What can AI do for retailers and customers?

Episode transcript, edited for clarity.

Bill Thorne: Welcome to Retail Gets Real where we hear from retail's most fascinating leaders about the industry that impacts everyone, everywhere, every day. I'm Bill Thorne from the National Retail Federation, and on today's episode we're talking to Nasrin Mostafazadeh. She's the co-founder of Verneek, a company that's building AI platforms that enhance the customer experience. We're going to talk to Nasrin about using AI in retail, what's behind all the buzz and how retailers can learn more about this revolutionary technology. Nasrin <laugh>, welcome to Retail Gets Real.

Nasrin Mostafazadeh: Hi Bill. Thank you so much for having me. What pleasure.

Bill Thorne: Thank you so much for taking the time because this issue, and I don't even see that I would call it an issue as much as it is an emerging technology. Everybody's talking about AI and one of the things at the NRF that we're really working with our members and other groups to determine what does this mean for retail? I mean, where does this come into the whole ecosystem of retail? And so you are going to help us understand that a little bit better. Let's get to it. But before we get into the substance of that, tell us a little bit about you and how you came to be where you are today with Verneek.

Nasrin Mostafazadeh: So my background is AI. I've, basically, my main practice is as an AI scientist. Now I happen to be, having started my own company, Verneek — If you want to go all the way back, I actually started in AI 18 years ago.

Bill Thorne: Wow.

Nasrin Mostafazadeh: By virtue of having gotten into robotics. So both actually, my co-founder and I, which is literally where we met, we were working on robotics and like ROBOCOP competitions. And then fast forward, turns out that was not at all the hardest problem in AI. Building these multi-agent systems that can basically figure out how to play a game of soccer or rescue a world that is in disasters in a simulated way. Turns out that was actually one of the easier problems for AI. At the time, this is now almost 18 years ago, I came across this thing called Moravec's Paradox, which basically entails that whatever is hardest for machines is actually the easiest for human beings and vice versa.

So actually the most basic cognitive abilities of, like, a three-year-old, four-year-old child turns out to be the hardest to replicate and imbue for machines. So I literally ended up applying my entire life to it to this day, which the name of the field is called commonsense reasoning. It's basically the studying, the science and basically engineering of building these most basic, commonsensical functionalities that we all have as human beings. Like how we basically acquire word knowledge, how we apply it in context, et cetera. This might actually be interesting to you, since everyone is now talking about AI, I have been around long enough to know how gradual the progress has been and how this is not an overnight progress, which is honestly the reason for a lot of fear that exists in the broader kind of mainstream is that people feel like, oh my gosh, I woke up one day and oh AI is taking over the humanity as if, right?

And what they don't see is that it really has been incremental. The example I always like to make, and this is literally how I've opened up every single research talk of mine in the last 10 years, is that I exemplify the motivating example that I came across that got me into the field of commonsense reasoning and language AI in general is as follows. So the task is super simple. You read a very basic sentence and you have to answer a question about it. The sentence goes as follows: The monkey ate the banana because it was hungry.

Bill Thorne: The monkey ate the banana because it was hungry. OK.

Nasrin Mostafazadeh: And the question that as a machine you pose is, what does it refer to? To resolve that core reference, and why can we do that instantaneously? Even a three-year-old child can do that because we have all this word knowledge, which are like billions and billions if you wanted to enumerate it.

And then we know how to smartly apply it in context without even realizing that we do that. And that's how we realized that "it" pronoun actually refers to the monkey, not the banana. So that problem was counted as exemplify for the holy grail of AI, that this is the hardest thing for AI to solve. If we ever solve this and problems like this, we are going to be able to see more general intelligence like cognitive abilities of human beings imbue the machines. So fast forward, believe it or not, this problem ended up being unsolved throughout my PhD studies, throughout my work as a scientist. So I always, when this is a visual conversation, I always showcase how the state-of-the-art in 2016 was failing on it. The state-of-the-art AI models, the state-of-the-art in 2018 was failing on it. And fast forward, how ChatGPT is now all the—

Bill Thorne: Yeah, totally.

