AI In Times of Crisis [Podcast]

Daisy Founder and CEO Gary Saarenvirta joins the Innovation Calling podcast to discuss how AI can help in times of crisis, and how companies can use data during the COVID-19 pandemic. This episode also covers what innovation means in the retail space, and how companies can make changes now that will set them up for a long term future.

Listen to the Innovation Calling Podcast:

For the official podcast breakdown, visit the Innovation Calling website.

Innovation Calling Podcast Transcription

Erin:
All right, we've got another great episode of innovation calling your way, but before we do, we've got to address the giant elephant in the room, which I guess is coronavirus and some things have changed. We're so diligently promoting events for GLO, the Global Leaders Organization, for our live podcast recording, and I'm sure it comes to no surprise that we've had to do some postponing, and some shifting. So, all of our live podcasting events will be postponed. However, Syya I do want to talk about what we're doing, which I think is so unique and so fun for GLO.

Syya:
Yes. So, we've gone to a full digital platform. Yes, GLO already had a digital presence for our global audience, but now we're taking the local chapters and digitizing that as well. So, when we are now hosting our weekly events, we're going to actually have breakout rooms, Erin. And so, there will be a Dallas, a dedicated Dallas chapter room.

Erin:
Yeah, we have, in those weekly events, I mean, we have an incredible lineup of speakers coming. You're definitely gonna want to just head over to innovationcalling.com as we announce these weekly, you're going to be able to register. Typically, these events are going to be closed to non-members. If you're a member or a first-time guest you can come, but we're opening these to everyone. However, with the, you know, insider meetings, that Syya just mentioned, those will be for members only. So, we know this is a time of stress for many of you, for concern of many of, for many of you, and we really want to help to bring a community together, provide great content, help you to get the help you need right now, and that's what the content is going to be for, and then specifically in these breakout rooms. Syya anything else you want to add?

Syya:
Yeah, I do really want to emphasize, you guys, Global Leaders Organization is a business networking organization, it's not just a meetup group, and now more than ever, because we are all business owners and leaders, and I'm certain all of us are going to have some challenges with our business through coronavirus. We've worked through this pandemic. We really need to be there to support each other, and that's why we're really emphasizing, you know, participate, um, you know, join in the dialogue, you know, give Erin and me feedback of what everyone needs, because I'm sure Dallas, even though we're the Dallas chairs, I'm sure everyone else probably can get the same feedback, and we can share that as well. So, that's my two cents about Global Leaders Organization and why we're so passionate about it.

Introducing Gary

Erin:
Yeah, so once again this is definitely changing, not changing quickly, but we're moving at a much faster pace than typical with, with weekly events. So, head over to innovationcalling.com, we'll have those posted there. You'll be able to register and that's going to be the best kind of centralized place for this information. So, all right. Speaking of coronavirus, and speaking of changing, we had a great conversation with, with Gary. Syya, who do we talk to in this particular episode?

Syya:
Yes. So, I absolutely loved our conversation with Gary Saarenvirta. He is the founder and CEO of Daisy Intelligence, basically one of the most kick booty companies that built AI platforms largely around the retail and Aerospace Industries, really known for the retail space. Gary is a preeminent authority on artificial intelligence here in North America, over 25 years’ experience working with, you know, corporations, large global to mid-tier enterprise, and has helped them reach profitable growth. You know, all their delivery for revenue targets, etcetera. So, he founded Daisy back in 2003. And I just got to tell you, Erin, I just loved his insights on how he can leverage Corona, well, leverage intelligence in the age of coronavirus pandemic.

Erin:
Yeah, I mean, how many, I mean, how many ways can a company pivot? I mean, there's just you, you are in dire, so many retail, I can't even get these words out, but so many retailers are just, they're in a lot of trouble right now, and, you know predicting, we joke about toilet paper, but, I mean, I've never in all my life walked into multiple grocery stores to see bare items, and how do you predict that? And how do you ensure, there's a shift now happening and an increase in sales, but what's going to happen when people have 50 rolls of toilet paper stocked up and now, they're not going to buy it for a while? And, you know, these grocery stores are working off very old, ancient data. So, you know, you got to innovate to survive. It's more important than ever in the retail industry, more important now more than ever too.

Syya:
Right. And as Gary pointed out, it's not just this situation that's happening, it's going to be after the fact. So, the most successful businesses are the ones that are going to be able to pivot and adjust those, that new data that has been unprecedented before to be successful in the future. So, without further ado, welcome to the show Mr. Gary Saarenvirta over at Daisy Intelligence.

What Daisy Does

Erin:
Alright. So, Gary, Welcome to Innovation Calling. We're so happy you're here.

Gary:
Thanks for having me, looking forward to this chat.

