Juan (00:00):
We might be the last generation of the blank page because in the future, what it's going to look like is Word is going to ask you, Hey Mike, what do you want to do today? You type something and it gives you a layout, a framework. Something you're never going to see is a blank page moving forward. Your kids, my kids are not going to see that. They're just going to have something already built out, initial point. They don't have to start from scratch for any task whatsoever.
Mike (00:39):
Please compose an opening greeting for the listeners of Retail Oriented podcast bout artificial intelligence. Oh hey, welcome back to retail oriented retail fans. I'm your host, Mike Fowler. I'm the VP of retail strategy here at the Sales Factory and today we've got a really special guest, Juan Hernandez, who is a great friend, a great soccer player, RI Madrid fan. We won't hold that against him and our resident AI guy who knows everything about artist official intelligence. He's teaching our staff constantly. He's always learning about it himself, so we're really excited to have him and talk about the greatest and latest trend artificial intelligence. Juan, welcome to the show.
Juan (01:26):
Thank you Mike. Happy to be here.
Mike (01:27):
Juan, if you can kind of level set us and kind of start us off with a little bit about your background. How did you get into the seat today? Where did you start your career and what kind of got you interested in AI and got us sitting here today?
Juan (01:40):
So as you know very well, the Sales Factory boomerang is a thing. So this is my second time here at Sales Factory. So when I joined first I helped the company develop the consumer insights practice. So this time I'm doing the same thing but with AI. So when I talked to Ged about what roles can I do back in February, we quickly realized that AI was the thing that was most important right now. So I became the AI guy,
Mike (02:09):
And so you're figuring it out on the job as you go, as everybody is today because it's emerging so rapidly, we're having to learn on the job, learn where there's applications, where there's not. Can you level set us also on what are the do's and don'ts? What are the basic best practices for ai before we get into how are we actually using it? Talk to us about what can you do, what can it do, what shouldn't you do maybe.
Juan (02:40):
Yeah, I mean AI took by surprise even the guys that are in the PALM two team at Google, those guys are working geniuses. They're even surprised by how these things are working right now. So yeah, everybody's sticking by surprise. So I think the safest way to think about AI or in this case generative AI, is think about it as your assistant, as your intern, so you can ask to do some work for you. You can always ask any questions, but always be mindful of if something looks wrong, double check it. Sometimes again, if it does, if it's easy, your intern, your assistant can do it, right? If it's a little bit more complex, it's probably going to need some or seeing some fine tuning to get it right,
Mike (03:27):
What you're prompting with and what you're entering in, right? We got to be a little bit thoughtful about what we're sharing and how do we approach the information that we are putting into prompts as well. Can you talk a little bit about that as well?
Juan (03:42):
Yeah, I mean it goes back to the assistant or intern mindset. If you tell your intern, go put this ball there, we'll put it there. Now if you tell it this ball is, I dunno, a faberge glass or something, he will treat it differently. So the more context you provide to the assistant to the intern, the better he or she's going to do the work. So the same thing happens with, that's prime engineering. Just think about all the contexts you can provide and you're going to see you get better results.
Mike (04:16):
You get out what you put in, right?
Juan (04:18):
Yeah. Sometimes
Mike (04:19):
The old adage, right, crap in crap out kind of thing.
Juan (04:22):
People think write me a song and sometimes they say, well, it didn't do a good job. Okay, but do you want to rap? Do you want Jay-Z style? Do you want Eminem style? Because remember this thing is predicting is sort of guessing what you ask, so if you give it more, clues, it's going to be more accurate in finding out what you want.
Mike (04:43):
I think that kind of helps us in terms of how do we think about it and how do we begin to approach using it. Can you talk us through how Sales Factory is starting to integrate this into our work within the retail landscape of how do we help clients win and that's kind of always how we're thinking about our work with our clients. It's all about how do we help them win, how do we create more sales? How are we integrating and folding AI into that?
Juan (05:14):
And that's a challenge, right? Because AI can do anything. So really the thing is about prioritizing and figuring out what we need to do first. We really are doing two things. First is trying to figure out what off shelf AI tools we can use to become more efficient and when you're more efficient, you provide higher value to our clients. So that's the first thing. The other thing is also we're thinking about how we can use all of our knowledge about retail, about consumer insights to develop consumer behavior predicting tools. So that's a big benefit because of working, being companies, and if you think about that in those scenarios, it takes maybe a quarter to launch a survey and to get the answers back. If we are able to create these tools that can predict consumer behavior, you can potentially get answers in seconds. Are they going to be a hundred percent right? No, but the speed and being sort of right, you can outpace the competition that way. So that's our goal here.
