#smallrooms podcast with Rob Kenedi

Listen to the #SmallRooms Podcast

Gary Saarenvirta of Daisy Intelligence chats with Rob Kenedi of #smallrooms podcast.
Entrepreneurs in Small Rooms Drinking Coffee is an unfiltered podcast that candidly talks with entrepreneurs and investors about the details of starting a tech company.

Rob Kenedi, the founder of #smallrooms, interviewed Gary Saarenvirta, founder and CEO of Daisy. We're happy to have a chance to sit down with Rob. In the podcast, Gary discusses his career and journey to founding Daisy Intelligence.

#SmallRooms Podcast Transcription

What Does Daisy Do?

Rob: Hello everybody. Welcome to Entrepreneurs in Small Rooms Drinking Coffee. I'm Rob Kennedi, and today we are here with Gary of Daisy intelligence. How's it going?

Gary: Great, how are you?

Rob: Awake, thank you very much. It's after Labor Day weekend, depending on when the show airs. We're tired. That's the that's the short answer. So, why don't you tell everybody what days intelligence does?

Gary: Were an artificial intelligence platform that helps our clients operate smarter. So, we take transaction data, for example, in retail, we take the transaction receipts data. So, every, every item purchased you get a paper receipt, you see one line per product. On your receipt, we have that data in a database, or for clients, typically several years, every single product ever bought, and that we help retailers to use that data to make smarter decisions. Some of the decisions retailers make that are very difficult for people to do are what product should I promote every week? You know, every week a retailer has to decide what do I put on the flyer? or what do I put on my e flyer? What emails should I send out? And they have to choose which products go on all those marketing channels. And if you're a grocery store, you might have 100,000 products and you got to pick 500. You know, 1000 choose 500 is, is a number that's more complicated than playing chess or go, it's more than the number of molecules in the universe. So, people are making these decisions. And then similarly what prices should I charge? How much inventory should I put on hand? So, these are quantitative problems that people are okay at. But evidenced by the profitability in grocery and retail is, you know, typically three-ish percent. We're barely cutting it, but computers can do those jobs much better. So, that's what we bring to the market. An AI platform that helps make the decisions for the clients. There's no dashboards or clients having to interpret this. It's, we give the recommendation, here's what we recommend you do, here's the products you should promote, here's the prices you should charge. And if they execute those recommendations, then we measure it and we show the incremental value of those things. And we've been able to show our clients that we can grow their sales by 5 to 10%. Which if your profitability is three, and we grow your sales by 5%, take off the cost of goods sold, were almost doubling your profitability. So, that's where, where AI is. I think it's going to change human decision making, that, that's what we're about.

Who Are Daisy’s Customers?

Rob: Is there like a sweet spot in time in terms of your customer size? Like if I were to sell something on Shopify, I'm too small, but you need a product catalog of x, or revenue of x to be customer.

Gary: I think it's, there's an affordability factor. So, I think, you know, our goal ultimately is to get down to mom and pop, you know, somebody pay us $50 a month, you know. We're not there yet, but that, that's the vision. That, because, you know, retail is tough. And we, were also, we have plans to go to other industries where, we're doing work in insurance doing health care fraud, identifying fraud and healthcare transactions or healthcare claims. And we plan to add an industry every year and a new geography every year. So, we're Canada in the US now. We're, we're, just signed a client who's got offices in the Caribbean, and Alaska, and Fiji, as well as Canada. We're selling into the UK this year.

About Daisy’s SaaS Model

Rob: And do you charge like, is it a SaaS model? You pay a fixed fee no matter how much you're doing, or is it based on the data set? Or is it based on the increase in margin? Or --

Gary: It's a SaaS model. It's based on the clients’ annual total sales, which is kind of a proxy for how much it costs us to manage that, you know, the volume of data, the size of the infrastructure. So, we base it on the clients’ total sales volume. In other industries, in retail that's how we price it, in insurance, because they're used to playing a claim, claim based fee, we can price it either as a SaaS monthly fee model, or a per claim fee, which is something that industry is used to. So, we have some flexibility how we pay, but it's a subscription service.