Erin:
Well, we definitely are too. Before you jumped on, Syya and I were doing, you know, a ton of talking about this because, as we're recording this, is it, March 25th I believe it is. But we are all social distancing, we're all recording this remotely because we are not allowed to be in the same room together. And it's because of the coronavirus, which I'm gonna have you just talk a little bit about what Daisy intelligence does in just a moment, but I feel like what you do is going to be more important now than it ever has been. And we're going to see, you know, things that I don't know if they were predictable with AI, we'll talk about that just a moment, but, I mean, just something that's happened that is just like something that we've never seen in our lifetimes. So, to kick it off, can you just give us a brief overview of what Daisy intelligence does, and who you specifically serve?

Gary:
So, for our retail solutions, we work for the retail category managers and merchants. They're our users. We help them execute processes and make decisions that are beyond human ability. So, the idea of deciding what products to promote every week. It's really the combination of products that matters, and in the past what we've learned is that customers don't buy products they buy solutions. So, it's the combinations, if you're promoting ground beef, that customer who's making an Italian dinner will buy pasta, tomato sauce, bread, cheese. If you're making hamburgers, you’re buying buns, condiments, produce. Ground beef is a product where the use case is large, nobody just eats raw ground beef. Contrast that to a case of water, you don't need to have another product, to purchase another product to use water, so it has a small use case. And so, it's the combinations that matter, and furthermore, you know, if you promote one brand, if, you know, you promote Coca Cola, Pepsi sales will go down, and vice versa. When it's on sale, people stock up, lots of stocking up going on now, right? And so, all those ripple effects are practically infinite, and if you had, if you had 50,000 products as a brochure, and you had to choose the best 2000 combinations to promote, competent toric mass says 50,000 choose 2000, is 10 to the power of 3600 possible combinations. And there's only 10 to the 80 molecules in the universe. So, this is really beyond human ability is what I'm trying to illustrate, and that’s where the technology can help, and then add on top of that, what combination of prices should be charged? Because the combination, pricing is the same issue, you know, if you promote chips and discount chips, pop sales will go up, even though pop is at full price because of these product relationships. So, add price to that, deciding what combination of prices to charge, and then figuring out how much inventory to allocate, you know, what's the forecast? how much should I put in every store? Those are the three things that Daisy does, helps retailers decide what combination of products to promote, what combination of prices, both promotional and regular, they should charge every day, and how much inventory, you should allocate to the stores. Those are core three weekly deliverables most clients plan weekly, and we just give the answer. So, our vision is to take the human out of the loop, let people do what people are really good at, and let machines do what machines are really good at. And so, that's what Daisy does, we deliver the answer, and we know that the machine can do better than people. On average, our clients see 3 to 5% total company growth using only 50% of our recommendations, because people are still in the mix and kind of poking their fingers in where we believe, in the long run, they shouldn't.

The Advantage to Using AI

Erin:
So, we're talking about ripple effects. We're talking about, you know, we're seeing something and these past few weeks’ grocery stores, bare, were joking about the toilet paper, right? I mean, there's, so with your analytics, and with any analytics and AI, we're using a ton of data in order to be able to make those decisions. How does something, like what has happened in, with the whole coronavirus, how does that impact the business, and is there an advantage or disadvantage to utilizing AI to be able to make these decisions?

Gary:
I think it's a complete advantage to use AI. Because again, it's beyond human ability. Especially in a time there’s dynamic change, and, you know, lots of change. So, you need to be able to monitor the changes. So, the things that the Daisy AI system learns from our clients’ point of sale receipts. It learns what's the halo of every item. So, what are all the items bought with a cheese block or bought with toilet paper. So, we know the Halo, we know what the price elasticity is, if you discount it, how much more is sold? We know what products get cannibalized, if you buy brand A, what brands get cannibalized. Or if you buy coffee, what does that displace? you, and people don't buy, they buy less tea. So, those are all the patterns the system looks at every single day. And we can see the changes and those patterns. And so, if you spot the changes and those patterns, then you can take advantage that, and that'll help you stock your shelves better, help you forecast demand better. Now, in a black swan event, it's going to be a crapshoot forecasting what's going to happen, but at least tracking this very granular change right down to the every store level, if you could respond more quickly on a daily or weekly basis to have some inkling of what to do, that's a huge advantage. Because this behavior is going to stop, and it's gonna go back to something closer to norm, and you want to be able to be on top of that and see when that, you know, because the curve will flatten. We've been talking about flattening the curve in healthcare, the curve will flatten in retail too, and you want to be able to see when that's about to happen, so you don't order 10 truckloads of toilet paper and be sit with five extra truckloads for the next six months.

The Difference Between AI and Predictive Analytics

Syya:
So, if I can take one step back here. So, we're looking at the historical data, if you will, of just the trends that you've seen with, with particular items and products. With Daisy, your application, your software does is not just looking at historical data, it's actually leveraging, and predicting, and anticipating, right? Cause I want to make sure I understand the difference between, you know, the talk about data analysis and analytics, etcetera, and then artificial intelligence. I feel like sometimes they get interchanged, and could you correct me on that? Or verify?