Nasrin Mostafazadeh: You know, sensation and ChatGPT would fail on it. If at the time when ChatGPT came out, you would just literally give it this question. It would say that it refers actually to the banana wrongfully so. But then this very same year just when the next GPT model came out, GPT four, it finally could solve it. I always open … talk with this to exemplify first and foremost how much we still have these to go right. To scratch the surface in even bringing human-like cognitive abilities to machines. But to also pinpoint this time, basically in the history of AI that, look, we are actually closest that we've ever been in giving machines common sense, which has everything to do with why we founded Verneek in 2021.

I could see through my line of research that in the academic world we were doing and in industry and big tech, that we had gotten the closest in actually unlocking the most basic common sensibilities in machines and I wanted to be part of the equation in basically bringing these technologies to the masses, your average Americans, in the most helpful way that you can actually have. Because if you can do the most basic reasoning, then you can build all sorts of really category-defining products on top of it that can help your day-to-day lives.

Bill Thorne: Well, you see these headlines, and I just saw the headline the other day about, we need to be careful because AI could mean the extinction of the human race and all of these wildly scary things that people are saying about this technology. Which I know is not new, but it sure is, everybody is talking about it. It's been around for a long time, but now you, I don't know that you can pick up a newspaper or watch a newscast or even talk among friends where people aren't talking about AI. It's insane.

Nasrin Mostafazadeh: It is insane and exhausting <laugh>.

Bill Thorne: I'm sure from your perspective it is exhausting, but why, is it just fear of the unknown? Is that it?

Nasrin Mostafazadeh: I mean, to be honest with you, I will take off my Verneek hat and tell you what I believe, like, as an AI scientist, what this moment in AI really means. First and foremost, as I've mentioned, it's not this one moment. It's been a very gradual incremental progress, albeit very exponential. Absolutely we've made strides that I personally would've predicted will take us five, six more years. From the scientific side of things, I honestly think that a lot of this fear-mongering has to do with the fact that a lot of our policymakers, a lot of mainstream people were not educated about what was happening in the field.

Bill Thorne: Mm-hmm.

Nasrin Mostafazadeh: I think the more we talk actually in these kinds of conversations, the more we shed light into where we come from and where we're going, the less that fear will be. And also the more we talk about the upside. The fact that this is a productivity tool. This is, like, literally start a whole company around building helpful AI because I believe that this is going to change the livelihood of your average Americans to live better, healthier lives. And that really would not have been possible without these kind of tech technologies and breakthroughs. With all of that said, the truth is there are so many scientists that are now like Geoffrey Hinton, me being one, that would come forward and want to sound the alarm, look, this what I'm seeing, it could really be that whatever, let's say 1% chances that it can really go rogue. It could have, of course, the supposedly super intelligence and all of that, which I, to be honest with you, it's very hard to decipher <laugh> the hidden agendas in terms of what is going on. The people who are insiders and talking about it, OpenAI being one, now with $11 billion that they have in their pocket, they're kind of sound, sounding the alarm for policymakers to go and regulate them.

Bill Thorne: Now that they've got it.

Nasrin Mostafazadeh: So there are all these other agendas that one can assume may or may not exist.

Bill Thorne: Sure.

Nasrin Mostafazadeh: Setting that aside, I do think that the truth is this is powerful technology. It's not just that it became powerful overnight. It has always been on the path to become powerful, not for the sake of becoming sentient and taking over humanity. People should think about how we would basically navigate the job displacement, which this is going to have at this certain rate that prior technology shifts have not had. And then second, we should think about misuses. The truth is, look, we have our upcoming elections, right?

Bill Thorne: Right.

