On 16 September we put around twenty senior retail leaders around one table in New York. No panel, no deck, no pitch. Just dinner and a question: what are you actually working on right now?
Five topics came up without prompting, and they came up in nearly every conversation. Taken together they are a fair read on retail AI priorities going into 2027. What follows is what the room said, aggregated and unattributed. The interesting part is not the list. It is what connects it.
1. Item discovery
The framing was consistent: how do we get shoppers to see more of the catalog, and see the right parts of it?
Three signals came up as the basis for relevance. What she is looking at right now. What she has shown interest in before. And the one that got the most attention: what she already owns or has bought.
That third signal is the one most retailers are sitting on and not using. Purchase history is treated as a record of completed transactions rather than as an input to the next recommendation. A shopper who bought a blazer in March is a different shopper from one who did not, and most sites do not behave that way.
2. Inspiration
The second theme was about bringing merchandising to life.
Every retailer in the room employs people with real styling expertise. That expertise shows up in campaign imagery, lookbooks and editorial. It mostly does not show up on the product page, which is where the shopper actually decides.
The question several people asked in different words: how do we help her understand style, build a look, and feel confident she will look good in it? That is a merchandising problem being solved with photography, and photography does not scale across a catalog of tens of thousands of items.
3. The digital closet
This one had the widest range of ambition of any topic, and the most energy behind it.
At the modest end, it is simply making the closet a more useful part of the web experience rather than a saved-items list nobody returns to. At the ambitious end, people described something closer to the wardrobe in Clueless: a closet that knows what you own and can build outfits from it.
What was striking is that nobody treated this as a novelty. Several leaders described it as a feature they expect to become a significant part of their site’s value over the next few years. It is worth reading alongside what Shoppable Closet does today.
4. Optimization, under real financial pressure
The fourth topic was a catch-all, and it was the one with the most urgency behind it.
Everyone is focused on conversion rate and units per transaction. That is not new. What has changed is why. Acquisition costs are higher, operating costs are higher, and the consumer is stretched. The room described genuine pressure to produce financial results, not engagement metrics.
That pressure changes which tools survive procurement. A feature that lifts engagement without touching conversion or basket size is a harder sell this year than it was two years ago. We wrote about the conversion side of that in how to increase fashion ecommerce conversion rate.
5. Fit and visualization are two different problems
The fifth topic arrived as one subject and should be treated as two.
The first is fit confirmation. Will this item fit my body, given my shape and size? That is a measurement question, and the goal is confidence that what arrives will fit.
The second is visualization, or what we call see on me. How does this look on someone like me, and does it suit me? That is a style question, and the goal is confidence that she will like how she looks.
They overlap naturally, which is why they get discussed as one thing. But they answer different doubts, they are solved with different technology, and a shopper can be sure of one and unsure of the other. Conflating them makes it harder to tell what a given product actually does, and harder for a retailer to work out which problem they are buying a solution to. Our piece on styling, not sizing covers why the style half is the one the industry has left alone.
What connects these retail AI priorities
Read the list again and something becomes obvious. In most organizations, those five topics are five different purchases.
A discovery vendor. A recommendation engine for inspiration. A closet feature built in house or bolted on. A conversion optimization tool. A fit or try-on provider. Five contracts, five integrations, five sets of data that do not talk to each other, and five separate cases to make to finance.
But the room was not describing five problems. It was describing one, from five angles. Every topic reduces to the same question: how do we help her decide, with confidence, using what we already have in the catalog and what we already know about her?
That is a single job. It is solved better by one layer that sees the whole session than by five tools that each see a fragment of it. The fragmentation is not a technology problem. It is an artifact of how the category was sold.
What we took away
Two things stayed with us after the room cleared.
The first is that the digital closet has moved from interesting to expected. A year ago it was a differentiator worth discussing. Now it is on roadmaps, and the leaders describing it were talking about scale and site value, not experimentation.
The second is that the financial pressure is real and it is sharpening how retailers evaluate. Nobody at that table was shopping for something impressive. They were looking for the shortest path between a shopper’s hesitation and a completed, kept order.
Thank you to everyone who joined us in New York. If you would like to see what this looks like on your own catalog, book a demo, or read the Revolve results.