Gary:
Sure. Yeah. I think the vast majority of what people call AI is just statistical analysis. I'd say 99.9% of when companies say AI, they mean statistics, only model what you've seen in history. So, the downside of statistics is you can only model what you've seen, you could never do anything you've never done before, because you have no, no data. It requires examples from the past. So, to learn anything new, you have to go do something new in the market, measure it, test it. So, learning happens only at the pace of time. Right? And it's a model, it's like, it's not, it's a one shot, you look at one pattern, see if you can learn from that. It's not really scalable to say does it model what happens to the business? And so, what the difference between what Daisy does is we do what the aerospace industry does. My background, I’m aerospace engineer. And so, you know, when Neil Armstrong landed the lunar lander on the moon, he didn't actually land it. He just said this is what I want to do, and then the computer figured out how to execute the flight instruction. So, we’d, we’d say that’s, you know, autonomous flight intelligence is the computer flying that lander, and then the human is the pilot. So, it's human oversight of autonomous machine intelligence. And so, in the same way, and then in the case of the lunar lander, it’s the laws of physics govern how the world works. And with the laws of physics, you can say what will happen if I do this? even though I've never done it before, because you have these mathematical laws. In the same way, we've created mathematical laws of retail, which are, encompass some of the things I talked about, halo, cannibalization, pantry loading, seasonality. You know, people buy different things at Christmas than they do in the summer, price elasticity, promotional elasticity, competitive effects. Those are the fundamental truths in retail. Everybody knows that. But we've assembled it into a set of mathematical laws that are like the laws of physics. And so, those laws are independent of the data, and I could simulate what's gonna happen if I do something different. And what we use the data for is to calculate the properties. Like in the laws of physics, you'd say, what's the force of gravity? You would measure that and plug it into the laws of physics. You'd say, how heavy is the Earth? you would plug that into the equations. So, from the historical data we calculate the features like halo, elasticity. We're just calculating these features and then we plug them into the math. And now we can simulate, say if you promoted this combination, what's going to happen? If you charge this price, this set of prices, what's going to happen? And we're modeling the whole company. So, with AI, you could do what you've never done before. So, because you can simulate, I can, I can learn faster than the pace of time. I don't have to go do it the marketplace, I can test it in the computer, I can do 100 million years of retail in one hour if I have enough computing power, because my simulation allows me to evaluate what's going to happen. Whereas predictive analytics, you're doing only what you've seen before, right? And with, the other difference is with, with statistical analysis, there's no action tied to it, so you build a model that says this customer is more likely to buy Coca Cola with an 80% probability, but what do you do with that? Does that say you're supposed to do something? There's no decision wrapped in it. Whereas, within the simulation, the decision is if I promote these products, what's going to happen? Well, the decision is promote those products. So, built into the AI is actually the action to take. So, it's a fully autonomous system. So, those are the differences. You could simulate faster than the pace of time, you don't have to have historical labeled examples to, to, to see what's happened before, the decision is part of the system, it's completely autonomous. And I'd say that’s what aerospace engineering calls optimal control, you know. And the, then human is the pilot, and then what do the detailed actions to take is what the system does, that 10 to the 3600, which is, you know, the computer figures out how to move all those levers and the human says, I want to grow sales, I want to build a flyer every week, I want to be the best baby category and fresh, and I want to be lowest price in the market. So, the human is the pilot, says this is what I want to do, and the AI figures out what's the best way to achieve that.

How AI is Leveraged by Daisy Clients

Syya:
So, I mean this sounds great. And if I'm a Daisy customer right now, I'd imagine they'd have to have some level of competitive edge right now over their competitors. How are your customers and clients leveraging, you know, Daisy's technology?

Gary:
Yeah, I mean we help our clients to answer those three questions, what products to promote every week, what prices, how much inventory to allocate, even how to layout your stores. And so, customers that have that intelligence, because their sales have grown, we've been able to, on average, I said grow customer total company sales by 3 to 5%, which if you're a billion dollar company, that's 30 to 50 million dollars, If you're a 10 billion dollar company that's 300 to 500 million dollars. So, massive gains. Grocery’s a 1% net margin business, so a 3% sales increase doubles your net income. So, companies who are getting 3% sales growth, that's coming from somewhere, it's, it's not like, the food industry isn't growing at 3 to 5%. It’s growing with, I mean, with population growth, and inflation, but you're typically stealing that from other customers. That means you're doing better promotions; you're doing better pricing that attracts more customers. So, yeah, it's a total advantage. Customers who have this will, will grow faster, decline less, they’ll be more competitively strong, and be part of putting the customers who don't do this out of business.

Was Daisy Able to Predict COVID Shopping Trends? And COVID’s Effects on Retail

Erin:
So, I'm curious too. Were you able to look at the, I mean, going back to what's been happening, you know, the Black Swan, we'll call it, worldwide, you saw the kind of movement, you it saw go from Asia, you know, through Europe, and now we're dealing with the North, you know, in North America, and of course it's like Australia is just gone on shut down and stuff too. But were you able to predict, like seeing what was happening throughout Asia, like be able to help customers say okay, we see, you know, A, B and C happening, we see people stocking up, it's time to make sure bottles of water are stocked, etcetera. Like were you, or price it a certain way, were you able to help predict any of that or give any insight into what was about to come?

