We delivered an AI operations manager for an oyster farm working leases in the ACE Basin — one that plans every harvest around the tide windows and shellfish-bed closures that rule the business, keeps the temperature-and-tag paper trail regulators audit without warning, matches each week's harvest to standing restaurant orders, and forecasts which cages hit market size next

The farmer works two leases south of Charleston, a few hundred floating cages of single oysters bound for the raw bars downtown. His product is superb. His business ran on a tide chart, a whiteboard, and a phone full of text threads — and all three were failing him at once. The harvest math alone is brutal: a workable window is where a low tide, daylight, crew availability, and an open shellfish bed all intersect, and a rainfall closure from DHEC can erase a planned harvest the night before while eleven restaurant orders sit on the whiteboard for Friday. The compliance load is heavier than outsiders imagine: every bag tagged, every cooler temperature logged from deck to delivery, every record kept audit-ready for the regulators who arrive unannounced — one broken cold-chain entry and a whole harvest is at risk. And the farming itself is a memory problem: which cages were seeded when, which got flipped, which are running toward three inches — knowledge that lived entirely in his head and left with him on the days he was hauling. We built him an operations manager that holds the whole farm. It pulls the tide tables, the weather, and the state's closure notices every morning and lays out the week's real harvest windows — not the theoretical ones — flagging the moment a rain event upstream threatens a closure so he can pull orders forward instead of eating them. It keeps the harvest log the way the regulation is written: tag numbers, times out of the water, cooler temps, chain of custody, all of it entered from a phone on the boat with wet hands and all of it audit-ready. It matches the standing orders against what the cages can actually give that week and drafts the messages to the chefs when the numbers move. And it tracks every cage cohort from seed date forward, projecting size from his own historical growth curves, so replanting and rotation became a schedule instead of a guess. First month, an upstream storm closed his best bed on a Wednesday; the system had already moved his harvest up two days and his orders shipped full. "The tide used to run this farm," he said. "Now we run it together."