Nasrin Mostafazadeh: We are going to have a machine that can really generate misinformation and disinformation at the speed and rate that no one had before. Let’s talk about those things. Those are real threats. In my opinion, as a scientist, the fears should not be about superintelligence, which may or may not happen in the next many decades. But should be actually very much about the misuses of as we speak. We will have, and I threw that classic example from 18 years ago with the monkey eating the banana. I told you how these models already are making, have a lot of failures as we speak. So extrapolating from that to being just super intelligent human being that can have this back and forth that you and I are having in a very nuanced way, and being able to actually have actions in the real world that will take time. I think that is a distraction from us as a society focusing on the real threats that we have to deal with right now.

Bill Thorne: Yeah. No, I agree with you 110000%. Retail and AI, this is where you’ve kind of focused a lot of your efforts, I believe. And so tell me, what does AI have to do with retail and what does retail have to do with AI?

Nasrin Mostafazadeh: Let me just talk about the underlying technology that we’ve built. So more than two years ago, having seen this inflection point, because of being insiders, we knew exactly what was going on. We wanted to basically build a domain general AI platform that can basically enable you to put voice or text interfaces on top of any digital or physical environment, if you will. So this domain general technology, after doing it, doing a ton of R&D et cetera, when we built the technology, now time was for finding out what would be the use cases that will be the best. Not just for the sake of the technology in terms of its efficacy, but actually for the sake of consumers. Because remember, our entire mission is to build the most helpful AI that can help anyone, everywhere at any time, which I think by me—

Bill Thorne: Not to take over the world, you mean.

Nasrin Mostafazadeh: <laugh> Not to take over the world. Precisely. I think by just the last three words I just used, which is how you started the podcast today, you can see why we ended up going after retail. We really wanted to apply our technology first and foremost in a domain that will interface with every average American and what better world <laugh> than retail, because really everyone uses it anywhere at any time. So, that was really one of the reasons for why we gravitated towards picking retail as the first contender for disrupting it. We had so many other reasons, honestly, it ended up being our blue ocean strategy because of the fact that we saw that because of the fact that retail was so under-digitized, under-innovated and underutilized, it was really in a very good position to seeing the true fruits of this breakthrough moment in AI.

Because, so there are no legacy systems within retail, as many as you can find in finance, health care, et cetera. It's so much easier to give them a box, basically that is AI plus knowledge in a box that they can then use from scratch. Starting from scratch, which is now so much easier to adopt, so much easier to integrate and beyond, which is really one of our main strengths. I would say that it is really a turnkey solution that we've built. Another reason was the fact that it's kind of interesting, but due to the pandemic, a lot of retailers had no choice but to have digital twins of their stores and get further digitized. So we saw this as an opportunity for our blue ocean strategy to be able to integrate and implement Quinn, which is the name of our AI platform, faster in this particular sector of the economy.

We basically went after a place that we had no competitors, <laugh>, and then the fact that it was actually primed to be disrupted and the data was there, the initial digitization due to the pandemic was there. So that's why we ended up having anything to do with retail, for the sake of our technology. You asked also the reverse side, what does AI have to do with retail? I do believe that actually with all the buzz around generative AI, a lot of people, of course, tout some other early winners, some other sectors of the economy such as marketing, media, et cetera, being the ones that are going to be prime to get disrupted. But I do believe that actually industries such as retail that aren't under-innovated and underutilized can really see this, the slingshot effects of this breakthrough moment in AI because of all the reasons that I said.

Bill Thorne: Quinn. So that's the name of the product. So how does that work in a retail setting?

