. For example because I worked on something closely related. Such an agent output is very useful, and taking the time to rewrite that to a short personal answer is not always worth it.