Reimagining Retail: How AI can help with hyperpersonalization and promising examples of AI in retail

On today's podcast episode, in our "Retail Me This, Retail Me That" segment, we discuss what AI can do for retail, who the biggest players are, and if hyperpersonalization is something people even want. Then, for "Pop-Up Rankings," we rank four promising examples of AI in retail. Join our analyst Sara Lebow as she hosts vice president of content Suzy Davidkhanian and analyst Carina Perkins.

Subscribe to the “Behind the Numbers: Reimagining Retail” podcast on Apple Podcasts, Spotify, Pandora, Stitcher, Youtube, Podbean, or wherever you listen to podcasts. Follow us on Instagram.

Episode Transcript:

Sara Lebow:

Hello, listeners. Today is Wednesday, February 21st. Welcome to Behind the Numbers: Reimagining Retail, an e-marketer podcast. This is the show where we talk about how retail collides with every part of our lives. I am your host, Sara Lebow. Today's episode topic is AI and hyper-personalization in retail. Before we get into that, let's meet today's guests. Joining me for today's episode, we have senior analyst, Carina Perkins. Hey, Carina.

Carina Perkins:

Hi, Sara. Thanks for having me back.

Sara Lebow:

Thanks for joining us on the later side in the UK. And in New York, with me, is VP of content, Suzy Davidkhanian. Hey, Suzy.

Suzy Davidkhanian:

Hey, Sarah.

Sara Lebow:

You're not really with me. You're a borough away but.

Suzy Davidkhanian:

I'm closer than Carina. I also want to say thanks for having me back, but then, it's like doubling up, so hello. I'm happy to be here.

Carina Perkins:

I stole your line.

Sara Lebow:

Welcome back, both of you. Let's get started with free sample. Or did you know Segment or I share a fun fact, a tidbit or a question? This one was previously featured in our eMarketer Daily newsletter. It's a quiz. Do either of what kind of platform, Slack, the chat interface that we used to get our work done. What was Slack originally? It was not an instant messaging platform. What was it?

Suzy Davidkhanian:

Image sharing.

Sara Lebow:

Not image sharing. Weirder than that.

Suzy Davidkhanian:

Oh my god. a dating one.

Carina Perkins:

Social media networking.

Sara Lebow:

I like dating platform. That's a good answer. It's not correct, but points for creativity. Social media networking was a little closer. It was a video game over a decade ago Slack was called Glitch. Glitch shut down in 2012 and rebranded to Slack, which focused more on glitches chat feature than on having weird little avatars going on quests. I think that when I am on Slack, I feel like a weird little avatar going on a quest, but that's not what it is anymore.

Carina Perkins:

I wish I had an avatar now.

Sara Lebow:

I know.

Carina Perkins:

I wish we all had avatars for Slack.

Sara Lebow:

I wanted to change my slack picture to Lisa Simpson a few months ago, but I was worried people would get confused.

Suzy Davidkhanian:

Why would you pick her of all people?

Sara Lebow:

I like her. She's smart.

Suzy Davidkhanian:

Yeah, she's smart.

Sara Lebow:

All right, now it's time for our next segment, Retail Me This, Retail Me That, where we discuss an interesting retail topic. Today's topic is AI and hyper-personalization. AI was the topic last year in 2023 and I don't think it's gone anywhere this year. And retail specifically, we're seeing chatbots and personalized ads and personalized products and logistic management tools and visual try-on tools, you name it. But AI isn't actually new. It's been around for a while, so why are we talking about it so much right now?

Carina Perkins:

I think a lot of what's been happening in AI previously has been happening under the hood. So, it's been in the background of product recommendation engines, predictive analysis and things like that. And what's changed now with tools like ChatGPT is it's a more consumer facing things. So, there are chatbots where people are interacting with AI directly. And also I think Generative AI, which has become kind of big with ChatGPT is a game changer because it can generate new content and creative, so that really opens up the possibility of true personalization in real time.

Suzy Davidkhanian:

And if you think about all of that, AI is not new in terms of needing some sort of system to help retailers, especially with all the data that they have, make sense of it and find it from different areas of the organization. None of the systems talk to each other. So, it's like this data lake model where all of the data is somewhere and it pulls the information. It's just like machine learning. None of it is new, but Carina I think said it exactly the way almost everybody at NRF was talking about it, which is it is now much more consumer friendly and they understand and are part of that conversation.

