Loading summary
Transcript48 lines
- [00:00]
A
Foreign.
- [00:10]
B
Hello. Hello and welcome to another episode of the Digiday Podcast, a show for anyone who wants AI to do their grocery shopping. I'm Kamika McCoy, senior marketing reporter here at Digiday.
- [00:21]
A
And I'm Tim Peterson, executive editor, video and audio at Digital Media, and also someone who does want AI to do my grocery shopping. And I'm very close, Kimiko. I'm very close to actually having AI do my grocery shopping.
- [00:37]
B
We are touching on the topic of agentic commerce today. Hence Tim's push for grocery shopping and having AI do it. There's this big question of how far, realistically we can take agentic commerce and how far a AI bot can go in actually researching products and then buying them on your behalf. How far have you gotten, Tim?
- [00:59]
A
I. So this has been an entire project for me over the. Since probably the holidays. So the past, you know, we're coming up on five months where I have a whole agentic meal planning and grocery list system. So I have databases in notion that underpin all of this. And then I'm using cloud to power it. And so do you want me to walk you through?
- [01:29]
B
Please, I was just about to say, please walk us through this system from top to bottom.
- [01:34]
A
So here's what I have. I have created with Claude a Claude skill where I can take any recipe, whether it's a photo from my phone of a recipe in a physical cookbook or screenshots of recipes from, you know, we. I can Even just upload URLs to websites. I can submit that to Claude. Claude will then create a recipe page in my Notion database where it'll have the steps to the recipe, but also the ingredients. The ingredients are crucial because then I have a weekly meal planning system in notion where I'm able to pick recipes from my recipes database, have that populate in my weekly meal plan. Then what I'm able to do is have I have a notion formula that then takes the recipes for that weekly meal plan and creates a grocery list from that based on the ingredients in the recipes. That's just like a text list. So it's very basic. But then I'm able to take that text list, go back to Claude and say, hey, Claude, use the grocery list skill. Cloud will then take that full list, deduplicate things. So if I have potatoes for one recipe and potatoes for another recipe, it'll take into account the quantity consolidate that. I've also told Claude how my two regular grocery stores kind of organize their aisles so that Claude can group things into this is in the deli section, this is in the bread and cereal aisle. This is in everything. So then it just spits out a document that I'm able to copy and paste into Apple notes. So when me and my significant other go grocery shopping, bam, everything's already there. So at this point, the hang up for me is the actual grocery shopping.
- [03:38]
B
My brain is so frazzled and impressed right now because I go to the grocery store completely on vibes.
- [03:47]
A
It's just like, what am I in the mood for right now, basically?
- [03:50]
B
Basically. And you know, one meal will be made of three potatoes because I miscalculated how many I needed for the first recipe earlier that week. This is such an interesting system, but
- [04:01]
A
interesting is doing a lot of work there because I think you also mean to say, like, Tim, what the hell?
- [04:08]
B
It's. It's efficient. I'll give it that. But I am curious, mind you, I have a shit ton of questions now, but also, like, how efficient is this? Right? Because like, it's doing a lot of heavy, heavy lifting. But as we've learned through our conversations on the show reporting DPMs that we had, the buck stops somewhere. Where does it stop for you? Do you allow it to then go and purchase said groceries? I don't even know if it can on every platform.
- [04:37]
A
No, I could see. I have. I haven't looked into like finding a way to then automate this to plug into like instacart right to do. But that's not an AI issue for me. That's not a technology issue for me. I. That's a human issue. I do not trust another person to pick out my produce for me. I've seen the produce at my local grocery stores. Do you want me to tell you how many minutes I spend going through all of the Sugar Bee apples to find the ones that aren't bruised or don't have like some sort of mark where I say, not, not today.
- [05:11]
B
Well, how many minutes?
- [05:13]
A
At least five. I actually, you know, this morning I stopped at Sprouts to pick up some apples. And of the Sugar Bee apples, there were maybe 40 Sugar Bee apples there. Sugar Bee is my favorite type of apple. I like a sweet apple.
- [05:32]
B
I couldn't tell.
