Reimagining Retail: Examples of how retail is using generative AI and the most promising use case in 2025

Audio by Sara Lebow, Sky Canaves, and Yoram Wurmser | May 2, 2023

On today's episode, in our "Retail Me This, Retail Me That" segment, we discuss why retailers should be paying attention to generative AI and how brands and retailers can prepare for it. Then for "Pop-Up Rankings," we rank the top four most interesting examples of how retail is using generative AI. Join our analyst Sara Lebow as she hosts analysts Sky Canaves and Yory Wurmser.

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

Episode Transcript:

Sara Lebow:

Hello, listeners. Today is Wednesday, May 3rd. Welcome to Behind the Numbers: Reimagining Retail, an e-Marketer podcast made possible by Adobe. This is the show where we talk about how retail collides with every part of our lives. I'm your host, Sara Lebow. Today's episode topic is ChatGPT and generative AI in retail.

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 on.

Sara Lebow:

And also with us is Principal Analyst Yory Wurmser. Hey, Yory.

Yory Wurmser:

Hey Sara. Glad to be here.

Sara Lebow:

Glad to have you both. Okay, let's get started with our first segment, News and Reviews, where I give the news and our guests tell me their reviews. Today's story is a Retail Brew article from April 25th titled The Resale World is Paying Close Attention to Emerging Technologies. Resale is hot with Gen Z and it's using ChatGPT, is a rhyme that I wrote. The story talks about how resale is increasingly technological with Depop's CEO saying, "Hopefully in the future people will be able to upload flat photos of clothing and we'll be able to turn them into photos on a model." Carina, your review of this story in 60 seconds is.

Carina Perkins:

Sure. So I think AI definitely has a place in retail, as it does with all e-commerce. Some people love shopping secondhand because they love the experience of sifting through different products, but some people are primarily doing it because they're looking for cost savings and they want to be more sustainable. And buying secondhand can be a bit overwhelming sometimes when you've got such kind of a huge amount of products and variations. So you could really see the opportunity around kind of AI powered chatbots that could help you guide through what's available, what's in your size. I can see really a big opportunity there.

And then also from an operational perspective, having such a huge and varied stock can make some of the tasks of getting those products online quite time consuming. So things like automated product descriptions, again, could be a really kind of big benefit for resale. So you could see that it could drive efficiency and user experience.

Sara Lebow:

Yeah, I feel like the volume of stock and how different all of those items are in their listings is one of the biggest challenges for eCommerce for resale.

Carina Perkins:

Yeah, absolutely.

Sara Lebow:

So makes sense to me. Yory, your review of this story in 60 seconds is.

Yory Wurmser:

So, like Carina, I think it's really interesting to apply this to resale. My take more is on the seller side. I think for all these small sellers that don't have a marketing team behind them, it might be really great to create product summaries, to create more 3D images of these products in ways that until now, only a marketing department could do or could do well. So for a lot of resellers, so I think that's going to be one huge application. And the second is digital twins. This article mentioned it, with 3D images you can create digital twins, which might be great ways to track authenticity and the providence of these products.

Sara Lebow:

Can you describe what digital twins are?

Yory Wurmser:

They're basically digital versions of the physical clothes that you're buying or The physical object you're buying. So you can wear your clothes in the game or you can go into the metaverse, whatever that may be. But more generally it's something where there's a digital record of it online.

Sara Lebow:

Sure.

Yory Wurmser:

They can track where it's from and so forth.

Sara Lebow:

So really useful for authenticity if you're buying, I don't know, what do you buy resale? An Hermes bag? That's not something you'd buy resale. Well, maybe. Okay. 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 ChatGPT and generative AI in retail. Let's start with the basics. Not so basic I guess. But spending on AI centric systems worldwide will jump from 121 billion last year to $154 billion this year. That's according to the International Data Corporation. Generative AI is the topic right now in retail and marketing. How is generative AI different from regular old AI? Yory, I'll hand this one to you.

Yory Wurmser:

Yeah, I mean, you can just look at the first word, generative. It's about creating content, whether it's creating a text from a very simple text prompt or creating images or songs, 3D images from 2D images. So it creates content or in some cases enhances content from let's say 2D to 3D or low definition to high definition.

Sara Lebow:

And then typical AI does not create.

Yory Wurmser:

No, I mean they're both models, which means ways of replicating the way we think. So that's what artificial intelligence is, these are these neural networks that are how we think. Generative AI are those networks that can create content.

Sara Lebow:

Okay. So then, Carina, you just wrote our report on ChatGPT and generative AI in retail specifically. Why should retailers be paying attention to generative AI right now?

