What does it take to get AI right in retail? It’s not one smart model. It’s identity, inventory, and behavior data working together. Dr. Serena Huang talks with Veronika Durgin, VP of Data at Exemplar Luxury Group, about why pointing AI at everything exposes the tech debt humans learned to ignore, how to tell real AI work from AI theater, and the one thing she refuses to outsource: thinking.
Point AI at everything you own and it will find everything, including the outdated draft policy buried in SharePoint and the “final-final” table nobody ever removed. Dr. Serena Huang sits down with Veronika Durgin, VP of Data at Exemplar Luxury Group, who has spent two decades building the data foundations behind retail personalization. They unpack what actually powers those uncannily good recommendations, why a chatbot isn’t an agent, how AI surfaces the tacit knowledge teams never wrote down, and why the right first move for a new AI leader is old-fashioned prioritization instead of boiling the ocean. Plus Veronika’s dream for where retail goes next, and why the human part of shopping should stay human.
Timestamps:
0:48 - Meet Veronika
1:27 - How every vendor already knows what you want
3:09 - The machine learning we forgot about
4:19 - Why recommendations got so much better
5:17 - No free lunch: your data as barter
6:47 - Shopping assistants vs. actual agents
9:38 - Why everything with GenAI gets called an agent
11:09 - What AI should and shouldn't take off our plates
12:13 - The one thing Veronika won't outsource
13:33 - Drawing the solution by hand first
15:13 - AI exposes every imperfection in your platform
18:56 - Advice for the new AI leader: don't boil the ocean
20:51 - Cutting through the AI theater
25:11 - What data quality really means in retail
28:00 - Using AI as your QA assistant
29:40 - Personalization and the emotional side of shopping
31:26 - Where retail is headed
Links:
[00:00:00] Veronika: Not every business problem needs an AI solution. Just because you can doesn't mean you should, because it's not cheap either. But identify what the most valuable problem to solve is. It's just a simple prioritization that we've always gone through
[00:00:23] Serena: What does it take to get AI exactly right in retail? It's not just one smart model. It's a mix of identity, inventory, and behavior data that helps you know what the customer wants My guest, Veronika Durgan, is the VP of Data at Exemplar Luxury Group, who has spent two decades building that foundation.
Veronika, welcome to "Deep Geeks."
[00:00:48] Veronika: Super excited to hang out with you, Serena. You know I always do, uh, whether we do it, uh, digitally or in person.
[00:00:55] Serena: It's been so fantastic to watch you speak and watch your career, and I can't believe I get to do this today. So let's just get started. Now, I, um, I will be honest, I do a lot of browsing, um, on my phone at night.
And sometimes, this has happened for, for years, but it seems like recently things have gotten even more accurate. What makes it so good these days especially? How does it know? Magic. It's magic-
[00:01:27] Veronika: Magic ... Serena. Uh, no, I'm just kidding. Before we dive in, I just wanted to do a little discla- disclaimer that my opinions are my own and don't represent my current or former employers.
But with that is every vendor, every platform is sort of monitoring your behavior. So there is clickstream, every website you're clicking on, every advertisement, probably with Instagram, if you pause and spend longer amount of time, all of that is tracked. It, it's nothing new. I mean, everyone's watching, and that's sort of how every vendor is trying to market to you.
But ultimately, every vendor, every company is trying to provide you with services or products. So it's, it's not new, it's not dark magic, just, just know it's happening. And as this data then is collected, and for the longest time, for years and years and years, machine learning was applied to it, right? If you're searching and looking and pausing at specific things, we kind of identify the pattern.
That's what right machine learning is great for, and we try to predict what we think you will be likely to buy. We kind of try to figure out what you like, what you don't like. And this is generally, um, what I would say some of it is first-party data, data we collect about you, especially if you're logged in.
Um, if you browse specific websites and you are, um, like you signed up for those, right? You logged in, so we actually know who you are. So we're kind of collecting all of that information, again, with the goal is to present you with something that you absolutely must have.
[00:03:09] Serena: Yeah. And I think you brought up a good point.