- [05:33]
A
So always kind of bruised or like have some sort of weird discoloration or like some sort of mark into them where it's just like, did a small bird take a bite of this? I touched at least 20 of the apples before I picked out my four that I decided to put in my grocery bag.
- [05:53]
B
Well, that taps into like questions around Agent like how agentic commerce can actually be here. Right. Because for you the buck stops at like purchasing. But it also seems like you're touching on this idea of like how the agent makes recommendations, how you process what choice you make versus how you know an LLM would process which apple to then pick, pay for and then put in your instacart. Right. Why don't you trust these agents? Has there been anything, you know, you trust it up to do the research and everything else. But like, why does the buck kind of stop there for you? Has it made bad decisions in the past?
- [06:32]
A
It hasn't. Well, I mean, yeah, they've all made bad decisions in the past. Even like this very intricate meal planning and grocery list system has been a very iterative process. I've needed to check things and correct things as we've gone when it comes to agents E commerce to like having an AI do the buying on my behalf. Where the buck stops with is with the buck. I don't want them to be one, I don't want them touching my credit card information. I'm not interested in that. Two, I'm not willing to trust that it's going to put the decimal point in the right place when it's making a purchase. Or that like in the case of the groceries, that it's not going to get a number wrong. Where Instead of getting four apples, I'm getting 40 apples.
- [07:28]
B
Yeah, yeah. No, that's absolutely fair. And I would imagine you're probably not the only person to feel this way which then kind of puts agentic commerce in a precarious predicament, if you would. Right. Because even when you looked at OpenAI and other players kind of rolling these features out, we've seen OpenAI roll back that instant checkout feature that they had. You know, in my reporting, that's one of the themes that I'm seeing kind of show up, is that when it comes to the human element, like that last 10%, a person is preferred to kind of close that gap. But I'm curious because like when you look at like the same pattern that happened in E commerce and social, we were able to close that gap, whereas two years before we weren't.
- [08:12]
A
Yeah, yeah. Although I mean, even there. So I've had. Sorry to everyone who has one of these. I've had an Alexa device for since 2016. And every now and then it'll pop up and be like, I'll ask, hey, what's the weather? It'll give me the weather. And it'll be like, oh, by the way, such and such product that you've bought before is on sale. Do you want me to buy it for you? No, I do not want Alexa buying things on my behalf just because I don't. I want to be able to like, physically see the purchase I'm making and click the button. Even though, like having Alexa do that again, sorry to ask one of these in their vicinity as they're listening to this, even though Alexa is just plugging into Amazon and so it making that purchase, clicking that buy button, it still is following all the guardrails that I would be using. Like, personally, If I'm on Amazon.com making a purchase. But even then there's just something about like not tangibly being the one to click buy where I haven't been able to get over that hump.
- [09:31]
B
Do you think even in your own experience, like, and also you're reporting and like, you know, being in this space, that we'll get to that point? Because again, there was at one point where people said, I'm not going to use like tap to pay on Apple. That's terrifying. I want a physical credit card to pull out or I'm not buying stuff off of social media. And then lo and behold, TikTok shop is exploding.
- [09:53]
A
I think so. I think like, I think if one of the times, the device that I will no longer name for the sake of our listeners, if one of those times it was just like, hey, this thing that you've bought before is on sale. And I was thinking, oh yeah, I need to like buy that.
- [10:14]
B
Yeah.