Carina Perkins:

I think just because the potential use cases for retailers are so broad. You've got everything from chatbots that can deliver really kind of human personalized speech, to generating product descriptions, marketing copy, advertising creatives. You could even use it in the backend for HR emails and supplier negotiations. So it's really such a wide potential use cases across every aspect of the business. So it really makes sense for retailers to be looking at it now and thinking how they could apply it to their business and drive efficiency.

Sara Lebow:

Yeah, I was talking to a retailer at eTail who had used generative AI to write up all of their product descriptions for a ton of merchandise that they'd acquired, and were able to do weeks of work in a couple days. On that note, what is the most promising use case in retail for generative AI right now? How can we be using it right now?

Carina Perkins:

So, I think as it stands, really the most promising use case is around that kind of content generation. So it's using it to create product descriptions, marketing copy, blogs, social media posts. You can even translate your website copy into a wide range of languages. At this stage I think in the products development, you still want to be reviewing that copy and giving it a kind of human check, but you can drive big efficiencies by automating manual tasks and that will free up your human staff for potentially more creative tasks.

Sara Lebow:

So if I'm a retailer right now, I want to be looking at generative AI to create product listings, blog posts, and sort of content generation?

Carina Perkins:

Yeah, any content you could think that as a retailer or a brand you might want to be generating, really you can automate it and then you can still have a human checking that content over and choosing. So you could use it to iterate a 100 different versions of a social media post, but you would still have someone then looking through and making the decision on which one to post. So I would say at this stage, the kind of riskier use case would be anything that's directly customer facing without a human check because there are some concerns around hallucinations and inaccurate responses that need to be ironed out, I think, before that's a kind of really promising use case for it.

Yory Wurmser:

Yeah, I could totally agree with that. And I would just add that those guidelines are where all the work is going to be, I think, along these customer facing content. You really want to vet that and have clear guidelines on what is acceptable and what's not.

Sara Lebow:

Yeah, hallucinations in AI, that's when the like ChatGPT or something confidently tells you something that is completely made up, not true.

Carina Perkins:

Yes.

Sara Lebow:

Okay. So we have the now. Looking a couple years out, in 2025, what will be the most promising use case for AI in retail?

Carina Perkins:

So I think the most exciting possible use case is this idea of generative AI powered chatbots that can essentially work as a personalized shopping assistant and can give an individual a completely personalized experience when they visit a retailer's website. So like we said, there's a few issues that need to be ironed out, I think, before they could be really effective. You wouldn't want to be giving customers wrong answers yet. But I think eventually as well, there's a hope that you'll be able to really personalize and give product recommendations based on previous preferences, previous buying habits. So I think that's where the big potential of it lies, but we're not quite there yet in terms of development.

Sara Lebow:

Yeah, I was talking to a marketing professor this past week who was talking about the potential of hyper personal ads. It knows that I like buying jeans in this size, this color, and I prefer shopping on rainy days, and serving me ads specifically for that. Is there a concern about customers being creeped out by that or will we all just adapt into that new era of personalization?

Carina Perkins:

I think there is a concern around, more importantly than customers being creeped out, I think it's about data privacy and I think that there are some concerns to be ironed out around how these large language models use and store data before customers will be really comfortable with that.

Yory Wurmser:

No, I mean, I think that's right. I think privacy issues are one's where not only is it will customers be concerned about that, but they're going to be rules, probably government rules, around it. So it will have to be opt-in, and if people opt into it, then they'll probably naturally be a little less freaked out. But I think that the opt-in rules will have to be very tight.

Sara Lebow:

Sure. Okay, jumping back to your ChatGPT and retail report, Carina. What was your most surprising finding in researching for this report?

Carina Perkins:

So it was, and I'm no AI expert prior to looking into this for retail, so for me it was really that generative AI isn't new at all. It's been around since the 1960s when it was incorporated into a chatbot named Eliza. So I think what's really changed now and what's really exciting about products like ChatGPT and Google Bard is that these are really kind of easy to use interfaces that put generatively the AI in the hands of everyone. It's in the reach of everyone now. And now we've got a lot of providers coming in and integrating that technology and providing kind of commercially available products that retailers can just plug and play into their existing systems. So it's really speeding up that kind of AI revolution in retail, I think.

Sara Lebow:

Yeah, any real Jeopardy fan will remember when IBM Watson competed on Jeopardy as an AI bot in like 2011. Okay, and then what are two ways that brands and retailers can prepare for AI's rise right now?