Sara Lebow:

Yeah, it's almost like AI as an interface is new. The tech isn't, but the interface, the talking to it, the chatbots that's relatively new. Chatbots already existed. It's the branding that's new.

Suzy Davidkhanian:

Well, in that event, it's the dynamic, what Carina was saying around it's smarter and it understands what's happening and can tie things together. Kind of like how if you talk to Google Home and you're like, what's the weather? You don't have to ask every time, what's the weather in New York? It kind of already knows and you can have a back and forth with her. That's what's the generative part, right? Of AI.

Carina Perkins:

Yeah, I think it's that kind of real time contextual interaction, which is quite new with Generative AI.

Sara Lebow:

And is this something consumers want? I'm a little sick of hearing AI everything. It's a little scary. Do we like this?

Carina Perkins:

I think they want it when it's genuinely going to reduce frictions for them. So, I think I saw an IBM study and it said 86% of adults worldwide would like AI to help with researching products, and there was a similar proportion that said they'd like to use it to get answers. I guess it really depends on how good their experience with that AI then is. And I think in the kind of early iteration of these Generative AI chatbots, and perhaps people might use them a couple of times and think, okay, that's still a bit frustrating. But I think certainly among Gen Zs, there is an appetite for AI and shopping, but only if it's actually going to reduce frictions and not just create more frustration.

Sara Lebow:

Yeah, with recommendations and shopping assistance I think there's a tightrope to walk because on one hand you might like it if you feel that these products are picked specifically for you based on technology. That's reassuring. Technology can be smarter than I am, but on the other hand, if I know that technology is picking and delivering these products to me, I feel like I'm being told what to buy and I'm not so sure that I like that. That makes me feel like a sheep for advertisements.

Suzy Davidkhanian:

Well, I think you also said something really important, right? Consumers need to know that you're using AI and that the tools that they're using or your website has some sort of AI in the backend because before it did and nobody knew about it. And so, consumers didn't feel duped, but now they want to know about it. I think you can look at technology enablement in a way that widens the scope. So, it's not telling you what to do, it's just helping suggest other things to you that might not have been on your radar, but it could also reduce the scope. If the models are set too narrowly, then it only shows you what you already know you like. So, I think that also becomes how the retailer uses the model to either broaden the scope or to ensure a real sale and then they're just showing you what they know you already like.

Carina Perkins:

Yeah, I think there's definitely a danger of getting in a kind of feedback loop with that, whereas you only see things that you've looked at before and then you end up never seeing anything new and that then hinders product discovery. I think as well, there's a risk of intrusiveness. I think that although people like personalized experiences, there's a level at which you can get a bit too creepy. So, you're looking at something, you don't pay for everything in your basket and then you get an email with that product discounted. Some people will think that's great, some people will not like the idea that they're being kind of stalked online.

Sara Lebow:

There can definitely be a flattening effect where I'm looking at products, AI is feeding me more recommendations. I'm looking at those recommendations, I'm creating this feedback loop. I'm sort of being sent on this spiral. That sort of leads me down this muted suggestion. Building on a suggestion, I'm reading this book Filterworld right now by Kyle Chayka that's really talking about this algorithmic fatigue that you get with these sorts of things, and it does get rid of the thrill of the find of finding a product that you like and of looking through a magazine. Now we're getting into old media, but of curated styles that are curated across the board, maybe not for you. On the other hand, if I am buying a flashlight on Amazon and a product recommendation tool reminds me that I need batteries, I don't feel like I'm being told what to buy. I feel like, oh good, I forgot I needed batteries.

Suzy Davidkhanian:

Well, so it's that whole thing, right? It needs to help you do what you're doing better, faster with more ease. So, it's the removing friction, right? But I also think it's an interesting point. If the retailers are using the right "AI models," then it shouldn't be doing that narrow filter. It should be saying "All the people that are like you are also doing these other 72 things. And so, we wanted to surface those for you because maybe you didn't think about it." I don't know that that's Gen AI necessarily, right?