- [10:14]
A
I could see myself being like, all right, fine, let's roll the dice this time. And if it worked that time, now there's a paradigm shift for me where like, I'm more comfortable doing that kind of shopping. And so I think similarly with an AI agent, like, I'm not ruling out there being a time where again, I don't know that it's going to be grocery shopping if there's produce involved, but if it's just like buy this specific. Well, I was going to say can of soup, but even then I don't want a dented can of soup. I'm not trying to get, whatever, what is it? I'm not trying to do that. But if it's a low risk purchase and a low denomination purchase, I could see trying it because, I mean, I also haven't bought anything on TikTok shop, but a lot of people are buying things on TikTok shop. And for me it's just I don't have the familiarity with the platform yet to trust it. 20 some years ago people didn't have that familiarity with Amazon.com yeah and now
- [11:26]
B
even Amazon's got kind of some setups where once you reach a certain price point it'll trigger for that to be purchased on your behalf. So it's already got some agency systems in place. I could see that. I was just about to make the case for it. I could see this happening. Like if there's a shampoo that I like to use, right Any I, I essentially it's a subscription. Every time that shampoo goes on sale I would absolutely clear for Amazon to go and buy it. Amazon is a trusted resource, you know, that I regularly use. So no it wouldn't, that would scare me. Or like running shoes that need to be replaced regularly. Those are all things that I could absolutely see myself. Like Amazon's already got my credit card on file. You've already done the research, go ahead. And you know, anytime that that's on sale, make that purchase for me. That I could absolutely see myself doing.
- [12:12]
A
See I've gotten close, but not all the way there. So like I, I do triathlons and so I was looking at a high value purchase related to me and bikes recently to cycling these things called power meter meter pedals, the ones I was looking for. I checked the price history and like occasionally the price drops by like more than $150. So like a significant drop but it's a pretty irregular thing. I couldn't figure out a pattern when I was looking at the price history and when I clicked price history, Amazon's Rufus popped up which was Amazon's shopping assistant. As we were recording this this morning, Wednesday, May 13th Rufus RIP it's now just going to be the aforementioned AI voice assistant Alexa that's going to be taking the place of Rufus now on Amazon to. Com. Regardless, Amazon's AI assistant pops up, shows me the price history and then has the button there of like would you like to set an alert if this product drops under a certain threshold? Also would you like to have it where, when it drops below that threshold. Tim, we'll just buy it for you. Like don't even worry about it. Yeah, I set up to get the alert. I did not set up to do the buy just because again the trust factor, I don't, yeah, I just don't trust it. Clicking buy for me. And this is, this may just be a me thing. I get the impression a lot of people feel this way but we all, we've all for the Most part gotten over this in some way. Even my mom, who's very technology adverse, purchases things or has people purchase things on her behalf on Amazon.
- [14:04]
B
There is, I think, like the adoption curve or the gap is starting to get smaller. Right, between people like me and your mom that are now starting to adopt this behavior and be okay with it. But I think what that means in my reporting is that starts to encroach on who has traditionally had a stronghold in this space, which is retail media networks. They really built their business on people searching for products on their platform, on their sites. That's where their bread and butter is. And now if you've got more people that are starting their search on ChatGPT, Gemini, Claude and these other LLMs like your roundels and your Walmart connects, start to potentially take a hit from that traffic. If there's no longer eyeballs on their sites, because there are eyeballs now on the LLMs and transactions also potentially happening in these LLMs, what then is the case for a retail media network?
- [14:56]
A
Right. If I'm not going to Best Buy.com then Best Buy doesn't have inventory on BestBuy.com to sell to advertisers.
- [15:03]
B
Yes, precisely. And that puts them in a very. Another precarious predicament. When I talked to a couple of commerce media executives earlier a couple of weeks ago about this, this very same thing, the argument was, well, retail media networks have a lot of data and they can move off site, and that's something that they've already been doing. Well, now this puts it into question, like, do they become retail media networks in their true form or do they just become data brokers, kind of hawking their data to offside platforms? So I don't know. This becomes a competition between retail media networks and LLMs, or does it reveal
- [15:38]
A
that they were always kind of data brokers with some exceptions? Like, what is it? Planet Fitness has a retail media network or is it. I think it's Planet Fitness and just like who's going to planet fitness.com like to see, like, where are these ads running? Even like Kroger and Albertsons have retail media networks. Yeah, I've never gone to their websites. You know, maybe I am an outlier in that respect. And it turns out that, like, the majority of Americans are going to kroger.com or what have you. But I kind of get the sense that, like, a lot of these commerce media networks are really just data brokers, data providers.
