Shoptalk is a conference at the intersection of online and IRL. At its heart, its is about predicting (and debating) customer behavior. Everyone from CPG brands, to app makers, to Shopify itself is placing their bets via the technology that they’re adopting and partnerships that they’re nurturing, so it’s only fitting that the conference take place in the home of gambling and brotherly love: Las Vegas. But as we observed, even that last part has changed in 2026, as most Americans can now gamble from the comfort of their homes. But I digress. This is a Shoptalk post!
Take #1: A.I. was the biggest disrupting force.
This should come as no surprise if you’ve been awake for the last year or so. As soon as you walked in the door to the conference, Shopify was showcasing a magic mirror with a sense of humor that would “Roast or Toast” you based on what you were wearing via text appearing on the glass. After correctly identifying and describing your attire then insulting or praising your fashion and even life choices, it would recommend an accessory then print out a redemption receipt for the item. Matthew chose “Roast” and I chose “Toast” (after seeing how scathing this mirror could be), and while the LLM delivered funny zingers, it ultimately recommended the same socks for both of us. That was likely due to limitations in promo material. It was an entertaining demo, if not practical for scalable real world application. No one could really give us a serious retail application for the mirror that would put a Nordstrom associate in danger, for example. After all, if you’ve left the house to go shopping, you’d probably also be fine with, if not craving, a little human interaction. Plus, it’s already been well documented that it’s always much easier to tap into existing behaviors than to create new ones. You’d maybe use the mirror once as a novelty then never again. And that’s what was at the heart of Shoptalk: figuring out what users actually want to use repeatedly.
The mirror is just the first example, of course. A.I. is slowly proliferating to every part of the customer journey and funnel. Compelling uses that we didn’t expect included:
- Apps where users can generate images of clothing on themselves
- On-site agents that help customers not only decide between products based on their need, but assist them in building baskets
- AI-personalized checkout experiences based on customer data
These are just some of the dizzying ways A.I. was getting more involved further down the funnel.
Take #2: Agentic commerce is happening — but be wary of eye-popping numbers.
We attended a debate on Tuesday that has stuck with me this past week: AI Agents Will / Will Not Transform Retail. At first glance, this seems like an extreme take for the “against” side. Anyone who thinks A.I. won’t transform retail must be insane or a Luddite, right?
Well, not so fast. Andrew Lipsman is an independent analyst and consultant who came to the talk ready to throw down. When the “yays” in the debate threw out some eye popping numbers from a McKinsey report regarding agentic shopping and purchase (whereby a user completes the entire purchase journey on an LLM like ChatGP — including paying for the item), Lipsman had receipts. He pointed out that the numbers actually revealed that less than 2% of users in that particular study of over 4,000 people had self reported using LLMs to purchase something. He conceded that a growing number of users were arriving at sites lower down the funnel, but the traffic apocalypse being threatened by OpenAI and others beginning in Q4 of last year has not materialized. In fact, many companies who bet big on that were left holding the bag. They’re now stuck in long term contracts for a channel that does not convert. At least not yet. Things sure are changing fast, and it may not be long before users are offloading the type of shopping that they don’t enjoy to LLMs. The main hurdles right now seems to be trust and accessibility. One can easily imagine this tech becoming normalized as people get more and more familiar and comfortable with it. It’s clear that we’re not there yet, and betting too big right now feels very risky.
According to a poll conducted by Quinnipiac this week, the headline reads: “AI Use Increases While Views On It Sour.” Nearly 1,400 Americans were surveyed, and 76% they rarely or sometimes trust A.I. while only 21% said they trust it all or most of the time. Growing concern about A.I. was seen across all generations, with 51% of respondents believing that it will do more harm than good. That’s potentially not great if you’re betting on folks letting A.I. into their lives more intimately… at least in the near term.
Take #3: The brand website is not dead: in fact, it might be more relevant than ever.
In some of his closing day remarks, I was surprised to hear one of the hosts of Shoptalk remark that that there is data which suggests that time on site is actually trending up for ecommerce stores. This is perhaps contrary to logic and has implications on how brands should monitor and evolve their site content. If more users continue to come from LLMs rather than say traditional search or social media, the context shifts — potentially toward an even higher purchase intent. Meanwhile, Amazon growth in several Subscribe & Save categories have been flat or even negative over the past year plus. This could spell opportunity for brands to capture that revenue on their own sites. Time on site is certainly a KPI to pay attention to in 2026.
Another interesting nugget came from Emarketer: according to Amazon and Walmart, on-site chat agents are increasing AOV. Users seem to trust on-site agents more than off-site LLMs (like Claude or ChatGPT) further down the funnel. There were multiple vendors at Shoptalk offering these types of solutions for Shopify, including Zowie. I’ll admit I was a bit skeptical at first, but the numbers don’t lie.
Take #4: Brands of all sizes need to be where the conversations are.
The recurring theme that we heard from legacy brands and startups alike was that when conversions and metrics were flagging, finding their customers on the platforms they were actually using and tapping into cultural moments was vital.
For example, Coach found that it was not connecting at all with Gen Z. After lots of research (in the form of just hanging out with the youth) and trial, they realized that not being on Amazon was hurting them — the main reason being because that’s where customers do a ton of comparison shopping. So regardless of whether or not their sales actually occurred on Amazon, just being there noticeably boosted their consideration.
Likewise, Clinque found itself being buried in the noise of startup brands that were louder and flashier despite the new players having none of the quality, science, and cache of the legacy brand. Instead of trying to butt into conversations, they asked everyday loyal customers what their experience was with the brand. Whether it was year 1, year 5, or year 20 with Clinique, the message was clear: this stuff works! And that became the campaign. Why try and say anything when customers can say it for you? Compiling those testimonials into their marketing north star across all their channels (and television) added up to their most successful campaign in over a decade.
The big takeaway was that consideration habits are changing. Some brands have found that it takes upwards of 19 touchpoints online before someone tries your product. It’s really hard to show someone that many ads or pieces of content, so if you’re not in conversations and being shared, you’re going to get buried.
That’s the report in a nutshell from Shoptalk! Overall, it was an inspiring and revealing conference. With so much uncertainty regarding A.I. and tech, meeting and communing with other folks in this crazy space that’s changing as fast as we can adapt to it was a breath of fresh air.