It's just regular AI that we're all very accustomed to. The idea around the Walmart, oh, I'm doing a unicorn princess party, and then it shows you all the things that you would need. I would say that's definitely leading you into that funnel, but also kind of making your life so much easier. So, I think there's this interesting balance, but the end of the day it's whatever data you put into your model is going to help you get the right kind of in quotation kind of model in terms of the output. It always comes down to what are you trying to solve for?

Carina Perkins:

Yeah, and I think the conversational element of Generative AI is interesting because it's ability to change search, so if you're searching for something, you'd have to say, I want a pair of black trousers. Whereas if you can have more of a conversation about it, I want a pair of trousers. Describe the field, describe the width. It's kind of a much more intuitive search and perhaps will aid in product discovery.

Suzy Davidkhanian:

Well, and I mean at the end of the day, we all generally speaking, don't think of online shopping when you go to a designated website as inspirational, right? You're going to Amazon because you are looking for something. So, there is a chance, especially if you say, "I'm looking for a black dress to wear to a wedding in Spain that needs to also make it to the beach with me" or "Help me pack my suitcase to go to Spain. I'm only there four days and I'm taking a carry on." Then you're going to start to discover impulsive things that you hadn't thought you needed because of the model. So I think it also all starts with how are people searching and looking and impulse is really hard online.

Sara Lebow:

I mean Amazon is doing just that. They just unveiled a chatbot called Rufus that was intended to do just that. We're seeing similar-ish AI tools from Google Cloud, from Walmart, even from Ulta Beauty. I don't know that I want to interact with Amazon to get inspiration. I don't want to feel like the Amazon machine, which already feels kind of soulless, is telling me what to get. I still get the appeal though. I still get why this is a good move for Amazon.

Carina Perkins:

I think Amazon is really the king of personalization. They do it kind of the broadest way from the moment you land on a homepage, you're seeing a customized page based on your buying history, your real time behavior, the emails you are getting from them. Everything is personalized. So, I think that they do get it right to some extent, although for some people that does feel a bit intrusive perhaps.

Suzy Davidkhanian:

But is it because it's Amazon? If you had gone to Nordstrom or to J Crew or whoever your favorite store is and that didn't feel like the giant who is eating everybody else's lunch, would it still feel as, I mean I put words in your mouth, icky?

Sara Lebow:

It wouldn't, and I think one of the reasons for that is because on some of these places, there are fewer choices. So on Amazon, I have every choice in the world available to me, and it's like looking through all of that. It feels a little more curated in these other places. I mean, I think this is what's going to continue happening with AI. We're going to get a divergence of hyper-personalized tools that really work. But then I also think some curated websites, retailers, newsletters, pieces of content are going to rise above the pack as people look for strong creators.

Carina Perkins:

And I think you're right, the endless product recommendation starts to feel somehow not personalized in a way because there's just so much of it, right?

Suzy Davidkhanian:

But also not only that, the endless aisle is brutal no matter what. And then the content that surfaces if you think about search, I think we've talked about how I was looking for a Dustbuster from a specific brand and I didn't even pay attention, which is a shame on me because it's the world we live in that I breathe every day. I bought some random, no-name, Amazon brand because it was I search and I thought I was going to get exactly what I searched for, but I got some other random thing. So, I should have done a little bit more reading before I hit the one click easy buy. And once you get it, you're not going to return it. That's just annoying. But I do think that there is this endless aisle that AI is supposed to help make less, but somehow also makes it even more endless, right? And that's tricky too.

Carina Perkins:

Yeah, and I think the challenge as well is that AI and Generative AI and hyper-personalization, it kind of fits really well with retail media and paid listings and positions and things like that. And then you are really balancing out the customer experience with the kind of retail media objectives. So, I think it's a very fine line for retailers to be treading.

Suzy Davidkhanian:

Sample size of one. I fell for it.

Sara Lebow:

Suzy, you should talk to our colleague Becky, who told me that she owns three different vacuums and Slacked me the other day asking if she should buy a fourth.

Suzy Davidkhanian:

Oh, she should not. There is no reason why anybody needs that many vacuums.