- [16:24]
B
No, I think that's a fair assessment. And I think that's why the push to offsite in these partnerships with social media platforms? Ctv, fast channels and things like that becomes increasingly important because you can't necessarily make the case for inventory on your own site if people aren't coming there. And you've got to then partner with where eyeballs are, which is social media, which is ctv. A couple of these big players don't really have that problem. Amazon's got an entire ecosystem, right. With Amazon prime and things like that. Walmart's got Vizio now and a handful of others have struck some deals with like CTV providers and things like that. But the longer tail you go, the less space and, you know, levers that you have to spark those things. So I almost wonder if, like now they start to go spark those same partnerships with the LLMs to tap that inventory instead, to continue to make the case for themselves.
- [17:18]
A
Right. The challenge there is, does the value or I mean really the performance of retail media advertising hold up in these environments because, you know, ads on Amazon.com work. Yeah, because there's. It's more than just like the sponsored placement. You're able to see, like, the sponsored placement draws attention to it, but then you're able to see things like how many stars does it have? How many reviews does it have? Things that, like, where you're able to get over the hump of, oh, this is just getting pitched to me just because someone paid for it to be here. It didn't earn its place on the merits. It earned its place on the money with a chat GPT or a cloud. If Chat GPT. If I'm asking chat GPT, hey, I'm looking at. We'll stick with my cycling situation. If I'm asking chat GPT, what are the power meter meter pedals I should be looking at? And ChatGPT comes back with five different options that it's able to like, explain why the, you know, the pros and cons of them. But one of those options is a sponsored placement. I'm immediately discounting that placement because it's just like, well, you're not here necessarily because you are a good power meter pedal. You're here because you paid OpenAI to be here. You are the, like, kid whose parents donated the library at the university for them to get admitted. Yeah, I'm not trusting you right now.
- [18:57]
B
No, there's. And that is already out. Even aside from like sponsored recommendations. Right. There are already kind of questions about like, because there's no transparency into how recommendations are made in general, let alone a sponsored one, you know, these LLMs are crawling sites from influencer reviews, from blog posts and other third party things. And then I've been talking to brands that have like been absolutely burning the midnight oil, just cranking out content with the intention to get picked up, up by these LLMs. So there's like a big trust factor when it comes to having again, not even just a sponsored result in those LLMs, but like a result in general. So maybe there's still a case to be made for the retail media networks. They don't, they don't pass away today. They get to live to see another day.
- [19:45]
A
I don't know that they pass away, but I do wonder if the scope of those businesses needs to scale down, if they just become data brokers. Data brokers. Because again, if I'm not going to best buy.com to buy something and then Best Buy can't sell, doesn't have that same inventory to sell to me, maybe Best Buy can do a deal with OpenAI so that advertisers can say, okay, let me check a box to like use Best Buy inventory and I'll pay whatever, you know, percentage data fee in order to like have Best Buy and Best Buy gets revenue that way. Challenge there though becomes if I'm not clicking on those ads within ChatGPT because I don't trust them, then that negates the value for the advertiser. They don't necessarily want advertisers on ChatGPT to select best Buys data to like target their ads. So then retail media, which the whole holy grail of retail media is we can get people over that last hump to purchase. We're showing these ads when people are on e commerce sites when they're about to purchase anyways. And so we can just kind of like tip people over into purchasing your product. Now they become more on the brand side of things. It's more of the like mid to upper funnel. Let's get people aware of a product. Let's get your product in front of people who may have a propensity to purchase it at some point. That's already been the case in retail media. That's been part of the pitch. The challenge then becomes how do you prove that those ads led to the purchase? And how do you also like weigh? Okay, was it the ad that I saw on CTV three months ago or was it the ad I saw on Instagram two days ago?
- [21:48]
B
Yep.
- [21:48]
A
Or was it just like the answer I got inside Claude, which doesn't have ads? That is what led me to purchase?
- [21:58]
B
Absolutely. There Were already a lot of marketers that I've talked to and covering retail media that were kind of groaning and bellyaching about attribution issues even before, you know, chatgpt became a thing within the last couple of years. And I imagine that this only serves to compound said issues here. I don't know where retailers go from here. They're in a predicament where they've got to either just kind of own up to becoming data brokers here and then striking those deals with platforms like LLMs and things like that to continue. Because the writing is kind of on the wall here for them where more people are increasingly going to these alums for, for research and you know, if they start to put their credit cards down. If you start to put your credit card down, Tim, what then happens? You've got winners and losers in this situation.
