Reimagining Retail: The role of GenAI in retail (beyond chatbots), retailer successes, and getting customers on board

Audio by Sara Lebow, and Yoram Wurmser | Jul 10, 2024

Episode Transcript:

Sara Lebow (00:01):

Hello listeners. Today is Wednesday, July 10th. Welcome to Behind the Numbers: Reimagining Retail, an eMarketer podcast. 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 leveraging generative AI in retail. Before we get started, let's meet today's guests. Joining me for today's episode, we have principal analyst at eMarketer, Yory Wurmser. Welcome back, Yory.

Yory Wurmser (00:34):

Hey Sara, glad to be here.

Sara Lebow (00:35):

Glad to have you been a while since we had you on the retail show.

Yory Wurmser (00:38):

It has. It has been a while.

Sara Lebow (00:40):

Also with us, super excited to announce a special guest, CEO of the New York office for VML and their Global Chief Commerce Officer, Beth Ann Kaminkow. Welcome Beth Ann.

Beth Ann Kaminkow (00:51):

Thank you. Thank you for having me. It's great to be here and join you both.

Sara Lebow (00:54):

Yeah, excited to have you. Okay, let's jump into this topic. AI, not a new topic. Generative AI kind of is, although everyone's heard of it by now. Almost a quarter of retailers have already deployed generative AI to automate customer service and another 25% are currently trialing it. That's according to a Google Cloud and NewtonX study. But people aren't really thinking about AI in the retail context as much as they are for other topic areas. A survey from YouGov showed that people think retailers will be least impacted by generative AI in the next five years, putting it behind social media companies and search engines. So with all of that in mind, tell me what is the role of generative AI in retail? Beth Ann, let's start with you.

Beth Ann Kaminkow (01:40):

Well, first of all, that stat just makes me laugh because I think... and maybe I'm biased, I'll start there, but I would put it first and one joke that we like to say here is that isn't everything retail nowadays as well, by the way? So the other things that you just listed too that were a little bit higher with maybe the putting science and a few other categories out of the mix, but I would say we see it prevalent in retail already and coming for retail in a really big way, which I think is really good news quite honestly for people like ourselves on the kind of partnership agency side of the business, but really with the end goal of making the consumer experience in the retail space much better, doing things that weren't possible before.

(02:20):

But I think people really want it to be possible, and I'm not just talking minority report stuff, I'm talking actual value to the consumer, which we'll get into. But I think it's a great thing and I think retail should probably be one of the top use cases when you start thinking about the role that generative AI is playing.

Yory Wurmser (02:36):

Yeah, I couldn't agree more. When you think about all the touch points retail has with consumers, the channels, the media, there's so many ways in which generative AI can be transformative. So yeah, I would put it pretty near the top of the list.

Sara Lebow (02:51):

Yeah, I mean it's kind of arbitrary right? I understand why they had to break things out, but social media and search, those are both retail not only in terms of marketing for retail obviously, but also commerce takes place on both social media and search like in the form of Google shopping. So those are both retail platforms in a sense.

Beth Ann Kaminkow (03:09):

Exactly. And also the retailer platforms themselves or retailers are both of those two things too. How many people are starting their search on Amazon or on TikTok and TikTok shop before they're thinking about searching through Google now? And of course we can talk a lot about what Google is doing to turn their search and their new AI generated functions, I think they're calling them AI overviews that they've been piloting, but all of that is for the shopping feature, I think to get more enhanced.

Sara Lebow (03:39):

Yeah, I mean, let's talk about some use cases for generative AI. Websites using AI chatbots are seeing 23% higher conversion rates than those without. That's according to a February 2024 article from Glassix. It's a huge difference. Chatbots seems like a pretty obvious, although still nascent example of generative AI in retail. What is the role of these chatbots and also what are other examples of generative AI beyond chatbots for retail?

