The merchants started asking how he did it

What tipped them over

Brett's job is the part nobody sees. By the time a category manager walks into an assortment review with a clean recommendation, Brett has already stitched together sales history, space and planogram inputs, supplier files, and store-cluster productivity from half a dozen sources that never agreed with each other. For years the work was solid and the trail behind it was invisible. When a merchant asked "why are we exiting this item in these clusters," the honest answer lived in Brett's head and a spreadsheet tab he hoped he could find again. That gap didn't hurt the numbers. It hurt the credibility -- his, and the recommendation's.

What tipped him was a review that went sideways. A merchant pushed on a productivity call, and Brett knew he was right but couldn't show the path fast enough. He wanted the reasoning to be as visible as the answer.

The first problem they solved

He started on the Explorer trial, no card, and didn't change how he worked. His SQL, his Excel models, his Power BI views stayed exactly where they were. What he added was a work-context layer over the top: he cataloged his assortment assets -- the feeds, the cluster tables, the productivity outputs -- capturing names, types, paths, and how they related, never the raw store data itself. Then he framed the real work as use cases: this markdown recommendation, who it serves, what it's worth, which merchant it lands with. It fit the flow. He was documenting decisions he was already making, not filing paperwork on the side.

"The recommendation didn't change. What changed is that anyone could see what it rested on."

Scaling from here

The difference showed up in the next category review. When a merchant questioned an exit call, Brett pulled up the lineage -- this recommendation traced back through these clusters, this space input, this supplier feed -- framed as a use case with the value and stakeholders attached. The merchant could see exactly what it rested on, and the conversation moved from "do I believe you" to "let's decide." A couple of merchants noticed and started asking how he'd made his reviews so easy to follow. That pull is what reached his manager. The merchants would never be MyDataWork users -- they don't need to be -- but the sharper reviews were coming from Brett's setup, so his manager moved him from the trial to a paid seat and pointed a few teammates at the same approach. Private-by-default kept his in-progress work his own; the leadership rollup showed only the aggregate architecture and value.

Toward agentic & generative AI

With his assortment estate cataloged and mapped, Brett started scoping an "assortment-review prep" helper in Agent Studio -- something to assemble the traceable context before each review instead of him rebuilding it by hand. Because the estate was AI-ready, he could give an approved agent read-only context through Agent Access: the metadata about his assets and their relationships, never the raw store-level data. The prep gets faster; the exposure stays at zero.