Mike (06:13):
Right now. Well, even the greatest consumer research is not a hundred percent, so it's still the evolution and the speed is going to be really wild to see over the next few years or even months. It's happening so fast we can't even predict how quickly we'll get there. Some of our listeners are probably a little bit familiar with the consumer pulse, it's sales factories, consumer understanding surveys that we're sending out. We started this during the early days of Covid Weekly, trying to keep our finger on the pulse of what's happening with consumers, what they're thinking and what they're feeling. How are we today starting to weave AI into the pulse as well? Can you talk a little bit about that?
Juan (06:57):
Yeah, I mean I was very lucky when I got here that you guys did that and we have three years worth of data.
Mike (07:04):
There's a lot of data
Juan (07:05):
Because that's the one thing, again, going back to Gen AI, what it does is takes the data, identifies patterns, and then when you ask a question, try to figure out, okay, what of the patterns that I have been trained on are closer to what you're asking me to predict the answer to your question. So if you think about that, we have all this rich data about consumer behavior. So again, when we develop tools to predict consumer behavior, we are leveraging all of that data plus the primary research we do on our client because that will produce more accurate results.
Mike (07:41):
The more data and the better it is at predicting. And luckily we kind of got lucky on that, that we had three years of data collection that we had been doing of a litany of different subjects and everything from political stuff to health to retail things. We asked a lot of questions over the last three years, so we had a lot of data to pull from. So as we develop our integration of AI and what that's going to be able to pull out of all that data is going to be cool to see. So I think that's something for everybody. Paying attention to what Sales Factory is doing and the content that we're putting out, you're going to see some really interesting stuff coming out of that in the next year or so
Juan (08:24):
And the business jargon, that's our mode, right? Because anybody nowadays can train a model, but if you don't have the right data or if you have better data, then your model's going to perform better than the other. So that's kind of what is going to set us apart.
Mike (08:39):
Yeah, we'll say that it was all part of the plan, yeah, predicting the future, but it is all part of our at Sales Factory, our general approach to the market and understanding the consumer. That's been it since day one, so it is evolving daily in how we do that. Next, I kind of want to talk about this and how it applies to our clients and our clients' wins as well. So can we talk about a little bit about between concept and merchandising innovation, how are those cycles changing and how is AI impacting those cycles?
Juan (09:19):
Yeah, I mean we talk about speed on the consumer insights and I think it's even more dramatic on the product development because think about years ago, rapid prototyping was it just put something together quickly and test it and that still took weeks, months, whatever, right now you could go and use let's say Trend God something like cha GPT or one of these large language models and another image generator model, and you can create product concept in minutes, literally in an hour if you're very thoughtful, right? Because with you Trend God, you can see what are the trends in your category, you refine it and then you can ask the GPT to write the concept, the product concept to write the PR release, all of that, and then you can create some images and create a survey. I start testing again in a day, right? Yeah, so I mean
Mike (10:15):
It's wild from when you and I started to the ability to escalate that time table so quickly, I think back to client meetings that I've had on the manufacturer side with merchants and with retailers trying to develop new ideas and innovation for their customer base and kind of talking them through the timetable of that and getting the disappointed looks and we really need you to escalate that timetable and make it a little bit quicker and bring it to market faster. That's possible now in a thoughtful way. You don't just have to slap something together and just hope that it sticks and hope that it works. You can kind of still take the thought based approach and get something quickly, which is wild. Have you seen or are we starting to work on examples of how that's going to work with clients? Is that integrating into client work today?