Uh, maybe we go in a little bit deeper here 'cause we hear so much about AI as in GenAI, right? Right. But ML, machine learning, is a subset of AI that we seem to have forgotten that it has existed for a very long time.
[00:03:29] Veronika: No.
[00:03:29] Serena: And, um, I do feel like the algorithms have gotten better over time. Like I remember even, you know, three to five years ago when I said, when I get recommended things that you might like, they were not things I might like at all.
But today it's gotten so much better, like the dresses are really similar. Like, "Hey, this black dress is out of stock. Here's another black dress with a little bit more pearls or something that, um, might be really sparkly on you. Who knows?" My own experience is that the recommendations are getting better, that personalization and including the items that I might like drastically improved compared to three to five years ago.
And I'm curious if it's the algorithm or the quality of data or something else that is driving that.
[00:04:19] Veronika: Such a great question 'cause, um, I am also very much a shopper, and I get annoyed when I get recommendations that, um, are not what I like, but at the same time I don't necessarily want to share my data. So it's a great question.
I think Serena, it's everything. I think the technology improved drastically. We have a lot more power To process a lot more data, and we can also store a lot more data, so we can actually collect a lot more data. So I think it's really everything. And, and I would say the question of how much data you're actually sharing, such an interesting one.
If you want the best recommendations, if you want for your favorite vendors to, to give you the best possible recommendations to truly know you, it is actually in your interest to provide whatever data you feel is, is necessary, your preferences, your likes, your dislikes.
[00:05:17] Serena: I remember years ago when I would teach data classes to, you know, whether it's college students or MBA students, I, um, people are often shocked that they are not getting the coupons for free.
That when they sign up, even the grocery store, that email address, right? That is, that is sold. That's right. That is not free. So there's like this economic concept of there's no free lunch and, and there's truly no, yeah, no, no free lunch in this
[00:05:47] Veronika: sense. Well, it's, it's bartering, right? Like you don't necessarily...
When, when people talk about data monetization, they think about like you're selling it for money. You're not. Right. You're exchanging- Yes ... your data for something else. Yeah. And I think it's, it's totally okay. It's economics. Mm-hmm. It's business. Yeah. So if you want something in return, don't give it up for free, is, is all I can tell you.
Yes.
[00:06:08] Serena: Right. If it's worth it to you, yes- Yeah ... absolutely do it. Now, the other experience that I've had recently when shopping online is a little agent might pop up, right? It might ask, "Hey, I see you looking at this, you know, pair of shoes. Can I help you with it? It's currently available at this price." Can you tell us a little bit about just the advancement in that type of s- I'll call it, you know, shopper assistant maybe, and then talk about agents, um, a l- a little bit there.
Yeah.
[00:06:47] Veronika: What, what you're saying is, is this, there's a piece of functionality that tracks you real time. So your clicks, right? Like it's clickstream data, so somebody wrote a piece of functionality that's sort of reading that data real time and is having conversation with you. So there's, again, like I'm, I'm guessing, but there's like sort of GenAI component sitting on top of clickstream data.
But an agent to me, which is why it's, it's really fascinating, it's an autonomous system where you give it a goal and it just does whatever by itself without any guidance from you. For example, I want to fly to Chicago to visit you, Serena, in two weeks, and I want the cheapest tickets. And then all of a sudden my, my trip is booked.
So this automation, this process, this piece of software has all of my information, all of my preferences. It has access to all of the websites' logins. It can decide what the cheapest ticket is and actually buy it.
[00:07:56] Serena: Mm-hmm.
[00:07:56] Veronika: Good. So-
[00:07:56] Serena: Without you being involved ...
[00:07:58] Veronika: without me being involved. Okay. So it's not the same as a very helpful chatbot-
[00:08:03] Serena: Yeah
[00:08:04] Veronika: that can ask questions and search on steroids, right? Yeah. And bring information- Mm ... and interact with you. Those are helpful, and I don't know what we wanna call them, but to me- Yeah ... agent is, is a little bit more than that. Was it Claude or OpenAI where their agents just went off and hacked Hug & Face? Like-
[00:08:23] Serena: Yes.