How Daisy Was Started

Rob: Got it, got it. So, like starting one of these is not easy. You, you started this quite a while ago, I think you said to me before the show it was like 2003 when you got going, but before that you were not running your own gig, you were working for some others.

Gary: Yeah. So, I mean, I have, I went to University of Toronto. I have a graduate degree in aerospace engineering, a master's degree. And, which is applying computational math to engineering problems. And so, then I saw this opportunity in the world. You know, some friends of mine who went to school, we started a software company and that's where I got my first exposure to, kind of, business. And I saw that the business world was kind of woefully undeserved with technical people. And I thought, wow, this is an opportunity to build a career, I just, kind of, lucked into it. I started work for a company called Air Miles, the, you know, their loyalty program that they have, you know, it’s called The Loyalty Group, and I was there, I built their first data warehouse. And I started doing this analytics consulting about analyzing data and helping them make smarter decisions. That's what's always been my mantra, I thought that we can, we can leverage the information companies create. Ran into IBM, started using some of their technology. And IBM was interested, I traveled the world doing customer briefings, and they paid me to go to Asia and do training courses. And data mining was the terminology back then. And so, I think I've been doing this advanced high-end analytics using really sophisticated software for over 20 years. And then at IBM, you know, fabulous company. I never have anything bad to say about other businesses, you know, they’re successful. Wasn’t for me, I'm too much of an entrepreneur, and I thought okay, it's, start my own business. And so, that's what I did. I left IBM and started, you know, we called it a different company then, it was, it was Loyal Metrics. And fraud box were two solutions in retail and insurance, but we, kind of, merged them all back into, into Makeplain, and then we renamed to Daisy Intelligence this past year.

How Gary Became an Entrepreneur

Rob: So, the catalyst for you to leave, I mean, you're, you're not a kid out of school who had no job and, kind of, no debt. You, you built up a career. And you're like, ah f*** it, I'm just gonna start my own company. What, that's not a trivial thing to do.

Gary: No, absolutely not. I mean –

Rob: Especially working for IBM, if they're flying around the world, your life was probably not horrible, tiring, but not horrible.

Gary: I worked like, I worked a lot. I was always building new businesses in these other companies, and I guess, and I always thought that I needed to have partners and other people to start a business. And I did, when I started had a partner and it was very difficult. We started out working, working at home, and then we were in shared space. And I was lucky in that, because I was known in the space I was in. I immediately got a few clients who knew me, and I was able to go back to my existing clients that I worked with at IBM and other, other times. And, you know, Canada's a small market. So, I was able to pick up a couple of customers and get going and start building up my business.

Rob: Did you just do it on a consulting basis?

Gary: Yeah, it was professional services. I always had the vision to somehow build a platform for analytics. I realized that doing analysis is not a human endeavor, because it's impossible for human beings to do this. You know, if you, imagine in retail you have 100,000 products, how can any one human being get their head around what's going on? All the patterns, you’ve got 400 stores, got millions of customers, it's impossible for humans to get their head around this. I realized early on that this has to be some kind of automated thing, an AI platform. And so, I've been, kind of, working towards that, you know, the whole time. And that's what, that's what we started the business with a goal to do that. And then, you know, to fund the development of that we had to have some revenue. So, we did professional services and consulting, and then used the money from that to fund the development. And that way we didn't get any investment capital. So, I was able to, you know, we ran, you know, in the, in the 10 years before we really switched on the, this as a subscription service, we ran about 30 million in revenue through the company, and it was all used to fund development of the platform.

How Daisy Generates Profit and Services Its Clients

Rob: So, how do you, how do you do that? Like many, many times on this show, I espouse my theory of all services companies want to become product companies. And this is what I think. How do you keep focused on that? When you're, you've got people coming in with, I mean, homogeneous problems in the, in the, in the abstract, but when you get right down to it, you know, IBM's dataset is different than Loblaws’ dataset, which is different than Air Miles. Like they're completely different the way they choose to structure entities, and the way they group things together, that's really complicated stuff. How do you, f***, continue to invest in a product without, and service your, service your client at the same time?