Gary:
I mean the majority of our customers are in North America, so it's been --

Erin:
Ok.

Gary:
And, right, we have a, we have one customer in Europe and one customer in New Zealand. And, you, you know, so it wasn't, it's not that I have, if I had a more global market share and I was like everywhere, we probably could have taken the learnings from one area to another. Now, product purchase behaviors and different in different parts of the world. Although, we know that toilet paper, water, and all these staple items are being stocked up, so certainly you could on an aggregate level see what's happening across, across different retailers over time. And our goal is to be the biggest AI company in the world. I think over time we will be able to do that and have, kind of, a surveillance capability to see what's happening in some parts of the world and respond to that. And that's a value add that we can offer as we get bigger and bigger.

Erin:
So, in the industry. You know, we talked a little bit about this before I hit record and I want to make sure we cover this too. This, I mean, you're talking low margins. You're talking about, you know, retailers have to innovate, or you're seeing them die, right? Like they have to. So, in a situation like this where you're seeing this huge peak of sales happening in the, like, how can retailers manage to maneuver or shift in order to be able to survive after things? Kind of, you know, we can probably expect what goes up must come down, right?

Gary:
Yup.

Erin:
As we settle out, what can they do in this situation to, to survive?

Gary:
Well, I think they need to, you know, on a very granular level, track what's going on. So, what's happening now is sales is going up, people are stocking up, and, what we call pantry loading or forward buying, and that's going to stop at some point, and then we'll decline, and it'll drop below, all the sales will drop below the pre-crisis levels. Because now people have stocked up, and they're using their stock up supply, and then it'll slowly start to return back towards the pre-normal, but I think the post-normal is going to be different than the, the pre-normal because of all the people getting laid off. I think the product mix is going to move to more basics and essentials. And the average transaction size, the average basket, will get smaller because people are being more economical. The transactions will go up because you can eat at restaurants, there will be more people buying food and stuff. But I think the overall effect is that, that the increase of transactions and the offset in the basket mix. And for some cases will be less than the pre-norm. Now, if your business is already struggling, and you don't come back to that norm, that's an issue. And you need to be today, like we've seen a lot of retailers start to pull their flyer in the short term. They’re, they, they’ve like paused on the flyer and promotions, and we think that's completely the wrong strategy. A because you want to, you want to communicate to your customers and make them feel safe. That's number one. It's like, you know, in, in Canada 80% of households read the flyer, in the U.S. its still a big number as well, read the grocery flyer. And so, having customers see the norm. That's great. It's an opportunity to communicate to them. You need to then be offering discounts on basic items, telling them, hey, this is where we're on top of the supply chain, we're doing our best to stock up, we're gonna limit your purchases, we're, and we're focused on giving you discounts on these core items and we want to help you. I think that kind of communication with customers is super important, rather than stopping. When you stop fliers, what does that mean? Are you charging full price for everything? Is that, are you, are you're not, maybe you're not. You need to tell people what you're doing and communicate, because if you don't, customers will go to the big discounters, and the guys who discount a lot are going to win. Walmart's gonna win for sure. And then, if you're a customer, and you're not paying attention to how many transactions are happening as it starts to return to the norm, you might have lost some of your customers. So, you need to be fighting for your customers right now, because otherwise those customers are going to the discounters, going to the cheap deal, going to where they get the merchandise. And so, it's a fight for your life right now. Like everyone says, start early. You should have social distanced early, you should be fighting for your life right this second, and that means communicating to your customers, giving them great deals, telling them you support them, you value them, you want them to keep coming. Because if you don't, the retailers that do this, are gonna disproportionately win. And so, I think that's where AI can help, is help fight that battle now for those mid-market guys, and I feel, I feel for some of those mid-market guys. I think, yeah, they're, they're gonna really struggle. I think we'll see a raft of bankruptcies and retail sales, especially outside of the grocery industry for sure. I mean, we'll see lots of apparel and other retailers, but I think you'll see a raft of mid-market grocers going bankrupt in the next six to nine months. I find, unfortunately feel that way, and I'm, we're trying to convince our customers to follow some of our advice. And, you know, it sounds self-serving, you know, and this is what we do and we want them to keep using us, but we're not the only one saying this. If you read the McKinsey report that they published on what grocers need to do to survive, you listen to some of the pundits, the time to innovate is now. If you haven't done it already, do it now and get on this.

Barriers in Implementing Daisy’s Technology

Syya:
So, you mentioned the SMB market space, right? I think, I agree hundred percent of the big box retail retailers, they probably have the infrastructure, they have the resources, they probably have their own data scientists internally that can augment whatever, you know, with what Daisy’s, you know, offers. I feel like if I were a small grocer, for example, and I have a very modest IT budget, and I don't have necessarily the most robust data science group or if anybody, for that matter, can Daisy help those types of organizations. Is it, is there a barrier of entry to even take advantage of this level of, of data?