Beth Ann Kaminkow (04:08):

I mean, I think chatbots is a great use case just because it was an underperforming area for a long time and customer service has been broken for a long time, so it's been an area that's so frustrating, especially as people move to more online buying behaviors. And if you can't get help, if you can't get an answer responded to immediately, then you just kind of go down this rabbit hole of a mess and then you start hating that retailer. So I love the fact that this has been an area that's been enhanced immediately. And I mean the thing to me that's partly at the heart, and we focus a lot on just making our retail partners and our manufacturer partners recognize the importance of customer centricity, and I think this should ground almost everything that we talk about in my mind.

Sara Lebow (04:54):

Can you define that for me, customer centricity?

Beth Ann Kaminkow (04:56):

Yeah, so it's really organizing around all insight and knowledge that has to do with the consumer, and that's one of the reasons why Amazon is always sort of out ahead and winning because it's maniacally focused on what the consumer wants and what the consumer experience is, what's going to enhance consumer engagement. And no matter how hard it is organizationally, they all orchestrate around delivering against that. And I think all things gen AI just enhance that potential. We're going to talk about some of the downsides and dark sides, but for the most part we're using this to enhance that capability because we're able to get deeper understanding into the things that are creating friction right now to the consumer and then things that are behaviorally intuitive that will make sense to start to intersect and enhance through gen AI value and processes.

Yory Wurmser (05:42):

And there are a lot of different aspects to the customer experience that can be helped by generative AI. You have obviously customer service and chatbots and concierge, automated concierge services, personal shopping, that part, but also things like search itself can be much more conversational. You can have the product pages a lot more vivid with 3D images that are generated from 2D images, things like that, upscaling images, creating different backgrounds for product pages. There are just a lot of ways that shopping can get better that's beyond just customer service that makes the customer experience better overall.

Beth Ann Kaminkow (06:21):

Yeah, we talk a lot about cultural relevance or having things feel culturally in tune with who I am, and I think this is a great use case because whether that's a background or whether that's a skin shade, we see Sephora experimenting here, we see Amazon making big moves here, but it's the thing that makes you feel like there's a cultural context to the content, to the communications. And then the other thing I'm just going to underscore that Yory said, I'm sure we're going to use this language a lot, is conversational. Just everything becomes more human and I think it's a great way of talking about AI because everyone defaults to the fact that this is going to be impersonal and more about technology interfering when in actuality tech becomes way more human and the way that we dialogue with each other becomes way more intuitive and human and conversational.

Sara Lebow (07:08):

Yeah, definitely. I mean, we already search so much more in a natural language way than we did 10 years ago, at least I do. I'll ask a search question instead of doing boolean search terms that it will continue in shopping. The Amazon example is really interesting because we talk about the Amazon flywheel at eMarketer a lot where you have their entertainment side, their commerce side and their advertising. And I can think of ways that generative AI is impacting all of those from creating the ads using generative AI to placing those ads and enhancing product pages to also maybe incorporating the ads into their content in the entertainment. So really strong example of where it's everywhere. What are some other examples of retailers that are doing a good job with generative AI?

Beth Ann Kaminkow (07:54):

Well, a retail that I've always had my eye on has been Best Buy partly because I think they moved to the more omnichannel approach pretty early on, and they did a good job of at least making you feel like these were connected channels and that they knew your behavior across them. So I do find that their app is showing new design capability within it that feels very powered by AI, that they're creating more distinct shopping experiences based on what they know your past behavior has been, and also maybe predictive behavior a bit as well.

(08:30):

They are customizing your home screen to look different based on who you are and based on your own personal preferences and they're letting you have a role in that too. So it's going to continue to get smarter based on the amount of information that you're willing to give it. You just know that they're pulling in your shopping history and your membership status and loyalty and things like that, or it's informing the way that they again, bring up search results or interact with you. And I think they may seem like basic obvious things, but it's been a long time since retailers have been able to really deliver against that, especially with the complexity of multiple channels that you can shop in. And I think the more that retailers do that, just more out ahead and the better relationship they're going to be developing long-term with you.

Sara Lebow (09:10):

That Best Buy virtual assistant is interesting too because if you're Best Buy, you need a reason to bring people into the Best Buy app. You need a reason that I'm not just going to go into the Amazon app and search there or on Google, a reason that I'll look at Best Buy specifically.