Gary: I just had this, you know, this one, you know, this one vision, and I stuck to the vision that, you know, helping companies make smarter decisions. I thought that, I thought that every business out in the world, every government organization, any organization that spends a lot of money, could probably double or triple their profitability if they leverage the data they had. I'm an engineer. So, if you have data about a, about a process, then to make that process better analyze that data, and start making smarter decisions. And that's what engineering systems are. You build feedback loops, you watch a process, you feedback loop, and you make the process more efficient. And that, that wasn't being done in business. So, I was slavish to that vision and I've never veered from that. I’ve always felt I was right --

Rob: How, how did you accomplish that though? Did you like carve off a team to be building core product that would be servicing clients as it went? Did they just go and hide in the corner and you just fed them requirements? and they ---

Gary: No, it was, I mean, I'm, I'm unique in that I can do every technical task that our company does, so I did a lot of it myself. Like, so burning the midnight oil, you know, I can say I've, I've given up my health at many moments in time by like just working, you know, 16, 20 hour days months in a row, seven days a week, you know. And my family, you know, was used to seeing me sit on the edge of the couch with my laptop open, sitting there all night. And kids waking up to go to school and still see me sitting there and go: ‘daddy, did you sleep last night?’ ‘No, no honey, I didn't’ ‘I, you know, what do you want for breakfast?’ You know, I was like, that's a usual kid’s conversation. You know, so it was really burn the midnight oil. And I had some good loyal employees who helped out and did bits and pieces of it, but it's been, was largely self-done with two or three employees, who are long term, with me who contributed. And then the odd time we hired a contractor here and there. And we used all the Canadian government programs, which are fabulous, the SR and ED tax credits. Because we're doing really hardcore development, so I think we're the poster child for that program, doing this AI research, doing that. And we've gotten some IRAP grants, NRC grants over the years as well. So, using all of those programs. And just stuck to that vision, you know, we had a business, and generated some profit. And those profits, we directed into funding the time away from doing professional services.

Rob: So, how do you, I mean, you kept it small it sounds like?

Gary: Kept it small –

Rob: From the beginning, so how do you --

Gary: Well, our services business got large, peaked at like 70 people, right? So, we had 70 people at one time in like year three of the, of starting this venture.

Rob: So, did you, did you focus on the product and you had someone running the services side? Is that how you did it? Or how did you –

Gary: No, I did both. Because like, because the services was around data warehousing, and that's the practice I ran at IBM, data warehousing and analytics. So I kind of, I wasn't involved in the hands on doing of the professional services, I found good people to run the projects, but I was involved in selling them, and, you know, just did a bit of everything, you know, time slice as much as I could, and burn the midnight oil like crazy for years, and have this vision, and just never wavered from it.

Daisy’s Commonsense Approach to AI

Rob: So, how do you, you know, to, to start a technology company today costs nothing, which is why everyone's an entrepreneur, everyone's a startup, quote, unquote --

Gary: And that's why I started. I'm not, I mean, it was just brain and laptop. And, you know --

Rob: But in some cases, like, you know, I could put a game together in three months and launch it, and it'll either be successful or it won't. Doing artificial intelligence properly takes more than a few months. So, how do you know that you're moving in the right direction? Like at what point is the product A ready to test? Because, you know, the methodologies now are like, build an MVP, test the MVP, iterate, pivot, blah, blah, blah. You obviously knew there was a market, and you're trying to build a product for that market, but this is a complicated one. How do you know when it's ready enough? How do you know when to stop? How do you know when you’re going in the wrong direction? How do you know it’s spitting out the wrong thing?

