We delivered an AI seafood broker for a family dock on Shem Creek — turning a captain's crackling radio call into priced offer sheets in forty chefs' phones before sunrise, pricing the catch against the real market instead of gut feel, and writing the traceability paperwork nobody had time for

A third-generation dock and wholesale operation on Shem Creek came to us with a business that ran entirely on one man's memory and a flip phone. Every morning before dawn, the boats would radio in what they were bringing back — so many boxes of white shrimp, a few hundred pounds of triggerfish, two coolers of soft-shell crab if the moon had cooperated — and the owner would start working his phone, calling and texting the restaurant chefs he had known for twenty years, matching fish to kitchens from memory: this one takes all the swordfish he can get, that one wants shrimp headed and never frozen, the new place downtown will pay up for anything unusual if you call them first. It worked, in the way things work until they don't. Chefs who texted after 7 a.m. got whatever was left. Pricing was gut feel calibrated years ago — he was underpricing the premium catch on busy mornings because there was no time to think, and eating spoilage on the odd-lot species nobody remembered to offer. The federal and state traceability paperwork — where each fish was landed, by which vessel, on what trip — was a shoebox of deck slips that took his daughter two full days a month to reconcile, and one audit letter away from being a real problem. And the whole operation had a single point of failure: the matching engine was a sixty-four-year-old man's memory, and he wanted to retire fishing, not brokering. We built the family an AI seafood broker that starts where their morning starts: the radio call. A deckhand or the owner speaks the manifest into a phone — species, counts, sizes, which boat, which trip — and the AI structures it into a live inventory before the boat is at the dock. Then it does what the owner did from memory, but for every chef at once: it knows each restaurant's standing preferences, order history, and what they paid before, and by 5:30 a.m. every chef gets a personal offer sheet — not a generic list, but their fish, in their quantities, at a price the AI sets against what the catch actually is worth that morning: landings up and down the coast, what moved yesterday, what is scarce this week. First-come chaos became a fair window: chefs claim from their sheet by reply, the AI allocates against inventory in real time, and the odd lots that used to spoil now go out as a flagged opportunity — the two coolers of amberjack nobody used to remember to offer sold within the hour to the downtown place that loves a story for the menu. The paperwork layer runs underneath all of it without anyone thinking about it: because every lot enters the system tagged to a vessel and a trip from the moment it is radioed in, the traceability records that took the daughter two days a month now generate themselves, and when the state inspector visited in September, the audit that had haunted the family took forty minutes. The numbers after one full season: average realized price per pound up nine percent — almost entirely from the premium catch no longer being underpriced in the morning rush — spoilage down from roughly six percent of landed weight to under two, forty-one restaurants active on morning sheets versus the twenty-six the owner could physically call, and the daughter's two paperwork days returned to her every month. But the family's own measure is different: the matching engine no longer lives in one man's head. The owner still walks the dock every morning — he just does it as a fisherman again, and the business his grandfather started can finally outlive his memory.