Why AI Companies Want Your Support Staff's Data, Not Just Your Experts'

@ashishmehta1691 asked in the comments how you judge whether office work was done right when there is no single right answer. This video answers it: AI is tested most on work with a clear right answer (Carnegie Mellon and Stanford, 2026), the best AI agent in Carnegie Mellon's pretend company finished 38% of software tasks but only 13% of admin and 8% of finance tasks, and your support staff's data holds the written outcome (approved, paid, signed, on time) that shows what right looks like. Plus one property management repair request, start to finish, and where this data sits in your systems....