Gary:
Yeah, we want, so, I mean, we are going out there to try to help mid-market retailers for a really low price and say look, we're going to help you. We want to do this, kind of, analyze the changing mix, try to forecast what's going to happen, help them keep their customers, and we're offering this, you know, we want to survive as a business too. I mean, we're struggling, you know, employees are working from home, some of our customers have paused, and, you know, they're, they're concerned. You know, so we're looking for ways to be creative, and we want to add more value. So, customers who aren't, clients aren’t ours today, we want to offer for, you know, 10 K to do this analysis and deliver them daily, weekly trend changes, and all these things that I talked about, so they can help them survive. And then if we go all in and help you today, then maybe on the back end of this you'll be thankful and we can figure out how to work together in the long run. So, we're seeing this as an investment in the future. We want to go all in with our customers and new prospects to help as many of them survive and succeed as possible and give customers great choice in the future. And we're here if anybody wants to take me up on that. I'm more than happy to, you know, to help out anywhere. And our, as, our orientation as a company, we just want to help, you know. I mean, we, I find it sometimes frustrating with our customers. We're trying to help them do better and it's this huge change management struggle to get the help, you know. We're not here to do anything other than help them succeed, and make more money, you know, which means service customers better, you know.

How Daisy’s Technology is Implemented in Businesses

Erin:
Yeah. So, it's actually an interesting question. To implement AI, a lot of people probably are very intimidated by it, you know. Are we, to the first, to Syya’s first point, you know, are we big enough for AI? Should it even apply to us? And secondly, like, are we capable of actually implementing something of AI within our company? What kind of manpower, what kind of systems, does it take to be able to implement something in a, you know, smaller business?

Gary:
I mean, we are a software as a service, so our clients have to do nothing internally, except give us the data and do some process change. You know, every retailer today they promote, every, every retailer promotes, they all charge prices, they all do inventory allocation. So, we're just, instead of promoting milk, bread, and eggs, we're saying do milk, cheese, and pop this week and next week. You know, we're just giving them what they already have. They have a list of 100 items, we're giving them a list with a, with a different 100 items, so they can simply execute what, what they're, what they're, you know, they can do it already. We have all the, the systems, so they don't have to buy any systems, we just need to get their transaction log receipts, the, every single receipt on e-commerce and bricks and mortar. Our smallest customer has four stores. And is, you know, is like a $60 million a year business. We could do this for one store business, right? I say, if, if 1% of sales is a compelling number to you, you know, the cost of what we do at the low end could be 100 grand a year at the low end to do for a small retailer. And in the short term, we're willing to help anybody for, as I said, you know, 10 K onetime fee. Let's get your data and see if we can help you take advantage of this. And we're up and running in, you know, within, this, this onetime analysis, we'd be delivering in a week, give you answers in a week. We can just load the data into our system. And for, for bigger permanent let's, let's help you optimize flyer, and price, and promotions, and all of that, that would take, you know, a month to two months to get set up. And then start to deliver in a really big way. But in the short term we could be up and running and helping somebody in a week.

Syya:
And that was gonna be my next question is, is how quickly can you ramp up? Because, I mean, Erin, Erin has been part of, like, you know, AI type projects. And it's been frustrating to her because of the level investment and time it takes. So, is there a, you say a week that you could roll this out, where you start looking at data and analyzing. If I am a retailer, again I'll say grocers small, smaller, I'll say 4 stores, like your example that you've given, and I don't necessarily have a robust infrastructure, I mean, it is like spreadsheets upon spreadsheets of just who knows what's in there, is it that simple? Or do you need a little bit more organized data in order to process?

Gary:
Well, I think --

Erin:
Can I add to that question?

Gary:
Sure.

Erin:
Not even organized data, but like organized people too, like.

Gary:
Yeah, the human, the human part of it is really difficult, but most retailers, if you have an electronic point of sale, so you have electronic cash registers, they capture data. Most companies centralize that data because they do financial reporting. So, that's the core data that we start with, and then, you know, we capture their product master, so the description of this barcode equals this product. So, that data can be sometimes a mess in small retailers, but we know how to deal with that. And in this short term up and running in a week, you don't need to make everything perfect. The goal is to, that we can start to help them quickly. In the months to two months to get up and running, we fix all of the gaps in their data. And, you know, the big challenge is people, but I would say that the answer, if it's taking you months and years to build an AI system it’s not because it's not AI. If you'd need to bring 100 people in and build something from scratch, how the heck is that AI? We're bringing in a robot that's already built. It's in a computer software. I’m not building that, it exists, it's a product. I don't drop 50 consultants. I don't have consultants, and my employees, they’re not like building stuff on the fly, it's built. And so, we just pump data into our system and the week is how long it takes to get the data from you, to me, us to run it through our computer systems, and then look at it, make sure there's nothing insane in the way we loaded it, and then it's like, after that it's just keep feeding it data and it keeps spitting out answers. That's because it's a productized AI, optimal control and reinforcement learning is what we do. The rest of them are really statistical systems that bring data scientists sitting at a laptop. When it's 10 to the 3600, you can't have a million analysts sitting at a laptop figuring out what to do. It's, it's a ludicrous, I find it completely delusional the current view on what AI is, because it’s, they haven't figured out yet that it doesn't work. I kind of had this epiphany 25 years ago, because I've been playing with this tech for 25 years. 25 years ago, as I'm playing with neural nets and machine learning, all these people are playing with today, and I realized in about two or three years playing with it, it doesn't really work. I think the whole world is having the learning moment I had 20 years ago just by accident very luckily, and I've kind of moved away from that into what we, this aerospace stuff that I talked about.
Is Moore’s Law Dead?

