Latest / Retail Media Breakfast Club / Rufus Remembers Me
Transcript
- Rufus with Persistent Shopper Memory === [00:00:00] Last Friday, I asked Amazon's Rufus a simple question. What kind of person am I in real life? Its response stopped me cold. Not because it was wrong, but because it was so [00:00:15] specifically right. Rufus knew that I play pickleball. I have kids. I own a cat. I crave expensive Japanese camping gear, and I maintain what is called a comprehensive [00:00:30] skincare routine. [00:00:32] ~I,~ [00:00:32] ~this new memory capability appears to have rolled out just last week. Don't be fooled. ~This new memory capability appears to have rolled out just last week, ~but ~don't be fooled. This is not just a basic feature update. Rufus is building persistent memories [00:00:45] of who I am as a shopper, transforming itself from a search tool into something closer to a relationship. [00:00:52] A relationship more like the ones we're starting to build with AI tools in our personal lives. [00:01:00] Let's jump in. [00:01:01] It's true that Amazon has offered personalization for a long time, but here's what's different between traditional Amazon personalization and Rufus with a memory. [00:01:15] First of all, is the recommendation trigger. With traditional personalization, it's session based, so you just search for running shoes. But with Rufus, with memory, there's a longitudinal understanding, [00:01:30] like you've been training for a marathon based on six months of purchases. [00:01:36] Another dimension is follow up logic. ~Traditional personalization would be something like. Okay. ~Traditional personalization would be purchase triggers. So you just bought diapers, do [00:01:45] you want to buy more diapers? Whereas ~Rufus with memory could have, whereas ~Rufus with memory would have persistent preferences. [00:01:51] So for example, last time you asked for plastic free packaging. [00:01:57] Another dimension recommendation method. [00:02:00] Traditional personalization would be something like ~collaborative filtering would be ~collaborative filtering something like customers also bought, but with. Memory, it would look more like cross category intelligence. So [00:02:15] since you like to buy organic baby food, here are organic snacks for adults. [00:02:22] And finally, context duration. Traditional personalization may be more short term in nature, ~whereas a AI with. ~[00:02:30] Whereas AI with long-term memory could hold more conversational continuity across weeks and months. [00:02:38] This memory update forces me to reconsider something. I've been arguing that retailers [00:02:45] launching standalone chat bots. Are wasting their time that the ~winners will be those who integrate with the personal AI agents. That ~winners will be those who integrate with the personal AI agents that consumers already use daily, like Chat, GPT or Gemini or Claude. [00:02:59] But [00:03:00] Rufus with memory may change that calculation. If Amazon can make Rufus persistently useful enough, knowing my kids' ages, my cooking habits, my brand loyalties, maybe they can pull off [00:03:15] what I thought impossible ~getting consumers to use. ~Which is getting consumers to use a retailer specific agent as their primary shopping companion. [00:03:24] Still most retailers launching me Too, chatbots without [00:03:30] Amazon's data advantage are kidding themselves. A retailer chat bot that forgets me between sessions. While chart, GBT remembers everything. Dead on arrival. The smart play remains building [00:03:45] bridges to wherever consumers are already living digitally, whether that's chat, GBT, Gemini, or maybe even Rufus. [00:03:54] [00:04:00] When your time, budget, and team are already maxed out, keeping up with retailer platforms and media requirements can feel like a never ending labyrinth. That's why even the best brands bring in an execution partner. [00:04:15] A Costa group makes retail media execution simpler. They stay ahead of every platform, update, tech shift, and media spec, so your campaigns actually show up and perform without draining your team. [00:04:29] [00:04:30] Let Acosta group handle the chaos behind the scenes so you can keep building the brand. Check out Acosta Group and tell them Retail Media Breakfast Club sent you. That's A-C-O-S-T [00:04:45] group. [00:04:48] Let's talk about what this means for consumer brands. [00:04:52] Number one is that switching costs just went up when Rufus knows, ~a shopper always buys. When Rufus knows that ~ a shopper always buys [00:05:00] Tide. Getting them to try Persil requires more than winning a keyword bid. You're now fighting an encoded preference that strengthens with each purchase, and number two, quality becomes [00:05:15] defensible. [00:05:15] Brands with high satisfaction scores and low return rates, ~get a ~get algorithmically entrenched. ~Your post-purchase experience.~ [00:05:23] Your post-purchase experience, the packaging product quality customer service now directly impacts [00:05:30] future discoverability. You've probably seen the little tag on some PDPs that say customers usually keep this item. We could see similar personalized messages based on shared interests like. [00:05:45] Customers who love the camping brand Snow Peak also tend to shop this brand, [00:05:50] This is overall great news for established CPG brands who are trying to defend themselves against private label and scrappy unbranded [00:06:00] sellers. When~ When AI consistently. ~When AI consistently steers shoppers toward what worked for you before displacing trusted brands becomes harder [00:06:10] and what retailers should pay attention to. Other retailers face a [00:06:15] genuine challenge here. ~Amazon's advantage isn't just having an AI assistant, it's having three decades of behavioral data.~ [00:06:16] ~Amazon's advantage isn't just having an AI assistant, it's Amazon's advantage isn't just having ~Amazon's advantage isn't just having an AI assistant, it's having two decades of behavioral data to train it on. And the breadth of category insight that's fairly [00:06:30] unique to Amazon. While Walmart and other retailers are building capable AI tools, they're likely working with a fraction of the historical online purchase data. [00:06:42] But this isn't insurmountable. [00:06:45] Smart retailers have options, one pooling data through partnerships. I've started talking more about retail media alliances and federations, and these types of relationships can ~help aggregate insights across retailers and these types of programs could ~help aggregate [00:07:00] insights and create a more powerful data set. [00:07:03] Number two, focus on specialized categories. Retailers who have really deep expertise in specific verticals could outweigh broad data. If I'm shopping in the [00:07:15] beauty category, for example, that's an area where there could be really strong memory utilization there. And number three, leverage physical store advantages in-store behavior. [00:07:29] Data [00:07:30] remains a unique asset. This is an area where some retailers will have a significant advantage over Amazon. [00:07:38] This development challenges my view that retailer specific agents are doomed and that [00:07:45] only those who integrate with consumers preferred AI assistance will survive. As I've argued before, most retailers launching standalone chat bots are building digital ghost towns. But if Amazon can make Rufus [00:08:00] persistently valuable enough, [00:08:01] they might just prove me wrong.