Beth Ann Kaminkow (09:26):

Yeah. Just to show another counter to that, so we all are familiar with Rufus from Amazon, it'd be interesting to see some of these things aren't necessarily going to be around for the sustaining future.

Sara Lebow (09:26):

Definitely.

Beth Ann Kaminkow (09:37):

And I think about Alexa in that way. We went through kind of an Alexa phase where everything was voice and people were talking to her as if she was a family member, but how often do you do that now, even Siri for that matter from a search standpoint. And so I do think some of the things that these companies are piloting and Rufus may be one of them that will serve a purpose to even help consumers understand the value proposition that comes with AI, but then will run its course in terms of the actual way that the value proposition from AI continues to come forward in the relationship.

Yory Wurmser (10:11):

Yeah, I think things are definitely iterating very quickly, and I think that's right that a lot of these first or really second generation generative AI features will evolve pretty quickly. I mean, talking about a retailer that I think has done some interesting things. I mean any of these virtual try-on or imagine your space type of applications on websites or apps, something with Lowe's has with designing a kitchen or Sephora with a makeup try-on or any of the clothing try-on apps or features, and those are all generative AI features and I think are great applications of the visual aspects of generative AI. And I think Beth Ann had some great examples of the large language model applications.

Sara Lebow (10:57):

Yeah, Beth Ann, you brought up Alexa. That's an interesting example because I don't like Alexa. I hate talking to her. I'd never want to have a conversation with her, but Amazon's goal was for us all to do that, to order our stuff from her. It has not come to fruition. What's stopping generative AI from following that same path from people like me being like, "I do not want to talk to a virtual shopping assistant. I'm not going to do it," what's different now?

Beth Ann Kaminkow (11:25):

Well, I think part of what's different now is it's a natural extension of the behaviors we already are taking and adding value to every one of those touch points wherever that's possible. So for instance, when you think about... and again to some of this was coming from understanding our physical shopping behaviors and then what could and should be different online and how do you exploit the online environment, but what also just needs to be similar for that continuity and because it is our natural way of thinking about how we shop.

(11:54):

So an example is I shop by aesthetic a lot. I'm not shopping for necessarily a specific item that I can describe in a search term, but I'm shopping for, "I'm getting ready to go to Cannes for the creative festival," or, "I'm going to Coachella for a music festival." And so my way I conversationally want to put that in there in a retail conversation is if I'm talking to an actual shop person who's there to help guide me through the store, that can now happen online in a way that you get exactly the kinds of things that you want to see, get the inspiration that you want to see, you get other queries possibly back from them looking for a little bit more information even about your size or what kinds of things make you feel comfortable, what's your favorite color.

(12:40):

It just becomes something that there's a different type of engagement associated with it, but it's all there to add a different type of value to your experience.

Sara Lebow (12:49):

Can't disagree. As someone who three weeks ago Googled, what do I pack to go to Portugal, definitely use case where Google's AI were a few generations down the road or maybe even now, I guess I didn't pay enough attention. They could be showing me what products I could be buying in that instant.

Beth Ann Kaminkow (13:06):

[inaudible 00:13:06] hyper local right? And I love that part of it too is that it can feel and it can start to deliver against being very localized, it knows and so it's not going to just default based on knowing. It's actually going to start tailoring and being more the spoken personal because it knows where you are, where you're going, exactly what you're doing, what hotel you might be staying at. The information starts to inform a very different mix of result.

Sara Lebow (13:30):

Yeah. Anything to add there on your end, Yory?

Yory Wurmser (13:32):

I mean, just add that it can be hyper local, I mean with retailers can actually contain the data that it dives into so that it doesn't go on these hallucinations like you get for ChatGPT. It can be very precise type of recommendations that aren't off the rails in any way.

Sara Lebow (13:48):

Yeah, I think that's a good argument for a really targeted chatbot or customer service bot, something that fulfills a purpose. I was talking to the folks at Zola, the wedding planning website about a bot that they made for splitting the decisions and something that the person who worked on that said to me was that just working with ChatGPT, that asks a lot of the user you have to come up with the problem that you want it to solve and then how to ask it to solve that problem. Creating one of these very targeted kind of kitschy bots asks a lot less of the user. So I also think that's a place where marketing teams from retailers are going to need to get creative so that they're holding users' hands and showing them what they can do.