Nasrin Mostafazadeh: The very first use case of our platform, as I mentioned, is actually Quinn shopping AI. Quinn shopping AI is basically this very holistic AI that is going to be your trusted source of information, your trusted advisor, your trusted ally, wherever you go as a shopper. And it can help answer any and every question that you have throughout your journey, whether online, whether in a store, or wherever else you end up being at. By virtue of being holistic is basically this intelligent being that is an expert in anything that is in domain. So, Quinn, in the shopping AI, a use case that it has is an expert in health and nutrition for the ones that are shopping in grocery domain. It's an expert in recipes, again, for the ones that shop in, for example, in grocery retail domain. It's an expert in product assortment. So it has this deep semantic understanding of every product that a particular store carries, and its nutritional information. Its size, dimensions, whatever it is. It also is an expert in the store, information like the frequently asked questions of that store, it's customer service and beyond. So we've built this very purposeful AI in domain that has this deep domain expertise that then, of course, in return makes it really helpful. To give you an example, in the physical retail spaces that we have Quinn enabled, you can walk in as a shopper, you'll see these signages for, "Hi, I'm Quinn, you're shopping AI, ask me anything." And you just scan a QR code, and then this very seamless web experience pops up on your phone, and then you start having a dialogue with Quinn.

You can ask, "Where can I find the healthiest peanut butter that costs under $5 <laugh>?" Or you can ask, "What is your return policy?" Or you can ask, "I want to cook for my kids, they’re four and five, what do you recommend I make tonight that will cost under $30 as a whole?" "I'm on a keto diet, what is a snack he would recommend me that doesn't have nuts in it?" Because let's say I'm allergic to nuts. So all these very sophisticated questions are things that are already on the minds of your average Americans, but they literally never had help.

Bill Thorne: No.

Nasrin Mostafazadeh: They were left on their own devices when they would walk into the store and they just are left with 40,000, sometimes in your average store, SKUs that they had to just navigate and make decisions as to what they can and cannot buy given their health or budget constraints or even simply put their preferences.

Bill Thorne: And that's amazing. I mean, that is simply amazing. I know that we talked about this before. I'm not a shopper, I'm a buyer. This to me is like manna from heaven where I could basically ask the question, get to the aisle, reach for the product, put it in the cart, and move on to the next thing, as opposed to going down aisles trying to find that one paste or whatever it is that I need to get for the recipe that I have. And not only that, I could change my mind about what I wanted to fix for dinner. I could say, I want shrimp scampi, and it would give me all of the ingredients and tell me where to find it.

Nasrin Mostafazadeh: Exactly. Which is exactly what it does, and it does even the shortest route in the store.

Bill Thorne: That is just so awesome because I kind of do that already. I know my store, so I know what's in the produce. I know where the dairy is, and I know what to find almost in each aisle. It takes me a while. So if I had somebody figuring it out for me, that would be really awesome. So tell me what the response has been. I mean, it's been out there. You're doing it in-store. What are the stores saying? What are the customers saying?

Nasrin Mostafazadeh: Yes, good question. There's a third angle to it. What does the store manager and store associates say, and I want to start from that because that has been the best, I would say, surprise that we got. We just talked about AI and how job displacement, in my opinion, actually should be on top of everyone's mind in terms of the threats of this breakthrough moment in AI. Of course, although we knew that Quinn is going to be that augmented intelligence. I like to say AI should stand for augmented intelligence, not artificial intelligence, although we believe that in-store shoppers would benefit from it, but also it's an augmented intelligence for the store associates. We weren't sure what kind of response we will get. Having launched now and having been in all these retail spaces for the past few months, to our surprise, the store managers and store associates are actually our biggest allies.

Turns out these people post-pandemic, first and foremost, retailers cannot find people to hire to begin with. But let's say for those that they have hired, and they have on the floors, they're never paid actually to answer people's questions and a majority of retailers. What happens is that these individuals that are already spread out super thinly, get bombarded with questions every single minute of the few hours that they're in the store, and setting aside all the operation costs that that has for retailers, when these workers cannot focus on what they're hired to do, these people are not even supposed to know the answer to a majority of the questions. The questions on the top of people are not as simple as, "Where are the bananas?" No one asked, "Quinn, where are the bananas located <laugh>?

