Latest / Retail Media Breakfast Club / AI Agents Hate Ads Today — But Will That Last?
Transcript
- [Tues 9/2] AI Agents De-Prioritize Products That Are Sponsored - But Don't Panic Yet === [00:00:00] Last week I shared some new research on the preferences that AI agents appear to have in online shopping environments. A key finding was that AI shopping agents actively penalize products with sponsored [00:00:15] tags, reducing the likelihood of them being selected by eight to 14%. [00:00:21] Now, I've written before about how ag agentic commerce could threaten retail media's current mechanisms, both onsite and offsite revenue [00:00:30] streams. But here's my contrarian. Take. This ad penalty that we are seeing with AI agents is temporary and retailers still have multiple paths forward. Yes, today's AI models [00:00:45] may treat sponsored products like a credibility tax, but viewing this as permanent misunderstands both how AI evolves and how markets adapt. [00:00:55] Let's jump in [00:00:56] [00:00:58] Why the current ad penalty [00:01:00] is throwing the baby out with the bath water. Current AI models, inherit their ad skepticism. From us, the humans, GPT-4, Claude, and Gemini, we're all trained on human generated content. [00:01:15] Where we explicitly express distrust of advertising, even when our actual behavior tells a different story. [00:01:21] AI models have learned our stated biases. [00:01:25] But I really don't think this is gonna last forever. Consumer facing AI [00:01:30] platforms will evolve beyond simple ad avoidance. The innovation will come from the AI agents themselves, not the retailers necessarily, as these agents compete for users. They'll need to deliver genuine value, [00:01:45] finding the best products at the best prices, and sometimes a sponsored product genuinely is the best option. [00:01:53] And if you have an agent working for you that categorically dismisses sponsored items just [00:02:00] because. Their being advertised might miss a really good deal for you, or it might miss your preferred branded product. That's going to frustrate users who discover that their favorite product or a really good product was [00:02:15] avoided just because there was an ad running on it. [00:02:17] Just as these models have been trained on one set of assumptions, future updates can incorporate new ones. We see new models coming out all the time that act a little bit differently. This bias exists [00:02:30] today, but I wouldn't expect it to persist forever. [00:02:33] Here's another thought on retail media. My friend Scott Wino has shared on his podcast and newsletter Retail Gentech a pretty interesting vision where [00:02:45] retailers when working with AI agents could actually bid for shopping baskets in real time, and that this would be a way for AI agents and retailers to kind of work together and acquire customer orders. [00:02:58] So here's how it works. [00:03:00] A consumer tells their agent that they need taco ingredients for six people tonight. The agent identifies the basket, what they need. They need ground beef, do tears, cheese, lettuce, tomato, sour cream, hot sauce, and simultaneously [00:03:15] they send that request out to multiple retailers. [00:03:19] Then the bidding begins. Walmart offers a total price and $3 off the total basket. If the agent goes with Walmart, target counters with [00:03:30] $2 off, plus free same day delivery. Kroger bids $4 off for curbside pickup. These retailers. They can compete on price, but they may also compete on total values delivery, curbside pickup [00:03:45] samples, et cetera. [00:03:46] And here's where I'll build on Scott's vision. Retailers could provide brands an opportunity to amplify those bids or kick in something additional. Old El Paso spots tortillas in the basket and adds [00:04:00] a 50% additional discount on top of Walmart's bid if their brand gets selected or a new hot sauce brand trying to enter the market goes a bit further. [00:04:09] They might contribute another dollar to any retailer's bid that wins a trial purchase [00:04:15] of their new product. Now this mirrors how collaborative bidding already works in social advertising through platforms like metas, collaborative Ads and third party solution providers like symbiosis, brands and retailers pool [00:04:30] resources together to win the customer, sharing both the costs and the benefits. [00:04:35] But in this case, in the AgTech commerce case, instead of bidding for attention, they're bidding for confirmed purchases. [00:04:42] ~Now,~ [00:04:42] ~how does the ai,~ [00:04:42] ~so, so coming back to this. ~So closing the loop on this scenario, [00:04:45] the AI agent evaluates total consumer value, which is the price after all the discounts, the delivery, convenience, product quality, satisfaction scores, et cetera, all of their decision making criteria, which may be customized by the human that they're working [00:05:00] for. [00:05:00] The agent finally selects the winning combination. The AI platform takes a cut of the incentives. That's how they get paid, provided by the retailer and brand, which creates a clean monetization model. Without [00:05:15] compromising trust, the consumer knows that the baskets have been shopped around and bid on. [00:05:21] It's sort of like an affiliate model of the past where I buy through someone's code or a special link and it doesn't cost me any extra [00:05:30] money. It might, in fact, get a discount. I know that, you know, Joe Blogs got a affiliate commission from that. But I didn't pay any more for the product. I might have actually got a better deal, so I'm okay with it. [00:05:41] As long as it's transparent, [00:05:42] everybody can win. [00:05:45] Did you know that 60% of shoppers abandoned carts when product info is wrong? Before you ramp up your ad spend, fix the [00:06:00] shelf Acosta group's. Award-winning connected commerce crew is trusted by brands like Coca-Cola and Sanofi to deliver critical content updates. [00:06:11] Acosta Group handles it all. Then layers on media [00:06:15] buying. Learn about Acosta group's connected commerce capability at Acosta Group and tell them Retail Media Breakfast Club sent you. That's A-C-O-S-T-A group [00:06:30] Here are some other ways that retailers could monetize the agentic shopping future. And this comes from that research from, the group of researchers from a private, ai, software company, as well as [00:06:45] academics at Columbia and Yale. [00:06:47] Value added services was something that they suggested in the study. The study found that AI optimized product descriptions delivered double digit market share gains in 25% of cases tested. [00:07:00] Now, there is a little bit of nuance here, but if you go back and listen to my episode from last week, AI generated content ~didn't actually improve product win, uh, product.~ [00:07:06] ~It ~didn't actually improve product selection in all cases and in all categories. So just, you know, a little side note there before you go and [00:07:15] update absolutely everything with AI generated content, look into this because there's definitely some nuance here, but one thing that the researchers suggest is that retailers could offer, product optimization as a [00:07:30] service [00:07:30] uh, that might be a way to make money. I'm a little skeptical of that, uh, suggestion because I see a lot of brands really taking the reins here and going out and proactively trying to figure out product [00:07:45] content and syndication across retailers. They're really the ones moving very quickly here. [00:07:51] In some cases it could make sense, but I think there's a big question of who's gonna move first on this. Is it gonna be the retailers or the [00:08:00] brands? Another non-advertising monetization opportunity for retailers is building out better product review ecosystems. There are such limited options available here [00:08:15] relative to the demand, in my opinion. [00:08:17] Product reviews, as I spoke about last week, analyzing this research are so important to these AI agents. They are place a, a great weight on product review, [00:08:30] quantity and, the rating of reviews. As we know with Amazon it's a very delicate system where as a brand, you cannot incentivize any reviews. [00:08:41] It's a very sensitive topic. For good reason. People don't [00:08:45] wanna feel like reviews are being generated, unethically. [00:08:49] Unfortunately, with such strict ways of acquiring reviews, it means that getting product reviews ethically and at the same [00:09:00] time, getting more of them, especially for new products and new brands, is very difficult. I think more retailers have an opportunity to start building up a strong ecosystem of product reviews, and this is a great way to compete with Amazon if you [00:09:15] have really robust, healthy reviews. [00:09:18] And look at Amazon Vine is a program where it's self-funded, it's generative. Amazon charges for access to this program. Brands are more than willing to pay to participate [00:09:30] because it gives them a bump in reviews. Everything is above board. The disclosures are there. [00:09:36] So this kind of seems like a big opportunity for retailers to create, ethical programs for brands looking [00:09:45] to build more reviews and incentivize customers to leave honest reviews as well. [00:09:50] So bringing this all into land here, I think my biggest takeaway is don't nix the sponsored product ads in an attempt to get ahead of the [00:10:00] curve. I think we're really in a transition period here, not necessarily witnessing extinction [00:10:05] today's sponsored products might evolve into something else over time, but [00:10:10] Let's be honest, adoption of a gentech shopping is still really low. [00:10:15] We will have time to figure it all out. [00:10:16] [00:10:18]