Yeah, OpenAI. The--
[00:08:24] Veronika: Yeah. Yes. There was a goal given to this automation, and it did what it had to, uh, to, to accomplish that goal or it got stuck, right? Like, there's kinda like the, it's a binary debt that you either succeed or you fail. So I don't know how many companies are there yet. I also don't know how many companies are comfortable getting there.
Or us as humans, would you be comfortable with sending automation to purchase I don't know, but make a big purchase for you?
[00:08:57] Serena: Maybe not yet.
[00:08:58] Veronika: Maybe not yet. Maybe not
[00:08:59] Serena: yet. Yeah. Um, it's interesting. Okay, so you're saying that the definition, true definition of an agent has to be able to autonomously accomplish the goal, uh, given constraints and whatnot- Right
navigate it on its own without humans.
[00:09:18] Veronika: Correct. Yeah.
[00:09:19] Serena: And then, uh, so agree there, and then you said the chatbots are not agents. However, I think maybe this is just poor marketing, right? Or poor branding, 'cause we've been calling these type of, you know, customer service agents that are assisted by AI as agents.
[00:09:38] Veronika: We essentially, I think, and again, kind of maybe poor marketing, but we call anything that has gen AI in it an agent. Like, it seems like where we're all going, and maybe at some point, I mean, if this is how we communicate, it's fine. But to me, I think there is a difference between an assistant that is sort of helping you along the way and have a conversation with you, and something, an, an automation that truly all you have to do is give it a goal and it accomplish it without any intervention.
And the perfect agent, it can talk to other agents. Like, it does whatever it has to do to accomplish a task you gave it. I don't know, maybe chatbots are, like, very simple agents. Uh, I'm like- Right ... I'm willing to, uh, sacrifice- Okay. Okay ... but I usu- but I usually, um... Maturity of AI, when I, I talk to people and, and listen and everybody like, "Oh, we have a million agents," I think there is a different level of maturity based on the definition of an agent, right?
So you kinda like, everyone has chatbots right now, everybody's using gen AI for their personal life or, you know, search, for, for whatever it may be. But the level of truly autonomous agents, I don't think many are there yet. But I think that's where the real story is. Again, like, it w- maybe it'll get a little philosophical, but is the goal of humanity to completely outsource everything to AI?
I don't think so.
[00:11:07] Serena: Ooh.
[00:11:08] Veronika: Right? I
[00:11:08] Serena: hope not.
[00:11:09] Veronika: I, I, I mean, I, I- I truly hope not ... I, I'm like, WALL-E comes to mind. I don't wanna be that, right? Like, I, I think the goal of AI is to supplement and help humans, and for humans to outsource either dangerous work or low cognitive work So that we can spend time on things that we are g- that our brains are good at, which is right, like, like this high cognitive cause-effect creativity.
So, so that's why I'm like, I don't know. I, I, I'm-
[00:11:42] Serena: Yeah ...
[00:11:42] Veronika: I'm still a little idealistic about hu-humanity for the win.
[00:11:47] Serena: No, no, I think, I think that's, that's very realistic, and I hope our listeners catch on as well. I, I'm curious, as you think about, as you see the AI advancing, and seemingly almost on a weekly basis there's a new model, what do you personally choose not to outsource?
Because as you mentioned, there are things that maybe we want to do as humans forever, and we don't outsource that.
[00:12:13] Veronika: What I don't wanna outsource is thinking. So wh- which is an interesting thing. So I would still... And w- like if you think of like within data, what involves thinking, right? It's understanding a problem, designing a solution Troubleshooting.
Troubleshooting involves a lot of thinking. So now once you understand the problem and design a solution, then you can feed it into your favorite local, you know, LLM, and then it'll write code. But even then, still like, to me like, like you, you always kind of want to understand what that code is doing. So, and we get lazy 'cause like, "Oh, it just works, we'll push it out."
