Latest / Retail Media Breakfast Club / 'Nobody Wakes Up Wanting To Chat With A Retailer'
Transcript
- 'Nobody Wakes Up Wanting To Chat With A Retailer' === [00:00:00] Kiri Masters: What is the job to be done of a McDonald's milkshake? It wasn't, as business guru Clayton Christensen learned doing field research in 2005, to just be a [00:00:15] sweet treat. The milkshake, as it turned out, was being hired to keep morning commuters full until lunch. This framework, the jobs to be done framework, has persisted over so many years [00:00:30] because many businesses fall into a common trap, improving a product without asking what it was hired for [00:00:38] Today, retailers of any size are now shipping one, upgrading one, or [00:00:45] introducing ads to one. They all kind of look the same and act the same. [00:00:51] They find products for you. They compare options. They suggest recipes. Even the names for these little digital [00:01:00] servants are similar in their perkiness. Olive, Sparky, Milo. [00:01:06] And the one-noteness of them all makes me think that we might be losing touch from the thing that these consumers [00:01:15] are hiring AI chat assistants to do. Let's jump in [00:01:20] [00:01:21] Kiri Masters: A written interview in the fantastic newsletter, The Aisle, by tech journalist Jason Del Rey got me thinking about this. [00:01:30] Last week, Jason published an interview with Mukesh Jain [00:01:34] And Jain spent nearly five years at Walmart in product leadership roles across online grocery and search and personalization. He then [00:01:45] spent eight years at Amazon leading the team behind the product graph, which is the long slog of turning unstructured mess of product descriptions and box photos into attributes that a machine can actually query, and later held a similar [00:02:00] role on the Rufus team. [00:02:02] So he is particularly well-placed to talk about AI chat assistance, where they're at now, and where they're going. But Jain had an interesting framework that I really latched onto, which is [00:02:15] splitting shopping into three jobs. It's not just one [00:02:20] The first is the repeat purchase, and this is one where it's really not suited to a chatbot style of [00:02:30] interaction. You know exactly what you want. You want a bottle of Heinz ketchup in an eight-ounce size. You type it, you see it, you recognize it, you check out. A conversation here is completely unnecessary.[00:02:45] [00:02:45] Another type of shopping behavior is research. You want to read reviews. You want to compare different options. You wanna understand what options are available. That would ordinarily take a lot of work, but it can be made much more efficient with an [00:03:00] AI chatbot. Great use case. And finally, complex constraints. [00:03:05] You have a eight-year-old nephew you need to buy a gift for. The nephew is into space and Lego and probably owns all the [00:03:15] space Lego already, and you're looking for something under $100. Go. Now, that kind of demand prior to an AI chat assistant was completely invisible to the retailer. It wasn't how people were able [00:03:30] to search and filter in the past [00:03:32] and so Jason, at the end of this interview with Mukesh Jain, asked him, "What do shoppers really want from these tools?" And Jain actually answered by naming what they don't want. He said, "Nobody wakes [00:03:45] up wanting to chat with a retailer [00:03:47] If the answer could have taken one message and it took five, that is not good engagement. What they want, he said, is less work between I need something and it's on the way [00:03:59] [00:04:00] And in this space, there's been quite fair pushback to all these impressive numbers about Rufus and Sparky in particular, and how they're driving all this incremental revenue. And the pushback is that all of this adoption that we're hearing [00:04:15] is from consumers who are already maybe the most engaged ones, the most likely to buy already. [00:04:21] And Mukesh Shah had a sound test to kind of run that potential reality through. He said, "I'd judge [00:04:30] any of these products on how the repeat cohorts behave at month three and month six, not on month one, because every consumer's first query into an AI chat assistant looks like a keyword [00:04:45] search with a few extra words, because we're not used to this way of engaging with an AI chat assistant. [00:04:52] We're all carrying thirty years of habit. The question is whether those queries get longer and more complex [00:05:00] after that." poor audience targeting is frustrating, but for retail media teams it can be costly too. With Growth Loop's [00:05:15] composable commerce media solution, you can turn your first-party data into hundreds of high-value audience segments and launch campaigns faster. After partnering with Growth Loop, instant [00:05:30] commerce pioneer Gopuff scaled from a hundred syndicated audience segments to more than six hundred, and now it takes less than forty-eight hours to turn around a custom segment for one of their brand [00:05:45] partners. [00:05:45] Learn more about how Growth Loop is powering Gopuff's best-in-class retail media operations at go.growthloop.com/breakfast. That is [00:06:00] go.growthloop.com/breakfast Next is the harder question: Who's going to win here? Is it gonna be the horizontal assistants, the ChatGPTs, the Geminis of the world, [00:06:15] or will the retailers be able to offer something unique that brings consumers back? [00:06:21] And this is where Mukesh Shah is even-handed. He says, "The general assistants have the habit of hundreds of millions of people [00:06:30] already visiting them every day, integrating them into their lives, and they can see across different retailers, and they can compare products from anywhere and recommend across the whole market, which an agent embedded in only one store [00:06:45] structurally cannot do. [00:06:47] And retailers, for their part, they own the transaction, the data and the wallet, so they have significant assets of their own." So his verdict is, "I think they can coexist the way search [00:07:00] and retail coexisted. You have discovery in one place and the transaction in another. And so the open question is who ends up owning the start of the shopping journey?" [00:07:11] And that fight is still very much on. [00:07:15] Now, there was some interesting research that I saw a couple of months ago, initially from Kantar, and Kantar is... This was the first wave of its AI-enabled commerce [00:07:30] pulse, which they field from two thousand US shoppers. This first wave was conducted in April of this year, and they found that fifty-eight percent of consumers want AI tools that work across multiple retailers, [00:07:45] and forty-one percent of AI shoppers- already use AI to compare prices across retailers [00:07:51] Kantar's read is that retailer-specific assistants produce stronger satisfaction among the people that use [00:08:00] them, but platforms will likely have an advantage since product and price comparison lead utility [00:08:07] So Mukesh Jain's POV is that owning the checkout, the purchase history, and the card already on [00:08:15] file, that evens out the fight between the horizontal and the vertical agents. But Kantar's data suggests that comparison is the thing that shoppers rate the most useful, and it's the one thing that a single store can't do. [00:08:28] So this is the part of the episode where I [00:08:30] tell you that I was wrong. I've been arguing for a while that retailers building standalone chatbots were building digital ghost towns, and that the winners would be the ones plugging into wherever consumers already [00:08:45] spend their time. And then last August, Rufus rolled out persistent memory, and this meant that when I asked it what kind of person I was in real life, for the first time, it admitted [00:09:00] that it knew I play pickleball, I own a cat, I like expensive Japanese camping gear And once I saw that, I thought Amazon has totally proven me wrong because I thought the moat was memory. [00:09:14] [00:09:15] I thought that using an AI chat assistant that remembered my preferences was going to be a critical aspect of adoption of these kinds of tools. That retailers might lose because they forget you [00:09:30] between sessions while ChatGPT remembers everything. And this research from Kantar shows that memory isn't the deciding variable, breadth is. [00:09:41] And Mukesh Jain, who built this from the inside [00:09:45] at Rufus, says that nobody has really delivered that memory part yet anyway. That going from a search box to true assistant is the hardest part, and nobody has fully delivered on it yet. [00:09:58] So wrapping up here, [00:10:00] the job to be done for a retailer's AI assistant is not a single job to be done at all, and that is why the current chatbot modality isn't the final boss version of [00:10:15] AI-enabled shopping. The fact that even Amazon hasn't cracked the code yet should be encouragement enough to retailers to keep iterating on the chatbot form, function, and reason for being [00:10:28]