The questions are so sophisticated, why would a store associate know what probiotics do you recommend for my kids? They don't know. Or, "What cut of meat should I buy for this particular stew that I'm cooking?" All these expert questions are the ones that store associates and store managers love to delegate to Quinn. They see it as a way of increasing their productivity and also increasing their accuracy. They are very happy by the fact that now in the stores of Quinn-enabled, there's now this help. That you can really use as the store manager and the store associate yourself. So that's been absolutely tremendous. We've had all these, believe it or not, store associates and the store managers, this doesn't apply to just grocery retail, but apparel and your general retailers as well. We have store managers and services who email us and call us, even call my co-founder and I, with feature requests.

Bill Thorne: That's awesome.

Nasrin Mostafazadeh: One of the features that we now have is that Quinn has other languages in beta. It's not as strong as English, but it is live with other languages. And it is purely because, one of the stores had a lot of international shoppers and they were like, look, the shopper just was speaking in this other language. This person walked in and there was no help. Can you actually add it? And we sure did <laugh>, which really has been such an assistive technology, which we had not even thought about. So that's the angle on the store manager, store associates.

Now switching gears to talk about the response from the shoppers who are the end consumers that are getting the fruits of it as the end goal. Customers have loved it. We have now showcased that Quinn is a very highly efficacious solution for people to get advice because they're left again on their own devices. They love the fact that they can just in 10 seconds that it takes to utter a sentence, shorten their research and discovery time from like 30 minutes, to just how long it takes to just say it out loud. Where can I find the best shampoo for dandruff that costs under certain dollars, you can ask it to cost over a certain amount of dollars. By the way, it's funny to me whenever I travel and we go talk to people who are not your average demographic. I'll actually tell you this funny story.

This is more than a year ago now. The first time that we were talking about the product that we are spinning off and letting some private audience to interact with it, we were somewhere in the Bay Area, you can imagine the neighborhoods. And then someone asked Quinn, what is the most expensive champagne you carry from one of the stores that we were grounding Quinn experience on. And this is more than a year ago, Quinn failed. Then I asked the team, what happened? And turns out because it's AI, the way it trains is by virtue of questions that it gets. Then we know if we go and collect the ground truth on the right hand side to teach it how to operate, given a certain question. Turns out Quinn that day had never seen anyone ask about the most expensive thing. <laugh>

Bill Thorne: Oh, that is awesome.

Nasrin Mostafazadeh: That was before we were in live in the market. Of course, now that if you're live in the market, we have so much more data than no one else has, and we know better. So anyways, consumers love it, and shoppers love it, and not talking about retailers. The truth is our model is B to B to C, so the retailers are actually the ultimate parties that pay for Quinn to be this helpful AI that they launch for the sake of their shoppers. So retailers going in, we knew that this is a way for really making your customer base more and more loyal because it's this way of differentiating yourself compared to the other brick-and-mortar store right across the street. Because now people love shopping here. But the bottom line for retailers is that they're super low margin.

They're all homed in on increasing sales, which we knew by the way is going to be the case. But now that we've been in the market and we've collected all the data, we have actually shown that we have increased the sales in these retailers significantly. And we have also decreased the operation cost significantly by virtue of the store associates and store workers being able to focus on their own day-to-day jobs. So retailers love that part. I wish they loved to be honest with you, the part that this is <laugh>, you know, vehicle for better customer service.

Bill Thorne: Mm-hmm. <affirmative>

Nasrin Mostafazadeh: And the way of differentiating yourself and bringing yourself to this day and age of AI and all of that. But the bottom line is sales, which luckily for the sake of shoppers, we have proven that it does tremendously increase.