Sara Lebow:

All right, that is all we have time for in the first half. Now it's time for Pop-up Rankings, where we take a look at specific examples and we rank them. I want to talk about four promising examples of AI in retail, both on the customer facing side and on non-customer facing examples. Let's start with customer facing Carina. Can you give us an example?

Carina Perkins:

Sure. So, Zalando has developed a Generative AI powered fashion assistant that helps with product discovery with the conversational interface. So, it's exactly that story that Suzy was talking about earlier. You are going to a wedding in Barcelona in October. It needs to be suitable for the church, it needs to be suitable for the beach. You can have a conversation and it will suggest products for you. It might ask some more questions to refine the suggestions. You can say you don't like things, it'll give you different suggestions. So I think that's really interesting. It's still in beta mode. It's available in the UK, but I think having used it personally, I think it's a really promising start, but there are still some frustrations. It's not completely seamless experience yet. And I think that's really where we're at with Generative AI tools, to be honest, that a consumer facing.

Suzy Davidkhanian:

Well, I mean it sounds also like it's much more scalable than the Nordstrom personal shopper who's texting you when something new because they already know who you are and so now they're texting because you're a really great customer, and so they want you to know about the newest sale, the newest items that have come in that they think is going to match you.

Sara Lebow:

Yeah, I'm curious about how creator content is going to get folded into this. Maybe not this specific tool, but I am going on vacation in Portugal in the summer and I am getting TikToks about what to pack for that.

Suzy Davidkhanian:

Already?

Sara Lebow:

Well, yeah, we're planning. It's not that far away. So yeah, I'm getting those TikToks. I wonder how the creator economy will maybe get folded into these tools in the future.

Suzy Davidkhanian:

So, I think personally they'll be completely different, right? Because they might use AI to make their jobs easier, but I think the value prop of a creator or an influencer is a little bit different than I'm going to a giant Amazon-like website that has an endless aisle and that I am self-servicing, right? The creator is a little bit more push and the other one is a little bit more like I'm going to go get what I need.

Sara Lebow:

Yeah, it's that curational aspect. Carina, what is another example of a promising AI use in retail?

Carina Perkins:

So, this is an oldie but a goodie and it's Olay beauty brand. They have an AI powered skin advisor. So, you can take a selfie of your face and the app uses AI to tell the true age of their skin, which told me I was four years older than I'm, which I was quite offended by. It evaluates skin health and makes recommendations for a personalized skincare regime. And it asks you a few questions about problem areas that you think you have. So, that's a good example of personalization and potentially offending people.

Sara Lebow:

Ulta also has a skin analysis tool. A while ago, D2C shampoo companies were giving me ads where you take a quiz and it tells you what your hair type is. I don't know, is any of this real? Does any of this work? Does the AI skin filter know your skin?

Suzy Davidkhanian:

So, I think they're probably different, right? Take a quiz and get an answer is different than take a picture of yourself and that it analyzes the picture that it sees from a scientific perspective. I imagine it's right, but I don't know though, Carina, I would've never given you four years older looking at you. I would've given you 10 years younger. So, I don't really know. Maybe it is wrong.

Carina Perkins:

Thanks very much.

Sara Lebow:

Yeah, I agree. Okay, Suzy, can you move us over to the backend and give us a non-customer facing example of AI in action in retail?

Suzy Davidkhanian:

Sure. I think one of the things we didn't talk about upfront, which we talk a lot about at the office, is around AI not only needs to remove friction, but as we talk about all the time, it needs to add value for the consumer, make it a better experience, but also be easy for the retailer to implement. And at the end of the day, help them save money in some way, shape or form. So, in some of the research we did, I talked to a retailer who was saying, they also think about it in terms of three different stakeholders, one of which is their vendor partners, and they're using AI to understand refrigerated trucks. And so, if there are enough goods that are getting to a store spoiled, then that somehow triggers something and down the line helps them identify at what point in the journey of the item is there a mishap that's causing the spoilage, which I thought was very cool, right? Before that they probably had some sort of way of detecting that, but it was a manual process. So, not Gen AI, just AI.