Juan (11:10):
Yeah, yeah, so like I said, we're thinking about a subscription model for instance, to develop, to create category experts for our clients. Again, we'll take our consumer reports data, all the primary data we have for our clients, feed that into a model plus scraping social media manufacturers and industry websites, all of that into one model. So we create these category expert entities just to call it something and you can ask any questions you can think of and you can get asked just immediately. Then even down the road we think about multimodal solutions. Not to get too geeky about it, but something like you can upload an image, like a merchandising image, and say "what do consumers think about this?" Or "how can I make my brand stand more than this" and things like that, and the model is going to give you answers
Mike (12:01):
On retail oriented and as a retail fan myself, that's not geeky at all. It seems like the coolest thing to be able to think about just that level of feedback and speed. Again, it's unprecedented. That's no surprise to anybody since we're talking about ai, that it's unprecedented, but it's going to change the way that everything works in the retail world and in pretty quick succession that's going to happen fast. Can we talk a little bit about concept testing and how that's going to work in the real world and in ai? How do we think about AB testing and concept testing and messaging and packaging and all of that? In relation to this,
Juan (12:45):
We're working on another solution on the digital marketing area, paid media, and really what this will allow us is, think about this: When you create a campaign, you have to create the headlines, the images, and then put it all together and test all that. So with AI, you can quickly create multiple combinations and then even use AI tools to create your landing pages so you don't have to wait on developers to create all that. So again, really become more efficient and then you'll use AI again, on the backend to do the analysis so you don't have to test every possible scenario, you just test the minimum requirement and then you can find the optimal solution right away. So it's more efficient in the development setting up the test. It's more efficient in the analysis of the results and it's going to be more efficient in the actual the performance because you get to the optimal solution faster.
Mike (13:47):
Yeah. Let's talk about merchandising and from the standpoint of picking the right assortment, optimizing your assortment, coming back in because my team is in point of sale data every day. That's what we live in and we are constantly picking out the little nuances from the skew level, the store level to optimize what we're doing and see where there's weak spots that can be improved or where we're doing really well and what can be replicated from that to other stores, right? To bring all the sales up, talk us through how AI is going to integrate into that world of analyzing your assortment and your data and stuff and making recommendations that are going to make money for your merchants
Juan (14:31):
Before your team existed. I did sort of that on an ad hoc basis. Sometimes when the sales, I remember this case where a grill manufacturer had a problem with sales, sales were soft. Everything looked good on the surface. I started digging to the data and uncovered that there were issues on certain clusters, like let's say one cluster preferred smokers, and we were out of stock smokers on those stores there clusters that preferred gas grills were out of stock of, even though at the surface we didn't have inventory level inventory stuff, everything looked fine. So that probably took me two weeks to figure out. Again, right now, AI will do the analysis for you in seconds, and not only that, it will, if you feed segmentation data, experience data will tell you what kind of people live in a certain zip code so you know what the stores are, what is the composition of people that live around those stores, and again, you can take social media data, website data, scrape all of that, put all of it into an analysis, and it's going to give you many, many, many more dimensions to use as solutions that me working by myself in a matter of seconds that me by myself working for two weeks on it.
(15:49):
So I mean this is going to be transformational. I don't think anybody knows how it's going to look like, but it's going to be incredible.
Mike (15:59):
It's all positive, right? It's all aspirational and exciting and everybody's gets, I think everybody gets very excited about the efficiencies that are potentially there and the speed and the innovation that's going to come from this, but we still have to especially now be very thoughtful in how we use it and where we use it and how we integrate it. Can you talk through a little bit of what's the human role moving forward now and then what do you see in the future?
Juan (16:30):
There are studies out there that have proven over and over again that the best solutions, the best performance is when the model works with oversight of a human. It's not like you type an answer, you get the answer, you take it, no, you take it, you analyze it, you figure out maybe after you saw that first solution, you think about, okay, maybe we provide this context, I get a better result. So humans always can make the output better, there's no question about it. There's not going to be one case where this machine alone can outperform a human. I mean, if they do it individually, yes, but if we're together, the result is going to be better. That's something that for me is going to not going to change, and I heard the other day about pro,pt engineering is going to be like having Excel skills today. It's something that everybody in the office is going to use it at some point. So that's going to be like you must have skill.
Mike (17:32):
I think I emailed you the other day talking about prompt engineering and just kind basics for old guys like me. I remember when social media was just starting. I remember MySpace and the beginnings of social media and I a young man when I was figuring all that stuff out. I'm not as young a man anymore and figuring all these things out today. So I think it's good for us all to keep in mind there is a learning curve here. It is not as complex as it may seem. Starting to interact with it is a great way to learn it. So it's one of those things just like a lot of things where you get in, you start doing and you figure out more and more and you improve and your interaction with it becomes better, and as a result your outputs and what you're getting become better as well. If there was one thing that you would share with all of our listeners, both merchants and retailers as well as manufacturers and vendors to those retailers, if you were to give them one piece of advice of do this or think about this in terms of how you're going to market in the retail landscape today with e-commerce as well as brick and mortar, what would you tell them to be thinking about in terms of how to use and how to utilize AI?