Carina Perkins:

So I would say one way is really just to learn more about generative AI, so familiarize yourself with generative AI and its capabilities. And then all of the different kind of models that are available, consider their strengths and weaknesses to try and determine which is best suited to your needs. And as the technology is incorporated into retail, as you've got people testing it, really keep an eye on that and look at the successes and failures to understand what the opportunities are and what the potential pitfalls are.

Yory Wurmser:

Yeah, and I'm just going to go back to the concept of guidelines. I think setting clear guidelines on when to use it and how to use it is important because there's so many ways to go wrong still, especially when you're doing customer facing content. You want to be experimenting a lot on it, but you want to have clear guidelines on how to do it.

Sara Lebow:

Yeah, that makes sense, especially with copyright concerns, especially with hallucination concerns. I was talking to someone this week about at what point you should be looking at outside hires who know more about AI and they were saying a lot of us are in the same boat right now, so now's the time to be experimenting.

Yory Wurmser:

Yeah, I don't even think you need to hire someone because everyone is so new at it. Just find people who are really into it and willing to experiment and learn from it. And those are people you want to rely on.

Carina Perkins:

Absolutely, I think really trying to get everyone at every level really as involved as you can is a really good idea because it is going to have implications across business, so you want as many people to be familiar with it as possible.

Sara Lebow:

Yeah. This is a takeaway I keep hearing is whether you like ChatGPT or not, you should be familiar with it at this point. Okay. You can read more about what Carina has to say about ChatGPT and retail in her ChatGPT and retail report, which is available now. Or you can listen to both Carina and Yory talk about this topic in our recent ChatGPT webinar. Links to both of these are available in the show notes.

We're going to take a quick break before jumping into some specific examples. But first, a quick message from our sponsor, Adobe.

Adobe Research found that 73% of customers want to be delighted by great personalization experiences, yet just 18% of the retail brands they've interacted with recently met that standard. From integrating technology and data, to empowering employees, discover how retailers can deliver seamless omnichannel experiences in Adobe's 2023 Digital Trends Retail and Focus Report. Go to adobe.com/go/trends to learn more.

Well, welcome back from the break. Now it's time for Pop-up Rankings, where we take a look at specific examples and we rank them. Today we'll be ranking the top four most interesting examples of how retail is using generative AI right now. So let's start off with our first example. I'm going to you, Yory.

Yory Wurmser:

Yeah, I mean, I think one great example is Instacart. It's not a classic retailer, but they were actually one of the pilots for the OpenAI, which is the creator of ChatGPT for their API for a ChatGPT, which basically lets Instacart take these questions from users that are open-ended, that don't have a clear answer, like, "What should I make for dinner? "Or, "How do I make good fish tacos?" I think that that was the example they actually used. Instacart then combines it with its own AI to make product recommendations. So it understands the query, it makes it a recommendation, and then it connects to specific products. So I think that's just a great example of combining the language understanding with specific product recommendations.

Sara Lebow:

Yeah, I think that these plugins from ChatGPT and OpenAI are going to be one of the biggest growth areas in the next year. These opportunities to use ChatGPT's API, OpenAI's technology, and input that alongside your website, I think are what we're going to see a lot of growth in.

Yory Wurmser:

For sure. Yeah, no, totally agree.

Carina Perkins:

Yeah, I agree as well. And I think combining ChatGPT's with different AI capabilities is really interesting as well because it lets you overcome some of the limitations with ChatGPT as an interface. So ChatGPT can't perform a task like adding a product to a basket at present, whereas other AI can. So I think we'll see more of that kind of combining the two in the future.

Sara Lebow:

Sure. I mean already there's an example of a plugin with Expedia, I think also, where you can be like, "Where should I go this winter to stay warm and not spend too much money?" And it can plan a full trip to Fort Lauderdale for you.

Carina Perkins:

Yeah, and Klarna is launching a shopping plugin with OpenAI, and I think they've really stolen a march a bit on that one because that could be really interesting in terms of personalized shopping recommendations. And it allows people, it gives personalized suggestions with links to shop those products on their search and compare tools. So they've done well to get in on that quickly, I think.

Sara Lebow:

Yeah, it feels like a huge opportunity for Klarna, which really just felt strictly like a buy now, pay later platform even a year ago. Now feels like a whole shopping interface.

Carina Perkins:

Absolutely.

Sara Lebow:

Okay. Can I have another example of how retail is using generative AI from you, Carina?

