I don't disagree that very numerical tasks like revenue forecasting are not a good fit for LLMs. But neither did a lot of data scientist concerns themselves with such things (compared to business analysts and the like). Software to achieve this has been commoditized.