Bill Thorne: I will tell you, in my own personal opinion, if I went to a store and they had Quinn and I was able to shorten my time by five minutes, it would be worth it. I would be a loyal customer because time for me, time is very precious. And I will say this too, I kind of got a kick when you said you talk to these store associates, they're always being asked. And I gave the example of being able to go in and say, "Oh, I think I'd like to get scampi tonight, fix shrimp scampi for my family." And you get the whole list of the ingredients. I could only imagine going up to a store associate and saying, "Hey, I'm gonna do some shrimp scampi tonight. Could you give me not only what I need to do that, how long I need to bake it and the volume of what I need?" And they would just basically walk away from you. So <laugh> now I have something I can ask and get the answer that I need. That's pretty awesome. So we're getting low on time, but I wanted to ask you a couple more questions. One is, what is the best piece of career advice you've ever gotten other than mine earlier before we started, which was take a vacation.

Nasrin Mostafazadeh: <laugh> Yes. The best piece of career at advice I've ever gotten. That is a tough one. Like to pick the best of it all, so I'll go with this one because it just resonated with me and I've done it every single day. This was one of my mentors at Microsoft Research way back a couple years ago. He said, I have one piece of advice for you in your career. If anyone in any organization goes out of their way being helpful to you, make sure you tell their boss. And I've been doing that through and through and you see, helpfulness is really something that has resonated with me. I have a company around it now. So I really have this scale of help, helpfulness of people, which now applies to even Quinn. Whenever really someone is being kind, honest and nice to me, I make sure to give them credit and also definitely make sure that their boss, whoever it could be, hears it. I do think that I wish our society operated that way.

Bill Thorne: I do too. It's called karma. Karma, if you do it, there you go. It'll come back to you 10 times over. So before I ask you the last question, I have one other question, which is, Verneek, how'd you get that name?

Nasrin Mostafazadeh: Oh, good question. So where do I begin? This has a lot, like the onion has a lot of layers to peel. Let me peel two layers. I told you for me and my co-founder, Omid, our lives have been applied towards making human language the new machine language so that you can talk to anything. That's the bottom line. Instead of programming languages, you use your own language. So we knew that the company has to do with language processing, language understanding. The name Verneek is basically loosely based on the part of the brain that processes language and a speech understanding called Wernicke with a W, which is the name of the guy that basically discovered that part of the brain. Like marketing people will tell you that a better brand is with V, not W. So it had to be spelled with V <laugh> that we ended up going with something that has .com, .ai on all of those available being Verneek with V E R N E E K. And the way that it's spelled is also a testament for “ver” in Latin meaning truth, and then Neek meaning nerds and geeks <laugh>.

Bill Thorne: <laugh> That's perfect.

Nasrin Mostafazadeh: So it's nerds and geeks that are seeking the truth, which has everything to do with what we stand for as that helpful, trustworthy AI company that wants to be ultimately that trusted party that you can get information from.

Bill Thorne: That's fantastic. And last thing I promise. What excites you most about the future of retail?

Nasrin Mostafazadeh: What excites me the most about the future of retail, I do believe, is the fact that everyone will remember that this whole so-called generative AI moment actually disrupted it in a tremendous way, good for the sake of consumers being the shoppers, good for the sake of retailers and the brands, all the CPG brands. I always call it the trifecta that I think is going to be really a win, win, win for everyone. It really excites me to be able to put a dent into the advances that this moment in AI will bring into the retail industry as a whole. I do believe that we are very well positioned to play that role, and it will be really an honor to get to have played that role.

Bill Thorne: That's so awesome. You know, I think about it. The future generations I'm looking at, Jen's children, for example, and others that are going to never recognize what was before, but what we're beginning to recognize what will be is going to be their reality. And that's kind of cool. I'm a little bit jealous, but I'm fine with it. It happens generationally over and over and over again. Nasrin, it has been a huge pleasure talking to you today. I really appreciate it and I've learned a tremendous amount. All of our listeners know I'm a liberal arts major from a large land-grant university in the deep south, and these kind of things usually go right over my head. You did a phenomenal job of explaining it, that even I understood it's the monkey. I'll never forget that either <laugh> and thank you all for listening to another episode of Retail Gets Real. You can find more information about this episode at retailgetsreal.com. I'm Bill Thorne, this is Retail Gets Real. Thanks again for listening. Until next time.