Carina Perkins:

Sure. So I'm going to talk about a brand here, so Coco-Cola. It was the first brand to engage with the new Global Services Alliance that's been formed between OpenAI and Bain Consultancy. And so Coca-Cola is exploring the use of tools like ChatGPT and DALL·E to overhaul its marketing strategy. So it's creating and testing variants of advertising creatives and marketing copy. And I think it's a really good example of a brand looking to drive efficiency and really kind of scale up. And what that will enable Coca-Cola to do in time is that personalization. So I think that's another really interesting example.

Sara Lebow:

Yeah, I feel like Coca-Cola is leaning into almost like the gimmick aspect of ChatGPT, which could be a criticism. I think they had some sort of challenge where fans of the brand could submit images that they'd created through AI as potential marketing images. A really good way of leaning into loyalty, leaning into the excitement, and getting someone else to do the marketing work for you.

Carina Perkins:

Sure. But I think, and it's also doing quite serious work on the backend, and I think what's really interesting is, again, this idea of personalization of individualized marketing. That kind of the scale you would need in order to create those creatives previously wasn't really possible. So it's something that generative AI is now enabling brands to do. So I think it's really an interesting area.

Yory Wurmser:

Yeah, I agree. It's much more than gimmick. I agree, Sara, there are gimmicky aspects, but just that mass personalization is I think the future for a lot of marketing.

Sara Lebow:

Yeah. I mean Coca-Cola for a long time has had the cans with people's names on them. Probably going to get a lot more specific with those names. I'm mostly kidding there, but it's the same concept.

Carina Perkins:

Yeah, definitely. And consumers want personalized experiences when they're shopping. So I think this technology just has such huge potential for giving those experiences. And I think the names on Coca-Cola cans is a really good example of that, and this is the kind of next level of it really.

Sara Lebow:

Yeah. Now I can be sure every time that my Sara will be spelled without an H. Yory, can you give me another example of retail using generative AI today?

Yory Wurmser:

So I'm going to go once again to a platform, and that's Google. They have Shop The Look, part of their shopping platform. Sellers can submit 2D images and Google's generative AI turns it into a 3D image, which I think is a great enhancement for the product and for looking at the product. So I think that's pretty cool. Just any way you can make it more interactive or create a more enhanced image, that improves conversions and shopping.

Sara Lebow:

Is the next layer of that going to be viewing the look on your own body? Is that already here?

Yory Wurmser:

Yeah, I mean it is already here. That's exactly the next layer. In fact, I think Amazon is doing that to an extent as well. But, yeah, for sure, that's one clear application.

Sara Lebow:

Yeah. I don't want to know how clothes will look on me. Surprise me, have me stuck with something I don't like. But I mean, yeah, it's probably another good way to decrease returns as well, another topic that we've covered on the podcast.

Yory Wurmser:

Yeah, Snap and Amazon are just doing a ton of stuff on being able to try on these clothes virtually. So it's definitely here.

Sara Lebow:

Is Snap doing like shoppable try-ons?

Yory Wurmser:

I think so, yeah.

Sara Lebow:

Cool. Okay. Carina, can you give us the final example of retailer using generative AI?

Carina Perkins:

Sure. So I am actually going to give an example of a retailer now. So this Zalando, and it has said it's going to launch a ChatGPT powered fashion assistant by the spring. So it's going to use the OpenAI technology to allow customers to ask questions using their own fashion terms and words to help them navigate Zalando's product assortment. So the example they gave was asking, "What should I wear for a wedding in Santorini in July?" And the fashion assistant will be able to understand that it's a formal event, that the weather in July is going to be, and provide product recommendations based on those factors. And they've said that in the future that could potentially be combined with customer preferences such as brands they've shopped before and products available in their sizes. So this is a really kind of consumer facing launch, so it'll be really interesting to see how it goes and how it's received by customers.

Sara Lebow:

Man, I want to go to a wedding in Santorini in July.

Yory Wurmser:

No, mine are mostly in North Jersey, so Santorini sounds pretty nice.

Sara Lebow:

Santorini sounds great. Yeah, I'm sure the clothing recommendations will be very similar to what you're wearing in Santorini.

Okay, that is all we have time for today. Thank you for joining me, Carina.

Carina Perkins:

Thanks, Sara.

Sara Lebow:

And thank you, Yory.

Yory Wurmser:

Glad to be here.

Sara Lebow:

Please give us a rating and review wherever you listen to podcasts and follow us on Instagram @behindthenumbers_podcast. Thank you listeners and Victoria who edits the podcast, using her own neural network. That's her brain. We'll be back next Wednesday with another episode of Reimagining Retail, an e-Marketer podcast made possible by Adobe. And tomorrow, join Marcus for another episode of the Behind the Numbers Daily.