But I actually still wanna read through it and understand what it's doing, and if I discover something new, I want to learn what it means. Being able to think logically through a thing is, is like if I lose that, like I-- that's the worst thing that can happen to me.
[00:13:18] Serena: Yeah. Are there guardrails that you started to create just for yourself, like certain activities that you would never use AI for?
How do you make sure you don't outsource your thinking?
[00:13:33] Veronika: I, I still very much draw diagrams by hand. I think in pictures, I understand pictures better. So to me, designing a solution is actually drawing, drawing it out. It's like it's a diagram. Um, and that helps me kind of logically break it down, sort through it, understand, and then I can like feed it into whatever and like ask it to write words.
Um, which is okay. Like I, I'm very accepting of, uh, documentation written by robots. I'm like, as long as it makes sense, I don't really care. To me, I think again, kind of like that simple diagram of a business flow or an architectural design is, is kind of where... A, again, this is what I enjoy doing, so I'm like, I'm totally biased.
Uh, but that's where I'm at.
[00:14:23] Serena: Yeah. I'm picturing you drawing diagrams and maybe, I don't know, data lakes. Um-
[00:14:31] Veronika: Lake
[00:14:31] Serena: houses.
[00:14:32] Veronika: Yeah. Lake houses
[00:14:32] Serena: I'll be drawing. Warehouses and lake houses, yes. Oh, oh man, there are so, so many jokes about the different bodies of water that... And houses that we're building on top of the water.
But back to the bots for a second here. I think I, I'm seeing, you know, as someone who used to draw dashboards and design them by hand as well before making them in Tableau and whatnot, I've started to see a lot of vendors sell chatbots on their solutions. What are your thoughts on the maturity of that?
Um, have you personally had any experience?
[00:15:13] Veronika: So you have to really, really scrutinize it. You have to really, really know, like know the domain really well to really kinda understand the data you're looking at is correct. The other interesting observation I have, and, and maybe I'll write a blog about it, actually I think I will.
Uh, I've worked with enough platforms, data platforms that I thought were designed really well, like thoughtful architecture, great engineering team, people with deep knowledge. And then you put AI on top of it, and then you quickly realize how imperfect your data platform is. Inconsistent naming conventions, beta, final-final Tables that should have been removed that are still there.
Like all your typical sort of, uh, skeletons that humans know, right? 'Cause kinda like anyone who's working with data picks up on this and they... Like, our brains very easily kinda resolve that. They're like, "Oh, this table's bad, this table's good. I'm gonna ignore that one." AI doesn't know that. AI has access to it.
How would it know? So like, what, what, what I find incredibly amusing is my, my level of perfect, great platform all of a sudden is like, ooh, all of the tech debt, everything that we kinda... Again, it didn't bother humans. Like, your BI was re-pointing to the right set of tables. Anybody who's writing queries knows, right?
Like this tacit knowledge that we never documented, we all know, we all learn, we resolve it, doesn't bother us. But then you put AI on it, and it's like all of a sudden it's not so perfect anymore. But I think that's okay, because we're also always taught humans have to, we have to document everything. It's impossible.
It, it's such an unreasonable expectation because our companies are not static. They constantly change. Knowledge changes. We, we talk to each other, we make decisions, you know, in meetings. We don't write them down. We update three pieces of information, not the fourth one. So it's, it's kinda like humans are imperfect and we, our brains kinda are okay with it, but AI not so much.
[00:17:29] Serena: That's such a good, um, and really clear explanation of how AI can actually expose what's broken or, you know, whether it's your tech debt or something else, um, or decisions made a long time ago that were okay but everyone learned to ignore, it suddenly shows up. It's like AI is the kid in the candy store and just sees everything.
It's like, "Oh my goodness, I have access to all of this and I'm going to use all of them." But the human can discern and choose not to do that.
[00:18:05] Veronika: Correct.
[00:18:05] Serena: But AI cannot.
[00:18:07] Veronika: And then the question is then, do we really need for AI to have access to everything? 'Cause remember, we had this rule forever, don't boil the ocean.