Erin:
So, we're looking at the keynote computational data as it's being analyzed, you know, we're talking about Moore's Law right? our comp, our ability to compute just pure raw data has, I mean, I have heard arguments that people tell me Moore's law is dead, that we can no longer see that level of doubling performance every 18 months, do you agree or disagree with that?

Gary:
It’s flattened out for sure. I mean, there's laws of physics limits, but I mean, I mean, so when I was an undergrad in the 1980s, you know, the computer power I had then compared to now it's been, it's astronomically different. Even when I founded Daisy in 2003, the computer equipment I own today would’ve cost like $2 billion in 2003. The amount of computing we do today is astronomical compared to what I used to do in the 80s. I used to drive around, parallel computing for me in the 80s was, I had three IBM X 86’s, like the very first ones. And one at my lab, one at my house, one at one of my buddies’ house. I'd be driving my car from computer to computer to run stuff. It would run for like days, weeks, months, right? And today on my phone, I can do on my phone what I used to do with like supercomputers at the University of Toronto in the 1980s. So, although Moore's law has flattened out, we've also learned how to use GPUs and FPGAs, new hardware architectures that speed things up. But it's a big compute world, there's enough computing to do the computing that I talked about. And we, the computing we do today, you know, we don't have a gigantic GPU infrastructure, and when we scale and get more and more, you know, I think there's no limits. It's not a big data world. This, this hype of big data is completely misguided.

Misconceptions About AI

Erin:
Can we talk more about the hype of, and the misconceptions of big data AI? What about jobs? So, a lot of people are probably freaking out. How dare you, you come in and you take this technology, and all these great people who would have been doing all that work before are now unemployed. Can you talk about that misconception?

Gary:
Yeah, sure. I don't think people will be unemployed. Like I said, it's, the humans the pilot, right? And so, we're elevating the human being to play a more significant role. I think that the job of deciding what to promote every week, and what the price week after week after week, you got to pick. Okay, I’m the meat manager, what's going in, what's going in the promotion this week? Oh my god, I’ll do sirloin tips and roast beef. What am I doing next week? Oh my god, week after week after week, with really little knowledge of what's going to work or not work, except that I know if I sell this at a buck, I know I'll sell 1000 pounds of it, but that's it. It doesn't mean that the business is going to do well or not well. That's so difficult. Now, if we elevate the person to say, my goal is, these are my objectives, these are my goals, this is my strategy, and then the AI figures out the details, I think that elevates the role of the human. It lets the human be the, be the pilot, and elevates them. Now, in, in some, in retail, there's a lot of manual data entry. Unfortunately, some of that will be replaced, because retails been kind of laggard. All the planning is done by people entering thousands of lines into spreadsheets, and unfortunately that will go away. You know, and other industries will have less, kind of, human displacement. But it's not the AI, I think part of it is just that it, retail has been laggard in implementing automation, and, you know, data, automating data entry could have been done decades ago. It's just that the retail hasn't invested in that. And for you to survive, if your priority is to survive as a business, then I think you need to, need to look at this technology. And some people will be displaced, but what that allows you to do is be more profitable. And when you're more profitable, you can lower your prices, as smart retailers don't bank that money, they lower their prices to stay competitive. That means the cost of living for consumers goes down for you and me. My real mission at Daisy is to lower the cost of living for humanity and reduce poverty. Because if we make retailers smart and be efficient, prices will go down. If we help insurance companies, we work in insurance as well, if we help eliminate fraud and set the right prices, the cost of insurance will go down. Cost of banking, cost of health care. If we lower the cost of everything by using Intelligent Systems, then the cost of living goes down for all of us. That's the game stakes for AI, and that's what we should be focused on. That's what we're trying to do with our customers.

Erin:
Can I be evil and nefarious? And say like –

Gary:
Sure.

Syya:
I understand what you're saying, but wouldn't it be as, if I were a CEO beholden to the stocks, you know, price and all that good stuff, If I can just increase my profitability, and not necessarily transfer it to lower pricing, wouldn't that be something that most people that are of that mindset do to just make more money?