Gary: Well, we were, I mean, we were, so we were solving, you know, my vision was always solve the core problem in every industry. So, in retail, it's what product price availability, you know, so it's product price inventory. So, retailers do that every day, and they don't do well. So, that's a problem. I didn't have to go find a problem. And every industry has a problem like that. In insurance, it’s claims, in banking, it's under, its, you know, loan approval, and risk, and treasury, and telco, it's network provisioning. And so, you pick the core largest cost line item in an industry, and then you go work on that problem. And so, we learned a lot through consulting, and professional services, and doing analytic projects. And then it’s, at the end of the day, it's common sense too. It's like you’re, why do people buy ground beef? Because they're going to make hamburgers, they're going to make a pasta dinner, they’re gonna make, you know, like a Mexican dinner. And when you see the patterns in the data you see, you know, every product has a use case. And so, at the end of day, its common sense is what we're driving, and that's, you know, that's, that's, what was the point that we could test, that the system is putting out valid recommendations. You know, when we see, you know, we don't, you know, we can't look at everything the AI machine is doing, but when you see it's spitting out that a good product is ground beef, because it has a use, because it has 1000 use cases, and every use case means a customer will buy two or three more products, that our system is more likely to recommend ground beef than a case of water, because a case of water has, the use case is you buy the water, you take it home, you don't have to buy another product with water to consume it. So, you can see the commonsense output starts to make sense, and we evaluate it from that perspective. And we designed a system that takes into account all of the commonsense things that you think it should, like affinity, what are, what other products are bought? You know, brand switching. When you promote Pepsi, coke sales go down, and vice versa. When you promote coffee, tea sales go down, when you promote tea, coffee sales don't go down, because tea drinkers also drink coffee. So, you see all the patterns, you know, when things are on sale, consumers buy two or three weeks’ worth because it's on sale. So, they forward buy. You see, so, price elasticity, you lower the price, people buy more, you promote it on TV, people buy more. All of these commonsense things that you know have an impact, and you've, and we've built a theory around that, and that's what the AI machine does. So, when it's spitting out common sensical decisions that the clients evaluate. And we've kind of worked with one client at a time, one or two clients at a time, over the years to pilot, and work with them, and make sure that the output is real. And then, you know, when it's ready for prime time, it’s spitting out common sense, and --

Retailer Decision Cycle

Rob: So, are they, are you interfacing? Like they have the data, they have the data and you're sucking it in, or you're connecting to it and running the AI on it, is that how it works?

Gary: Our clients send us the data. It's our infrastructure. So, we, we get the data. It's not real time. I mean, you only need to do analysis as fast as you make decisions, right? So, analytics, if you're, you know, retailers typically make decisions on a weekly basis. So, they decide the weekly flyer, the weekly prices for the flyer, weekly inventory allocation to stores. And so, that's the decision cycle. That's the cycle of analysis, you analyze on a weekly basis. Now you can real time decision, but it's based on a weekly analysis.

Is AI Necessary in Retail?

Rob: Yeah, yeah. So, how do you like, the stuff that you articulated, that list of things, they all do seem like commonsense. So, do you need an intelligent machine to tell you that when people buy hamburger ground beef, they buy buns, and they buy ketchup, and they buy, do you need a machine?

Gary: Yeah, we do. Because there's, there's 10s of millions, there's hundreds of millions of patterns. So, I gave you one example out of, out of millions of examples, right?

Rob: So, is it that you feed it patterns, you, you expect the outcomes, and then you say: ‘now go find Interesting stuff’ –

Gary: No, it finds the patterns --

Rob: That’s what I’m saying. Like, you, you validate it with stuff that you know makes sense, and then you're like, I think you're not an insane machine, now go tell me other cool stuff.

How Daisy’s AI Works

Gary: Yeah, for sure. We, I mean, so it's, we've built a mathematical theory. So, independent of the algorithms, or the data, we've built a theory of business, a theory of retail, in every industry we work in. And so, like Einstein's theory of general relativity, we built the theory of retail. And based on that theory of retail, we, the, you know, we solve that equation, like you can solve the theory of, Einstein’s theory of relativity, to figure out how to get to Pluto. The people I work for, they hate these examples, because they don't think anybody understands them, but I don't know, I keep using them because I’m an engineer –

Rob: I'm a physics nerd. So, I get it. Maybe our listeners won’t, but I, they’re pretty smart.