Juan (18:52):
Don't expect generative AI to solve your problems. One thing that I always talk about is the double diamond framework. So you've got a problem. You can use AI to refine your problem statement or peel the onion, if you will, to get to the root cause. Then once you do that, you can use again AI to help you branch out and uncover all the possible solutions because that is going to help you find the best solution. That's why it's kind of called the double diamond because you start with one problem, you brainstorm all the possible problems, find your problem statement or refine your problem statement, come back to one issue. Then you kind of use it to expand to kind of brainstorm with the gen AI tool to find the best solution, and then you come up with that so unique solution.
Mike (19:49):
It's just going to broaden our capacities for thought on problems and solutions on both sides of the double diamond, as you say, it's going to help us to add more context, add more different possibilities for ways to tackle problems. That's a great advice.
Juan (20:10):
One more thing I want to add on that is we might be the last generation of the blank page because the future, what it's going to look like is Word is going to ask you, Hey Mike, what do you want to do today? You type something and it gives you a layout, a framework, something that you start with. So you're never going to see a blank page moving forward. Your kids, my kids are not going to see that. They're just going to have something already built out, and initial point. They don't have to start from scratch for any task whatsoever.
Mike (20:45):
That's interesting to think about as humans and our constant drive to create and innovate and conquer and having that piece change, at least change what it looks like is going to be interesting in how future generations adapt to that and what that means. I think that's a good one. That's a good statement to end on, but I never let any guests go without some hot seat questions. So hot seat questions are super simple, right? I'm just going to ask you a question and I want you to just quick fire, give me the first thing that pops into your head. Don't think about it. Just what intuitively comes out. Alright, so first question is what was the first computer game that you couldn't get enough of as a kid?
Juan (21:28):
I don't remember the name of it, but it was like a car chasing game where the police caught up to you. I don't remember the name of it, but it was like monochromatic. I just played for hours, even though it was the same thing over and over
Mike (21:44):
Again. Can't get wrong with a car chase ever. What podcasts are you loving right now? What are you listening to? What do you love?
Juan (21:51):
Revisionist history from Malcolm Gladwell. It's my go-to, I love it. Because of my role now, there is, there's one called this week in AI, and this is a crazy thing. The average I would say is like 90 minutes long per episode, but those guys actually have to rush to get all the content they want in. This tells you the pace of change on Ai. Most podcasts, 20 minutes are even too long. This one is 90 and still they cannot put everything in.
Mike (22:29):
So you've got history podcasts and what's happening today and in the future podcast. You're pretty well-rounded on what you're listening to. Last one, and this one is an important one to me. I love a sense of humor. Do you think AI will ever develop a sense of humor?
Juan (22:44):
I mean, to some extent it already has, right? If you ask, tell me a joke in Seinfeld style, it will tell you something.
Mike (22:52):
Oh my gosh. Wow.
Juan (22:54):
I mean, it might not be original, but I think it would be funny. I bet
Mike (22:58):
Could be original to you and that's all that matters. If you haven't heard it before, then it's original to you. Yeah. Juan, thanks so much for being with us. We certainly learned a lot today. We are excited to see what's coming out, so keep an eye on things that Sales Factory is putting out with and around the topic of ai. I think it's going to be exciting, especially for retail focused people and people that are interested in what's happening in the retail landscape to understand how that's integrating in because it's going to be a part of our lives, right?
Juan (23:29):
Yeah. I mean it is already in, yeah. Yeah.
Mike (23:31):
Well, thanks a lot for coming on and we appreciate you. See you next time. Yeah,
Juan (23:36):
Thanks for having me.
Mike (23:38):
So there you have it. If you're finding your way in the retail world, trying to figure out how do you integrate AI into your, from merchandising to product assortment decisions to product development, we hope you learned something from this episode. Juan had some great nuggets to take away there. If you got any questions, we'd love to engage with you and have a conversation about it. So drop a comment in the comment section and we would be happy to reply. Also, if there's any other topics that you're especially interested in hearing about on retail oriented or guests you'd like us to have on, drop a comment in the comment section, or you can email me directly at Mike.Fowler@salesfactory.com and we would love it if you subscribe to the channel so you're always in the know and up to speed on all the content that we're putting out. Click the like button and don't forget to hit the little bell icon so that you always know when a new episode drops. And remember when you're thinking about retail and thinking about the retail channel, it's always about selling in and selling through.