Focus on one thing at a time, define it well, scope it. I mean, that's I think where kinda like From, you know, relational structured data, AI needs a semantic layer. And you're not gonna model all of your data in semantic layer. You're gonna model data that helps you answer specific questions and then kinda like incrementally add to it.
So why did we all all of a sudden decide that AI is so magical, we need to give it absolutely everything? And then it, it's, it's gonna give you answers based on the imperfections that we generated over, you know, decades of, uh, tech.
[00:18:56] Serena: I'm curious if you have suggestions, something concrete and tangible for a new AI leader who maybe inherited a lot of data and platforms they didn't really ask for, didn't fully understand, and now the leadership team is pushing hard on go all in on AI.
What's your recommendation for them so they don't boil the ocean?
[00:19:18] Veronika: Right. I think a lot of companies, I hope every company is using AI in engineering as far as, um, just helping engineers sort of move faster. We Googled answers, how to write code. Now AI kinda does it quickly. So there's like this personal improvements, efficiency improvements, and I hope every engineer has access to that, right?
Your copilots, Claude Code, et cetera. But as far as using AI to solve a business problem, that's the tricky one. And I think specifically for that, again, not every business problem needs an AI solution And the la- like the thing is like just because you can't doesn't mean you should because it's not cheap either.
But identify what the most either valuable problem to solve is. Maybe you solved this problem in the past, but it's expensive and it's too long. S- resolve it with AI. Like it's just a simple prioritization that we've always gone through. The most spectacular, if you have to like do something beautiful and shiny and maybe it's easy to do, but it kinda gets the buy-in from some business leaders that are skeptical about AI.
It's like just prioritize a problem based on various things. Um, maybe it's the quickest one you can do. So I don't have anything different for AI from any other sort of prioritization. I think it's just the same prioritization as always.
[00:20:51] Serena: Yeah, that's fair. Prioritize what's going to be most important for the business, right?
Whatever outcome you're trying to optimize, it's a, it's the same thing and start there. I think there's just so much hype and excitement right now and the fact that there are now bots everywhere and every event has turned into an AI event- Mm-hmm ... says a lot. And I have written about this before, the AI theater that is out there where both companies and employees feel like they need to show up differently as they're super excited about AI.
This is the next thing. And frankly, not everyone necessarily feels that way and not every problem, to your point, requires AI. How do you decide or figure out what is real and what is not?
[00:21:46] Veronika: Yeah, that's, that's a tough one and it's, it's so true. I remember I was at a conference. I was speaking about data modeling.
I, I had barely anybody in the session Like, who cares? There's all of these AI sessions that are like AI, AI, AI this, that, and the other thing. What I think you, you realize is, so there's obviously, you know, vendor presentations. We all know AI didn't change that, is there's a lot of exaggeration happening during those presentations.
There are also presentations from various companies where they're trying to present in the best possible way, which is also slight exaggeration, so which is okay, we all know this.
[00:22:31] Serena: We
[00:22:31] Veronika: understand. I-- We un- we understand that. You're not gonna go on stage and, and tell everybody how much, you know, you f- like how many times you failed or how like- Right
nobody, nobody is interested in that and- Yeah ... nor do you want to kinda present your way in a very sad sort of way. Um- Right. That's it. I think to me, hallway conversations are always a lot more interesting and real, kinda like tight round table discussions. A- and again, like the, the other thing in is, is incentives.
Understand, uh, whoever's on stage or whoever is, what, what are their incentives? Look, we're in the capitalistic economy. The goal of everyone is to make money. So if you are speaking with a vendor, their goal is to present themselves in the best possible way so that you buy their service or software.
Nothing's wrong with that. If you have a person on stage, their incentive is to present themselves in the best possible light. So understand that, that that is totally okay. When you're in a small group, the incentive is to just share information and learn from each other, so you get sort of different story, different perspective.
So kinda like just understand where you are and what's driving that specific conversation. And then again, apply your critical thinking to sort through that, of the messages that you're getting
[00:23:56] Serena: And let's talk about data for a little bit. I know you've spent your career seeing, working, building with data, and I personally know a lot about the power of having good data and clean data.