Gary: So, the general theory of relativity is just like, to go from here to Pluto, you have to, you solve the theory of relativity, you plot a trajectory. So, you can solve the theory of relativity to get a trajectory. You can also solve it to calculate the solar mass. There's a whole bunch of different things you can solve using that theory. So, we solve our theory of retail, and it gives us a trajectory of products to promote every week. Every week there's a list of products, or a list of prices, or, or inventory allocation. So, our theory has all these things embedded in it, and we solve that theory to find these answers.

Moore’s Law and its Effects on Computing

Rob: So, how come, and I think you sort of answered this before, but the hard part is, you know, creating this machine that's intelligent, the actual artificial intelligence is smart enough to parse all this data and come to interesting conclusions. That requires a fair bit of engineering. It's not just you and your laptop, I imagine the team is growing to be smarter, to make the machine smarter. You could have taken, I mean, 2003 was early days, but you could have at some point taken some sort of capital to bring in a million engineers, and do all the machine learning, and all of the things. You've chosen not to do that.

Gary: I think 2003, if you look at Moore's law, the cost of computing was, was orders of magnitude more. So, just to give you a sense of the cost of computing, we bought infrastructure that has enough disk space to store all of the retail data for North America for a decade, right? And has like, more than a terabyte of memory, 20 servers with 400 processor cores, and that cost like $40,000, okay? 10 years ago, that would have cost $4 million, right? And so, 10 years ago, I knew that we weren't ready, you know, we were only selling to the big guys, and I, you know, is waiting for Moore's law to kind of come to us. We're doing one or two customers at a time. Then this whole interest in AI happened, this, so, there's a confluence of events that said: ‘Okay, now we're ready to go.’ And so, we were chipping away at it while, and we were paying the bills, you know, I was making a living doing this. And I just didn't feel that the market was ready for it, right? And so, timing in the last couple years has shown that Moore's law is gone, gone down, the cost computing continues to go down. Moore's law is kind of flattened out now. We're not sure if we're gonna break through the next level. I'm sure there will be some new breakthrough coming --

Rob: Inter-dimensional computing –

Gary: it's been, it’s been, kind of, stagnated, you know?

How Daisy Raised Capital

Rob: But okay, so now, now it's a sexy time for, for these kinds of businesses. You know, capital is, I mean, venture capitalists are a little bit wary now, but cost of money is relatively cheap, computing power is insane, AI is hip, we can talk about that, what that means in a second, but you could say: ‘okay, now is the perfect time to find a bunch of people, double or triple the machine, make it smarter.’ Like there's the competing power, then there's adding the intelligence piece, you could get raised some capital.

Gary: Yeah, we have. So, we have raised capital this year. So, that's, I decided two years ago that it's time to put the pedal to the metal, and we went out and raised capital in the last couple years –

Rob: It’s like a friends and family kind of thing?

Gary: We did an angel investment, we went to the, all to the, the Canadian Angel groups here in the GTA. There's a bunch of registered Angel networks. And so, I ran into those and I, we did one round of investment through the angel network, we did a few hundred thousand dollars. And then we found a couple of private equity investors who are very successful Canadian businessmen, very wealthy individuals, and we had two of them invested earlier this year. And so, we raised, you know, $1.5 million earlier this year.

Rob: Wow, that’s pretty significant.

Gary: And the team has grown dramatically since then, so –

How Daisy Used its Funds

Rob: What did you think the money would be useful for?

Gary: Market expansion. And so, we've hired salespeople. We've hired some more tech people, you know, the more, you know, the tech people using, you know, government funded programs that, you know, we can offset some of the cost of the tech. But it was primarily market expansion, sales and marketing activity, and that's what's really, you know, we've gone from, you know, 2 customers to 10 customers in the last seven months.

