Hong Kong is where that question gets asked first. The desks here run coverage across the mainland and the region, where a single position can rest on hundreds of filings, transcripts and local-language sources, and where the research task is honestly measured in days rather than minutes. A system that returns one good answer does not settle it. What matters is whether the work still holds when the market has moved underneath it. The afternoon follows that problem into the four places it actually lands: the buy-side desk, sell-side research, the data layer underneath both, and the risk function that has to decide when to pull an agent off the task. Two firesides, three panels and a live demo, followed by a reception.