Gary:
No, you’ll lose, you’ll lose market share. I mean, look at what Walmart, Walmart is a super smart company. Every five years they go we're reducing prices 15%, right? It’s a race to the bottom in price in retail, and the way to do that is to get efficient. And because low, lower pricing, it creates a perception with consumers that you'll track more market share. So, if you don't lower your prices and stay competitive to the market, unless you're a very specialty boutique offering, if you're a general kind of grocer or retailer, you need to stay price competitive, otherwise you'll lose market share. So, you can, and I think if most retailers are very afraid to increase prices, they're like, they err on the side of being way less price, which is why margins are less than 1%. So, the smart CEO is not going to bank the money, they're gonna invest in innovation and price. And I think, I think that's the way to the future, invest more in innovation in the short term, so you can do more of what we've been talking about here today.

How Daisy Can Help Companies Innovate

Syya:
Are there any other things, you know, you talk about, about, like being able to, I can't remember the term you use and I'm sorry, but like, I know if you put chips on sale soda is gonna increase. You talked about that, being able to predict certain times of year. Are there any other things you can help companies with, based on the data that you serve, that can be bucketed under innovation? I think a lot of times companies say what does that mean? And the word innovation is just thrown out, you got to innovate. Well, what exactly does that mean? and how, again, your company by serving the data, how can they help do that outside of what we are doing?

Gary:
That's executing all the processes you execute efficiently. So, in store, how should you lay out your products on the shelves? So, you put product in the front page of your flyer, on the front landing page of your website. What goes on the end caps in your store? What goes in the middle aisles? For how, to what height, which shelf do you put it on? You put on the top shelf, the middle shelf, the bottom shelf? So, how to layout the stores. What's the assortment? How many brands should you have in every category? All those decisions. How much labor should you have on the, on the, in the store. Maybe you should make less price changes. One of our, one of our prospects said, you know, I spent $5 million labor a year changing those little price tags, sticky tags on the shelves. And they said, if you can reduce the price, number of price changes I do to save me two and a half million labor, I'll split the money with you, right? And so, what we do can also optimize labor and store. How to layout the store, how much labor should you have, how much sales people you should have, when in the store, when does the traffic ebb and flow, where should you put your stores, you know, close to the population. The whole value chain around running retail, around all these core processes, you know, supply chain, you know, how much inventory should you have on hand, what's your safety stock, when should you replenish, should you replenish at 10 units, five units, or 20 units. You know, all of these kinds of decisions are all optimization decisions, and that's what AI does and what our system does, because we have the laws of retail, we would use the same mathematics for every one of those problems. And that's a benefit of having a system, like an autonomous AI system like I described, because if you're doing statistical modeling, you have a different statistical model for every problem, who’s to say they're not working at odds with each other. Vendor A and vendor B are building two different models that are actually cannibalizing each other, and the net effect is zero. That's what I feel the effects of statistics is. The current branch brand of AI has this net effect of zero, because nobody takes into account the interaction effects between supply chain and this. For example, if I price pizza, frozen pizzas at a buck 99, I might sell more units than I have refrigerator space for the store. So, shouldn't I take into account refrigerator space? How can I do space planning without knowing the, the price I'm going to charge? How should I know what price to charge if I don't know how much space I have? What if the pizza manufacturer can only make 5000 units a week, and I think of putting a price that sells 10,000? Well, I need to know my supply chain’s ability to meet it to me. How long does it take the pizza maker to get the pizza from his factory to my store? Right? Like all of that needs to be considered and needs to be done with one central brain. Our vision is to have the central brain to make these really complex, these million moving parts decisions, and let the business decide the strategy. Like, what's my branding, what am I best at, etcetera, etcetera. You know, that's the role of a human.

The False Positive Issue: Why Predictive Analytics Doesn’t Work

Erin:
So, you've articulated that there are other, maybe competitors, that are arguing 95% accuracy in AI. Are you, is this part of what you are talking to, or alluding to? That it's, 95% accuracy is not as impressive as you would think?