Does Daisy Plan on Expanding its Customers?

Rob: Are you staying on the same size of market customers for now? Or are you actually starting to go to smaller customers?

Gary: We're going to, in the US grocery market is our interest, is the retail priority. And we're doing other industries other than grocery, but, and grocery is kind of the pinnacle of what we do. You have large transactions, you have high customer frequency, customers buy every week. So, that's a, that's a great industry. And then the United States, there’s like 600 grocers, right? So, it's a huge market. And the, there's a lot of large grocers. So, you know, we're going at, you know, few hundred million to the biggest guys. Yeah, we're going to the, you know, all the way from, kind of, you know, there's a whole bunch of $200 million grocers that you've never even heard of, that are, that have 30, 40, 50 locations, that are, that are amazing companies, and so --

How Does Daisy Perceive Competitive Threats?

Rob: Crazy. So, how do you deal with, like interestingly, in Canada, we had on the show a company, I think we may have talked about this, where we had a company called Rubikloud that also does like big data, not necessarily for groceries, but it's for retailers. There's a number of entrants in the space that are newer than you, you've been at this longer. How do you perceive the competitive threat to those guys, or do you see it that way?

Gary: I mean, I'm not so arrogant to think that we're the only people out there trying to solve some of these problems. I do think that our experience and working with retailers, like, so we've worked hand in hand. And I've spent 25 years, and I've learned a lot working with my clients. And I think that's the most valuable time I spent. So, we have a lot of business expertise, and domain expertise. That's why we don't, you know, a lot of companies build dashboards and, you know, reports then they expect the client to figure it out. I'll tell you now, the retailers we work with, the people who run the retail businesses, they don't have time to do any analysis. That's why, you know, I've learned over 25 years, wasn't, wasn't through a great idea that I came to this realization that you just have to give decisions to people, then they can decide whether they want to use your decision or not. But there's no time for analysis on the on the client side. And so, that's, I think, and I haven't found a competitor yet who's doing, who said: ‘I believe in my stuff so much, that I'm going to take it to the goal line.’ Like when I when I tell one of our clients to put this product in the front page of your flyer, I mean, there's a gut check, we better know were right, right? Because if we get it wrong, that's millions and millions of dollars. So, we believe in our technology, we know it's right. And we're able to take that leap forward to say this can run your business, right?

AI and Change Management

Rob: How do people, how do people viscerally react? it's one thing if it's like, you, it's almost a human quality, or like, this person has been working in, in my in my business. You know, Galen Weston's been working in Loblaws for a while. I think he has some idea of what the customer wants, you would argue. And then here's a machine, you know, where you've got the expertise, you have those relationships, but the machine is saying you need to put bananas on the front page. And they're like: ‘ah!’ Is there, is there a visceral reaction to a machine recommending you things? Do people, or do people take it?

Gary: There's, I mean, that's the biggest change management hurdle, is the people. There's no, there's no, there's very little process change, because companies already have processes to do these things. So, we just plug in. There's a, there's the secret spreadsheet going through everything company that's got the list of products for this week's flyer, or this week's promotions. And then all we say is turnaround for a millisecond, and we put our list in and they just keep on going, right?

Rob: I see, but tell where it came from, so that --

Gary: You know, but there is, they know we do that, right? So, I'm just illustrating that we don't, there's not a big process change, but the people change, for the people to buy into it --

Rob: That’s what I’m talking about.

Gary: You know, that, that's the challenging piece, but then we just go back to say: ‘look, this is impossible for humans to do this.’ You know, it's like lotto 649, you know. 49 choose six is like 3 trillion, 100,000 choose 500 is like 10 to the power of 36,000, right?

Rob: But, I mean, that's rational. I mean, it made me laugh, there was the, the commercial on for the lottery yesterday I just saw, and I was like: ‘you know, I think you're more likely to die in a plane crash getting your, your lottery ticket, than you are actually winning the lottery.’ But people, that is rational thinking, right? Like, that is rationally correct. But people make decisions emotionally, I think.