I, I remember a company was trying to do something so simple as building a chatbot that will answer policy questions. Mm-hmm. And they compile all the policies. It was all PDF files going back all the way to when. They looked at it, cleaned it up, and was like, "Great, the bot is ready to go." Who knew the bot had somehow found on SharePoint somewhere a very old, outdated draft policy and started spitting out very incorrect information.
[00:24:48] Veronika: Mm-hmm.
[00:24:48] Serena: And, and like I think about that a lot, and it takes AI to surface that bad data- Mm-hmm ... um, as we said earlier. But I'm, I'm curious in the retail space, what does that actually mean? Can you make it real for us, like the quality of data and what it means to shoppers and what it means to the companies?
[00:25:11] Veronika: I think data quality is really sort of the same across the board. It doesn't matter retail or not. Basically, data is generated either by applications or devices, right, to support capabilities of those applications and devices. So as long as that data that's needed for that specific piece of functionality is there, right, you're great.
So what we do in data analytics space is we take all of the data and we try to put it together to kind of see your customer 360, your, uh, order three s- all of your kind of like three sixties, and that's when data quality issues start popping up. "Oh wait, I can't actually join this data set with their data set," or there's actually a lot of either like, I don't know, nulls or questionable values, or this field was used for something and now it's used for something else.
So the actual semantic meaning of that data in that field is different depending on the day. Um, so this is kind of what we're running into, and this is what AI is also easily kind of like showing us. That's why I don't think that just connecting AI to everything, to all of your systems will actually help you at all.
So we need to sort of put harnesses around our AI. This is where this context semantic comes through. Yeah. You only have access to this- Mm-hmm ... whatever that is. This is your box. You're staying there. Right. And we defined what's in that box.
[00:26:45] Serena: Yeah.
[00:26:45] Veronika: So in your example, unfortunately, your box was, "Here's our entire SharePoint, not just this folder," right?
[00:26:54] Serena: Mm-hmm.
[00:26:54] Veronika: And then of course it found everything. So it's, it's a very, very wide data set. So there are a lot of different systems supporting different functionalities of the business. So that's where I think is, is kind of like tricky. You know, remember the three big Vs of big data: variety, velocity, volume.
So I think variety is what's prevalent specifically for retail, again, because it's just so many different things that happen in different systems and different functionalities. So that becomes like y- your, your breadth of like, and each system has its own data quality issues, right? So kind of like reconciling all of that, uh, becomes a little challenging.
[00:27:40] Serena: Never really thought about that. I think that's such a, uh, interesting point that you have all these, well, potential data silos, but you- Yeah ... you're connecting them obviously. Do you see what retailers can do now because of AI that they maybe couldn't five, 10 years ago?
[00:28:00] Veronika: I think it's almost like AI could be your QA, right?
[00:28:04] Serena: Oh, okay.
[00:28:04] Veronika: Like, you kind of think about it like, you don't want to un- like I just said, I don't want to unleash AI onto all of my data everywhere and just start asking questions- Mm-hmm ... and make decisions with that. No. Right. But what I can do is unleash AI and say, "Reconcile these two things."
[00:28:20] Serena: Mm. "
[00:28:21] Veronika: Tell me what's, what's not working.
Help me identify all of the naming inconsistencies. Help me identify all this, like, weirdness. Help me identify things that don't make sense. Like, what is this field? What does it mean? What's in it?" Like, kind of like have your own personal QA assistant, QC assistant to just bring this to light and then decide what is worth fixing and what can actually be fixed, because not everything can be fixed, right?
Not everything will be prioritized to fix because, again, businesses have problems to solve and not every data quality makes it above the line.
[00:29:00] Serena: I like that. And plus, that is usually very tedious work- Yeah ... for humans to do and-
[00:29:06] Veronika: Well, very boring, yeah ...
[00:29:07] Serena: yeah, very boring. So, but, but that's a good use of AI, for AI to do something that it's good at and that's boring for humans to do.