Beth Ann Kaminkow (14:34):

Yeah, I think just to build on that too, what I also love is all the behind the scenes things that this is enabling that people may never really actually see, but they're solving for either pain points or they're just optimizing even a physical retail experience because you have supply chain data and understanding linked to local insight and data and geography specific information to what is popular right now, what is driving a lot of query, and then how do I refit and be more modular about my actual store footprint to adjust really quickly and adapt to what I think people are going to want the demand or how they're going to want to show up in store and then navigate a store footprint.

(15:16):

I just think... or even be very quick about experiences that can enhance a shopping experience because you're able to get a different level of intelligence collected behind the scenes through these large language models that can then inform changes that can be made on the store. I just think... I always love it when technology comes in right at the moment that you really need to start solving for the pain points that exists that is really holding something back. And so I think people are going to just start to feel enhancements hopefully in so many areas of retail.

Sara Lebow (15:49):

This is a technical question, probably more for you Yory, but the problems that Beth Ann just described, how does generative AI take us further in solving those as compared to AI that's been around for quite some time now?

Yory Wurmser (16:02):

Just making it hyper contextual and I mean, it's better at taking in a ton of information and picking out the important elements and then producing, giving back an answer that responds directly to the question instead of meandering back and forth like I'm doing with this answer. It gives back an answer that directly answers the need of the consumer that is aware of the full context that the consumer is asking or looking for.

Sara Lebow (16:31):

Gotcha. So those inputs on location and store, they're all coming in at the same time rather than separately.

Yory Wurmser (16:40):

Yeah.

Beth Ann Kaminkow (16:40):

Yory, I have a question for you based on things that we get from our clients a lot, and that is we are always trying to emphasize the need for pilots right now and just sort of testing and learning and just even AI applied to not just the pilot that you're running, but also the ability to AB test and get a lot of great feedback is pretty tremendous as well. How would you suggest that our clients prioritize what they're piloting or how they're thinking through the use cases and the jobs to be done?

Sara Lebow (17:08):

Very good question.

Beth Ann Kaminkow (17:09):

Yeah. Yeah.

Yory Wurmser (17:10):

I mean, it's a great question. I think it goes back to, first of all, what is the area where the consumer experience would be most enhanced by AI? And then to answer that, I think it's twofold is one is where do you already have the data in dispersed areas that you can use to create a really rich answer that would serve the consumer as well? So somewhere where you have data that's diffused and you can't really bring together in a coherent answer, but an advanced LLM, a large language model, could bring back a really clear answer. So something which takes into account context, location, shopping history, demographic information, things like that, bring it all together and then answering it. I think that's an area where you could really prioritize pretty well.

Sara Lebow (18:01):

I went to two different generative AI conferences recently, one from Tech Brew and one from Brand, and I'd been to both of them the year before, and something that I felt really changed this year as opposed to the year before is last year people were talking about what can AI do, what are all the places that we can plug it in, where can we put it? Now people are talking more about what are our problems that we already know exist and then moving from there into how AI can solve them. And I think that's a much more useful way of approaching things is problem first, customer first, what you were saying, customer centricity rather than AI first.

Beth Ann Kaminkow (18:37):

I totally agree.

Sara Lebow (18:38):

That said, the lift on generative AI isn't as high as AI, so I do understand the plug and play experimentation going on right now.

Beth Ann Kaminkow (18:46):

Yeah. Yeah, I totally agree. And the other thing that we just partner our clients with is thinking through the different platforms and not necessarily having to go at this alone. It is thinking through, "Okay, there's the Instacart, there's the Ubers, there's the Metas, everybody that kind of intersects with their retail experience for the consumer now and not just think about themselves isolated from all of these other platforms and partnerships that really enhance the experience for the consumer." And because everybody is getting into the generative AI space with different capability and utility, functionality, how do we map those things together and match them up in a way to solve some of the problems in a really cool shared way?