Gary:
yeah, I mean, if I do the simple math, like, when you're predicting things, you're predicting rare events. Anything worth predicting is a rare event. Like if you live in, if you live in the UK, you will, you take an umbrella every day because it's going to rain 30% of the time. If it rained 1% of the time, you'd want to go I want to predict the weather, because I want to decide should I bother to carry an umbrella or not. So, anything that's a rare event, you want to predict. So, like a 1%, I'll do, I’ll do 90% and 1% just for easy math. So, let's say you're trying to predict the, out of, out of a million transactions, you want to know how many transactions are going to have, I’ll use Coca Cola again, a caffeine free vanilla flavored coke. Let's say 1% of a million transactions has that, 10,000 transactions will have the vanilla flavored coke. And you want to build a predictive model that says, you know, with 90% accuracy, let's see if I can predict which transactions are going to have the coke, because I want to give them a special offer that'll make more money, let's say that's an example. So, 90% accuracy means on the 10,000, you got 9000 out of the 10,000 right. Awesome, right? But you got 10% of the 999,000 wrong, right? The 990,000 left over, you know, of the million transactions, 10% wrong, you got 99,000 wrong. So, you got 9000 right, 1000 wrong out of the 10,000, and 99,000 wrong out of the, the big group. So, it, actually your 90% model is less than 10% accurate, because you got this false positive issue, right? and that's the issue. If false positives cost nothing, like you want to carpet bomb the world with emails, there's no, there's no negative, there's some customer negative impact, but there's no cost of doing this, it’s inexpensive to send, send emails, then you can go do, that that false positive rates no issue. But if a false positive rate is has a cost to it, like you, you, like in the case of medicine, you know, if you diagnose somebody with a false positive and you put them through all this medical testing, or you do fraud detection and you have to spend money investigating that prediction, and rare events is a huge issue. Lookup false positive breaking events for autonomous cars, you'll find a whole bunch of videos of cars just randomly breaking, and the car manufacturers have turned those braking systems nearly off. I've seen videos of two cars driving 60 miles an hour, one pulls over and there's a parked car in front. You know, it was a test, the cardboard car, and the car behind drives right through without even touching the brakes, because of false positive issue. If you're using statistics to decide when to hit a break, you know, like emergency braking is less than, it’s a 0.01% event. So, using predictive analytics is completely the wrong problem. Nobody talks about that; it slays my mind how nobody talks about the false positive issue. In, in, in any, and everything is a rare event. Anything worth predicting is a rare event. Otherwise you just carry your umbrella with you every day. And nobody talks, I like, do they not know? Or do they know and they don't want to tell us? Like either one of those scares the hell out of me, I don't, I don't get it.

Will COVID Bring About Further Innovation?

Syya:
That’s absolutely fascinating because you think, you would think that in an industry, again, I'm picking on retail, when the razor thin margins are so razor thin that there is no room for error, I'm shocked right now that you're, you're even telling me this. But I'm not surprised at the same time, because retail does tend to lag behind technology. So, what would it, do you think, do you believe this whole coronavirus pandemic, do you think this is going to be that, you know, straw that broke the camel's back for this entire industry to become much more than, you know, embracing of technology?

Gary:
I wish I could see yes. I mean, I don't know. I hope so, because I think that would be a boom for businesses like us, it would help us, it would help them, you know. I hope that this is the one, because if not, there'll be more and more people going out of business over the coming years, because the world is not going to get any less complex. It's not going to get any less competitive, and technology will continue to improve, so how long do you wait till you invest in technology. And the, the cost of the technology is going down. Like the cost of our stuff is, you know, we're like a, a, a fraction of a penny of what like the big IT vendors would charge to go build a system like this, you know. And I tell my customers for every dollar you pay us, you should get at least 10 in net margin dollars back. And if not, I'll quit, I'll fire myself, I’ll apologize for wasting your time, I'll give you your money back, you know, because I know this stuff works, it's a no brainer. It's just your willingness to, to act and have people play along. It's the human change that's our big struggle.

Who are Daisy’s Customers?

Syya:
Okay. And I know we're getting late on time here, so I want to be respectful, if I were a business owner and I see Daisy, and I say this makes sense, I'm interested. Out of curiosity, where do you sit? Do you, who do you talk to? Is it the IT guys to get access to the data? Is it the marketing guys who, who, you know, leverage that data? Who is your actual customer in this context?

Gary:
I mean, so the buying customer who buys is typically the C level people, you know, in mid-market retail. So, we get access to the CEO, or the CFO, or the chief merchant, right? The head merchant might be a C title, or an SVP of merchandising. So, those are the people we talk to. And they're, they're bought into net income growth, right? So, but if the CEO says yes, and we've had this happen, the CEOs like I'm all in. And then we struggle with the users who are the day to day category managers merchants. So, it's like we're trying to build what's the value proposition for the day to day users, because they're afraid of their job. And so, we're working on trying to make the experience using our system, more, more fun if you can call fun doing work, but more pleasant, less threatening, convince them that it's a, it's a job elevation opportunity. And so, we're working hard at putting an interface on the technology that makes it more user friendly, because the C suites bought in. But if the users don't buy in, then we won't achieve the results that we have, you know, the ideas is take the people out of the way, right?

How to Contact Gary and Conclusion

Syya:
Exactly, exactly. So, if, you know, I'm a business leader and I want to learn more about Daisy. Gary, how can we get a hold of you?

Gary:
Yeah, you can go to our website daisyintelligence.com and look at, there's, you know, my email address is there. You can find me on LinkedIn. Look up my last name, Saarenvirta. S, double A, R, E, N, V, I, R, T, A. I’m the only one, me and my brother, and my son, and my daughters. You'll find me, my email, my work email address is at my LinkedIn profile. I’d be more than happy to talk and see if we can help, you know.

Erin:
Thank you, Gary. Very much. We’ll of course include that all over, at our notes page over at innovation calling.com. Syya, do you have any other questions?

Syya:
I do not have any other questions. Gary, thank you so much for your time, and it looks like that wraps it up for another episode of Innovation Calling.