[00:29:16] Veronika: I would also say, actually, I shouldn't say boring. Data, any good data analyst actually thrives on this work. Mm-hmm. But any good data analyst tends to go into a rabbit hole that just takes a lot of time and might not matter. Yeah. So to me, maybe AI can do the first pass and look at all rabbit holes- Mm-hmm
and then you kind of prioritize what is actually worth paying attention to and, and trying to solve. Oh.
[00:29:40] Serena: Okay. Interesting. And on, uh, personalization front, do you see how AI is really helping with personalization in different ways than maybe even traditional ML?
[00:29:57] Veronika: I think for me personally, I was googling something and this Google AI assistant, what I loved, just loved about it is it was finding reviews.
So I'm not a very tall person. I always have hard time reading, like finding reviews on specific things, and it was just finding them for me. So for me, that was like, it's not necessarily personalization, but it was so helpful. It's also like it's emotional. So you walk on the street and somebody says, "Hey, this looks amazing on you."
Made your day, right? Imagine you say AI gave you a recommendation for, to try something, you're in the store, you put it on, you came out, and the store associate's like, "Oh my God, this is you." You maybe you weren't sure, maybe it's not your typical thing, but somebody just said, "Oh, girl, you gotta get it." One, it made you happy.
Two, you just got something new that you normally wouldn't. So to me, again, like this is just art and personal and it, there's, it's emotional and I kinda, I want it to stay
[00:31:05] Serena: human.
[00:31:06] Veronika: Yeah.
[00:31:06] Serena: Yeah, I like that. No, we found another thing that we're not going to outsource, so
[00:31:11] Veronika: No, no, absolutely not. The two of us will go shopping together
[00:31:15] Serena: forever.
Yes, we, we made a plan. As we close out, um, where do you think retail is headed now that AI is everywhere?
[00:31:26] Veronika: It's a total, like total dream. I am hoping that what will happen with retail is it'll be a little bit more like bridal stores, where you go in person and maybe there is some personalization if it's kinda like a brand store that you've always been at, they know who you are, they know what's in your closet already.
That location has just samples that you can try on, and there is a person who, again, that stylist that actually is really awesome at their job, that can kinda help you navigate all of these options. And then once you choose what you want, it's shipped to you.
[00:32:07] Serena: Ooh, okay.
[00:32:08] Veronika: Because I think there's still desire to touch and feel and try on things.
[00:32:14] Serena: Yes. Yeah.
[00:32:15] Veronika: And there's also desire for that human connection and, and exchange because again, there's like-
[00:32:20] Serena: Yeah ...
[00:32:20] Veronika: there is, there's emotions involved and we, we're emo- emotional creatures. So, um, this is my, not prediction, it's a dream. And I hope there's like more of these like small locations where you can actually go- Yeah
and, and try it on and have a conversation.
[00:32:36] Serena: I like that. I like that a lot 'cause, uh, I, I agree there's the magic of trying on something-
[00:32:42] Veronika: Mm ...
[00:32:42] Serena: in person, but then that also cuts down the inventory if you have to hold. Right. And the- Exactly ... square footage and the electricity you need to power the store and-
Right ... all of that. Exactly. Um, and still keep the human good at styling people and having those human conversations, um, employed in the job that they love. So I like it. So
[00:33:04] Veronika: I don't know. We'll see, we'll see what happens, but this is kinda just an idea that I have.
[00:33:09] Serena: Yeah. Yeah. Well, when, you know, Deep Geeks hit, I don't know, episode 200, we'll invite you back and we can- Yeah.
check back in, okay? So thank you for joining us.
[00:33:19] Veronika: Thank you for having me, Serena. It's a date. We'll go shopping.
[00:33:23] Serena: All right. Let's do it. Thanks for listening to Deep Geeks. A huge thank you to my guest today, Veronika Durgin. If today's episode made you think differently about data and how AI gets built, share it with someone who needs to hear it.
Find Deep Geeks on Spotify, YouTube, or wherever you get your podcasts. Until next time