Sara Lebow (19:27):

Okay, so we've been really bullish on AI. There are definitely things that generative AI is not good for. What shouldn't we be using generative AI for?

Beth Ann Kaminkow (19:39):

We obviously have a tremendous amount of very talented creatives, and I wouldn't be doing them in injustice if I didn't say there's certain aspects of the creative process that still need a lot of human intervention. I think they've all not been slow to adopt. I think they have actually really adopted, especially when it comes to selling concept and when it's some design elements and some visual elements.

(20:00):

But I think when you talk about the integrity of a brand and you're not just thinking about loads of content for PDP pages, but you're really thinking about creative experience and the things that your brand is going to be at the heart of them, the level of human intervention and ideation and things that you still want to kind of purely come from human capital, yes, augmented by and generated by and supported by AI, but still very much about the artist and the originality that it still comes from, I think a human creative is an area that we see having to still maintain based on the incapability right now still. I mean, it's not to say that's not going to continue to evolve, but right now based on the concerns of brand safety and imagery and visual ability right now is just not there.

Yory Wurmser (20:49):

Yeah, I'd 100% agree with that. I think anywhere where precision is extremely important, Sara, I think you were mentioning earlier about some health related and insurance for instance, something where you can't have any type of hallucination or faulty answer, I think that's where you still need human oversight or direct human response, and then definitely creativity. I couldn't agree more with Beth Ann about that.

Beth Ann Kaminkow (21:14):

Yeah. And just to underscore your point too, the misinformation, anywhere that misinformation could still get pulled in or bias information because these larger language models are still training on something and some of that is still biased in terms of what it's suggesting. And so I think those are things that need a lot of human intervention. So it is true kind of augmentation versus replacement.

Sara Lebow (21:35):

Yeah, I mean the AI is trained on biased data. That said, as a person, I've also been trained on biased data, so it's hopefully something that I can critically think through and maybe the AI can eventually, but yeah, it's just working with the data it's been given. Wrapping up, can you guys give me a top line philosophy retailers should have when approaching generative AI? How should they be thinking about it?

Beth Ann Kaminkow (21:59):

You are usually talking three Cs. I've came up with these three Vs. So mine are vision, value, and visibility. And so I think starting to your point about a consumer-centric vision out versus generative AI starting at the start, I think it's thinking through what are the outcomes that we're trying to drive here? What are the pain points? What are the business solutions? We're talking retail, so what are the selling solutions that we need? Starting with that really well-defined vision. Value to me is both value in terms of this has got to bring value to the consumer, sometimes it's efficiency value, but if something is enhanced for me, if I feel the true benefit from it, then I'm on board. But also values driven where again, the commitment, the trust, the credibility is established and really important. And then visibility comes from the discoverability of great search and brands are trying to break through and it's getting more crowded and competitive and complex. And so how do we use this to enhance their visibility to the right people at the right time?

Sara Lebow (23:01):

Yory, any final thoughts?

Yory Wurmser (23:03):

Yeah, no, I mean I think that sums it up really well. I would just say it's see where it makes the biggest impact, what's the outcome that you think makes the best impact for your customers? And also the trade-off, is the technology ready there? How much would it cost you to actually get a solution that measurably improves what you already have? So just that, the cost benefit analysis.

Sara Lebow (23:26):

Okay. That is all we have time for today, even though I feel like we could keep going for another 30 minutes. Thank you so much for being here, Beth Ann, thanks for joining us.

Beth Ann Kaminkow (23:34):

Thank you for having me. It was fun talking to you both and I agree we could probably go on for at least another episode.

Sara Lebow (23:40):

Thank you, Yory.

Yory Wurmser (23:41):

Yeah, I agree. It just was a lot of fun.

Sara Lebow (23:43):

Thanks to our listeners, to Victoria who edits the podcast and to I am adding in as the fourth V to vision, value, and visibility. We'll be back next Wednesday with another episode of Reimagining Retail, an eMarketer podcast. And tomorrow, join Marcus for another episode of the Behind the Numbers daily.