Everything below comes out of the conversation with Jill Scott about Jason's questions. The through-line she gave us is simple: the tool's job is to get a rep from a cloth number to a complete outfit the client can see — because clients who buy combinations keep wearing them, and clients who buy an orphan jacket don't.
They buy a really beautiful sport coat… and they're like, I've got tan pants at home. Then they go home and it's kind of a weird shade of tan… and now they become less confident in what to do with their new sport coat. And so then it sits in their closet. Jill Scott — the company's pants lag, described as a person
Every feature we build should be a step on this line or make a step disappear. Today the line breaks at step two — which is exactly where Jill stalls, and where I stalled with Brian Daly the same morning.
I'm like, oh, this would be really cute to show this client. And then I go in and there's no mannequin yet… I don't have everything I need currently. Jill Scott — why step two is first on the list
Sequenced by what removes a stopping point, not by what's most interesting to build.
One tap on any cloth produces a usable mannequin, choosing the best source automatically. Plus: photograph the actual cloth in the folder and use that, because a lot of our stock images are poor.
Why first: it is the literal point where the tool stops being useful. Every step below depends on it existing.
Once the mannequin exists, turning it into a real-setting image should be the obvious next tap rather than a separate expedition.
Why: the render is the thing that sells, and the mannequin is only the means to it. Jill's five-panel capsules are this, done by hand, at roughly 30 minutes each.
A consistently formatted sheet the rep sends on delivery: the outfit, the pieces, how to wear it, what else it goes with. Generated on a button, not assembled by hand.
Why it matters more than it sounds: Jill already does this manually for every AI capsule she sells, and it is also the answer to teaching assistants — they don't need to learn styling if the tool formats the output.
Show the same jacket with genuinely different tops: button-up silk, tie-front blouse, knit, and so on. Plus: when a new women's model launches without photography, generate one immediately.
Why: named as the single biggest gap and the area of greatest opportunity. Women hit two objections — will I like the fabric made up, and will it look right in this style — and this answers both at once.
Pick the cloth; the tool proposes what completes it. Trousers shipped today. Next: shirts and knits, women's models, and priors drawn from what actually sells together.
Why: Sergio's expert mode told reps to stop worrying about the details and go sell. A 23-year-old can sell a sport coat but has no idea what to show if the client wants to have fun with the trouser.
A client-facing view: browse by the outfit pictured, by price band, by fabric, in-stock only. A client file whose swatch card is already populated, with a wish list, and this appointment versus last.
Why: today the app is for making things; selling happens from the camera roll. The prep is real work and it's done twice. "I've got too many things in too many different places."
Kept as an internal reference and an input to the AI, not as a client-facing artifact.
Why smaller than planned: Jill would never show one to a client and doesn't need to watch one assemble — only "if I was trying something risky." So they don't need to be pattern-accurate, which removes most of the cost.
| Capability | State | What it does for a rep |
|---|---|---|
| Cloth catalogue & search | Live | 42,000+ cloths, live stock, natural-language search |
| Sales attribution | Live | Answers "what is the tool producing" with real numbers |
| H&S best sellers | Live | Company and personal, by pieces, any timeframe |
| Trouser pairings | Live | Suggests what completes a sport coat |
| Ladies fit advisor | Live | Reads a measurement page, recommends the model |
| Lifestyle renders | Live | Works well — when a mannequin already exists |
| Mannequin creation | Manual | The blocker above |
| Reference sheet | By hand | Jill builds each one herself |
Open it to any top-25 producer who wants in. More real usage is the only way to find what's actually broken, and it turns "one rep's opinion" into evidence.
Access to real cloth, order and manufacturing data instead of derived data. Every recommendation gets better the moment this lands, and it isn't something that can be worked around.
Development has been funded personally. If it's proving useful, the people who built it should be made whole — and a decision about how it's supported going forward.