- [22:51]
A
And I mean that's going to happen sooner than later I think, because there's. So Google has universal commerce protocol which is basically like a standard, standard protocol around agentic commerce. It's the thing that as it gets adopted and supported will make it so that you can be in ChatGPT or Claude and get a product like recommended to you and then be able to connect to your Amazon account to purchase that product within Claude without having to leave Claude while Amazon stays the merchant of record while you are like making that purchase via cloud but directly with Amazon to the point of like even at the checkout you can kind of get a pop up that looks just like the Amazon checkout experience to like for me make me feel like, okay, yeah I'm within cloud but like I am going like directly to Amazon here. In fact, Amazon has come on as one of the members of the tech council for the universal commerce protocol. Walmart is one of the retailers that was a launch partner with Google and was instrumental in developing the protocol. So major retailers are already getting behind this to enable a ChatGPT, a cloud, a Gemini to be used to buy products, just like web browsers are used to buy products. The question then, as you were saying, is what does that mean for retail media? Because the retailers, traditional retail businesses are going to be fine. This is how they bridge their way into agentic commerce.
- [24:46]
B
Yeah.
- [24:46]
A
Do their retail media businesses go with them?
- [24:49]
B
Yeah, I think especially like the Walmarts and the Targets and the Amazons of the world, their first movers, saw the writing on the wall, have already started plugging into these places to the point that you just made. I think now you're moving into the conversation about Kind of these longer tail players who don't necessarily have those same partnerships in place or even the leverage to put those same partnerships in place. And I think those may be the ones that are going to be caught on their back foot here.
- [25:18]
A
Although then there's like also the question of how important are these retail media businesses to these retailers in general. Because it seems like Amazon has obviously built a very big and successful business around retail media. And it seemed like every other retailer is just like, oh wait, we can just be like making money off of the shopper data that we have and it's like pretty high margin and fairly low lift. Cool, let's do that. Amazon can continue to do that and it's like agentic commerce landscape. Can the other retailers and do they like need to figure out how or can they just say well that was nice while it lasted.
- [26:02]
B
I almost feel like it's the latter here because right. Even when I talk to media buyers and commerce experts, they don't necessarily consider Walmart, Amazon, especially Walmart and Amazon to be retail media networks. They're more so just media networks because they've got an entire ecosystem built around them. They've got their incrementality, they've got the attribution Amazon can tell you if somebody saw an Amazon prime ad and then purchase. Right. They've got that closed loop system. Some of these longer tail players to your point that are kind of like tertiary don't necessarily have those same things. And a lot hinges on their ability to be retail media networks in their true sense of, of the word. But again you kind of run to this issue now like do you lean into becoming a data broker? And that's just it. Given that there's so many changes happening within LLMs and search and whatnot and agent to commerce.
- [26:54]
A
And I think the deciding factor is what you identified. That is already the issue when it comes to commerce media. It's advertisers questioning are these commerce media networks actually driving sales for me? Are they getting me in front of new customers or customers who wouldn't already be purchasing my product or are they just like trying to take credit for sales that are happening anyway? So I think even setting aside the agentic commerce of it all, that's going to be something that the commerce media networks need to prove out regardless. And if they prove that out, then that's that will help them to continue to exist and attract some form of revenue in an agentic commerce world.
- [27:43]
B
100%. Well, by going through this journey of your meal prepping, we have given retail media network something to think about and
- [27:52]
A
I appreciate you giving me the opportunity. I am sorry. Obsessed with my meal planning system and grocery list system, so I will talk about it with you or anyone else who wants to talk to me about it.
- [28:09]
B
Well, that brings us to the end of this episode of the Digiday Podcast. Thank you to everyone for listening. And please don't forget to share this episode with someone who you think would enjoy it. You can even rate us and leave us a comment on Apple Podcasts. We'll be back next week with another episode of the Digiday Podcast. Thank you so much for joining us.