We delivered an AI Merchandise Planner for a two-store Charleston women's boutique buying from sixty labels a year -- one that moves every stranded size between King Street and Mount Pleasant overnight, takes the first markdown on every slow style on its own, reorders what is selling and chases every late ship window, while holding the second markdown, every cancellation and every new label for the owner.

The dress sold out in a size 6 at the King Street store in four days. The same dress, in the same size, hung on a rack on Coleman Boulevard for eleven weeks, and went to the sale rack at forty percent off in the same week a customer on King Street asked whether it could be found anywhere. That was the spring of 2025, and the owner only learned of it in July, going through the season's sell-through in a spreadsheet at her kitchen table on James Island. She opened the shop in 2012 after six years as an assistant buyer for a department store in Charlotte, and the second door in Mount Pleasant followed in 2019.

Eleven people work for her between the two. She buys from about sixty labels a year, most of them small, at four markets, and the stores take in roughly 1,400 styles a year across every size. Every one of those styles has a ship window, a size curve, a week where it stops selling at full price, and a second store it might have sold in instead, and for thirteen years the person holding all of that in her head was her. The point-of-sale system knew what sold.

The vendors' emails knew what was shipping. A legal pad under the King Street register knew which customers were waiting on which size. Nothing talked to anything else, and the Sunday night she spent reconciling them was the only time the business saw itself whole. The AI we delivered runs on her laptop, under her own account, and does the planning every night.

At 11:40, after both registers have closed, it reads the day's sales by style, size and store. Where a size has sold out at one store and the other holds two or more, it writes the transfer, prints the pick slip at the store that has the stock, and texts the opener there so the pieces are in the courier bag before ten. Where a style from one of the fourteen labels that ship at once has dropped under three weeks of supply, it emails the rep the reorder in the vendor's own format, by size, and logs the confirmation when it comes back. Where a style has reached its eighth week and sold through under forty percent at full price, it takes the first markdown itself -- twenty percent, in the point-of-sale system, with new tags queued to the label printer at both stores for the morning.

It reads every order confirmation that lands in the shared inbox and holds every ship window it finds. On the day a window closes without a tracking number, it emails the rep. Five business days after that, it writes the cancellation and puts it at the top of her morning note, because cancelling a purchase order is hers to do and nobody else's. The same is true of the second markdown at week twelve, of any reorder over $1,500 at wholesale, and of any label the stores have never carried.

Inside those lines it moves, reorders, marks down and chases on its own, and nothing it does ever drops a piece below cost, which was the first line she drew. The legal pad is gone. When a customer asks for a size the store does not have, the associate types her name and number into the note on the sale. When that size is received at either store -- the receiving scan is the trigger -- the AI texts her within the hour, holds the piece for forty-eight hours, and releases it to the floor if she does not come.

In the first season, 114 of those texts went out and 81 of the pieces were bought at full price. She was in Atlanta for four days in August, buying spring. While she walked the showrooms, the AI moved 31 pieces between the two stores, texted 19 waiting customers, took the first markdown on seven styles, and chased three ship windows, one of which turned up a label that had quietly pushed her delivery to October. She read all of it on the flight home, in a note shorter than this paragraph, with two things waiting for her: a second markdown on a knit dress and a cancellation she signed with her thumb on the tarmac.

The morning note is what she reads now, at seven, with coffee. What moved, what was reordered, what was marked down, and the one or two decisions that are hers. Full-price sell-through across both doors was 52 percent in the spring of 2025. The same season this year it was 63 percent, and the sale rack on Coleman Boulevard was half the length it used to be.

"I became a buyer because I loved the clothes," she said. "For thirteen years I was a spreadsheet. I'm a buyer again."