Sara Lebow:

Refrigerated trucks are such interesting technology. We should do a whole episode on the logistics of that. But yeah, I mean all these logistical uses are really great and I think across industry, not just in retail, this is a place we're going to see, not Gen AI, but AI in general really be powerful, really be useful by taking lots of data and making use from it, and making it parsable with plain language rather than with complex formulas.

Carina Perkins:

Absolutely that, and I think you really hit the nail on the head when it's all about the data, and I think companies have had this data for a really long time, but it's being able to analyze it, being able to do predictive analysis on what that means for the future that is really powerful and the AI can really help with.

Sara Lebow:

Yeah, I feel like a lot of times I'm at conferences and I hear people saying there is a such thing as too much data. We collect so much data, we don't know what to do with it. And so, that's a really strong use case for it, is training AI on that wealth of data.

Suzy Davidkhanian:

Well, I mean I think that's a whole other podcast too, right? There's such a thing as too much data that's sitting in different systems that they don't know how to get them to communicate with one another. And then there's data where retailers don't understand how to tell a story because they're missing parts and pieces. So, not everybody sees the transaction that actually happens. So, if you think about the dress that you saw at whatever website, you actually purchased it, but there are lots of vendors who just see you looking at it and it being sort of part of your browsing history. All of a sudden you're getting ads for the dress you've already purchased, but it's because they don't have the transactional data. So, there is a lot of data, but we don't have all the pieces to together and they're just going to need to be more partnerships to try and make the whole customer journey in data available.

Sara Lebow:

Yeah, I don't need more ads for the purse I bought that looks like a giant binder clip. I already bought the purse. That looks like a giant binder clip, obviously.

Suzy Davidkhanian:

I bet you did buy that, didn't you?

Sara Lebow:

Oh yeah, I have it. It's right by me. I'll show you after this call.

Suzy Davidkhanian:

I would like to see.

Sara Lebow:

Carina on the topic of personalization. Can you give us our last example of AI in action?

Carina Perkins:

Yeah, sure. And this is kind of backend, I mean its consumer facing in some respects, but I don't think the consumers are necessarily a hundred percent aware that AI is generating it. And I think it's an interesting one because it really shows the difference between personalization and hyper-personalization using AI. So, traditional personalization and email marketing campaign would've relied on historical data and broad segmentation. So, you might send a different version of an email or a campaign to different segments of your consumers. But Starbucks, I know not strictly retail, but we're going to go with it, has got a real time personalization engine, which means that it can send literally hundreds of thousands of variants of personalized emails to consumers with personalized offers based on their individual preferences and behaviors. So, that's literally the kind of drinks that they're buying that week. They will then get relevant offers with it. So, I think that's just a really interesting example of how as AI technology develops, it's moving from sort of broad personalization to really individualized experiences.

Sara Lebow:

And this is one of the places where it showing that it's AI is paying off and making it more attractive. So, a similar example is those Spotify day lists that they're giving now where it has a name and it's made just for you. They could not put AI behind that. They could just be making it random. But the fact that it feels personalized to me is very appealing as a consumer.

Carina Perkins:

Have you listened to the Generative AI DJ on Spotify?

Sara Lebow:

I haven't, but well, currently I'm getting southern Gothic religious music Thursday afternoon, which does not fit what I listen to at all. So, maybe Spotify needs to give its day list DJ a little more trading.

Suzy Davidkhanian:

Or maybe this is to your point, maybe they are using AI to broaden your taste and your scope of what's out there.

Sara Lebow:

Yeah, I will give Southern Gothic religious music, all the songs on it I listen to. None of these are religious. I don't think that boy genius and a 100 gecs are religious. But I digress. This is all we have time for today, so thank you both for being here. Thank you, Carina.

Carina Perkins:

Thanks Sara.

Sara Lebow:

And thank you, Suzy.

Suzy Davidkhanian:

Thanks for having me.

Sara Lebow:

Please give us a rating and review wherever you listen to podcasts and follow us on Instagram at Insider Intelligence. Thank you to our listeners and to Victoria who edits the podcast and whose intelligence is far from artificial. We'll be back next Wednesday with another episode of Reimagining Retail, an e-Marketer Podcast. And tomorrow join Marcus for another episode of Behind the Numbers: The Daily.