Gary: Yeah. But we think that, what we say is that, like you're doing great, you're a successful business, you know. Like, you know, you're getting an 8 out of 10. Like you're, you know, you're not like crappy at it, otherwise you wouldn't be in business, right? We’re saying, you know, we can do 5% better. So, by number crunching the heck out of the numbers, we can, we can increase sales by 5%.

Rob: Do they do some weird AB testing with you? How do they prove to themselves that you're okay?

Gary: Yeah. We do the testing, we measure it, and prove that our, our predictions, our rankings, our decisions correlate exactly with their decision making. They don't even have to use our recommendations. We can score, we can, we can give it a score --

Rob: I see. If you listen to us --

Gary: Yeah. We can say that, you know, we have this thing where we rank everyone's picks, we rank them from 1 to 100,000, you know, if that's what you do every week. And they’ll, and then we'll say, look, the average ranking of your flyer, or your promotion this week was 472. We would have, and then we need, that, when you look at next week was 250, we could say look: ‘the 472 week was worse than the 250.’ We can see that the sales follows our scoring, which is the output of our theory.

Rob: Does that work?

Gary: Yeah, it absolutely works.

Rob: Yeah. And people are like: ‘Ok, I get it. You’ve convinced me. Try.’

Gary: And then we measure it for a period, we've done other, another test, is do, you know, six months. Ok, we're going to use your system –

Rob: And then they do it on the side for their, on their end? Is that how --

Gary: No, they do the whole flyer, you know. We've done that where we, we then measure and prove to them the incremental lift. We agree on, have a, we have a promotion measurement methodology that we agree on and share with them, and we measure it. Or the third way is an AB test, if you can do two different markets, and we've done that. And we've shown for one client, you know, did an AB test, you know, our, our flyer picks beat theirs by, you know, 250%, right? And you know, so there's a clear example of, okay, we're doing 250% better on the, on the sales front, showing weekly sales lift of 5% over a several month period where they're trying to test. And then, recently, getting down to these rankings, our rankings correlate exactly to sales output. So, these proof points, and as we add more and more customers, then there's more belief, and it's getting easier to convince customers.

What Counts as AI?

Rob: So, you've capitalized on, literally capitalized on the wave of AI sexiness that is now in the universe. It's at the top of the hype cycle, as we talked about before the show, and it's like probably on the way down. We talked about like, what, how the press, or how people are using the term AI these days, and how you’re actual AI, and then there's a lot of not actual AI. How do you perceive the perception of the market right now? And where you fit in it with respect to that, versus like bots that are doing crazy stuff?

Gary: Yeah, like, so, I think, you know, the definition of AI like, you know, so there’s great technology out there. I'm not poo-pooing anything anyone else is doing. But if we're, I think you need to be clear with our customers, so that they understand what they're buying. So, it seems to me that, you know, the machine learning terminology, that a machine is learning, is around optimization. They say if you're optimizing a statistical model, like even linear regression would fall under the definition of AI according to what, what the, the way it's being used today. Which, because you have a bunch of coefficients, you have an algorithm that's figuring out those coefficients on its own, you didn't program the log for the coefficients, but it was optimization.

Rob: So, what's an example? A non-mathematical example for our listeners? Like, what is, what's an example of that, in the real world, that's not actually artificial intelligence or machine learning?

Gary: Yeah, I would say like, you know, how you pick, you, you know, customers to market to, you know. Like that, that example of, here's how we market the customer. I think some of the chat bots are not, are not, the, they're doing, they're exploiting historical patterns. So, if all you're doing is fitting a model to what you only know, then that's not really, there's, there's no artificial intelligence component. You're just building a mathematical model to fit the patterns in your historical data. Whereas learning, like, so, learning would be, think of how you learned as a child. You know, you had the, you had the different shaped blocks, and you remember, you had that little, some kind of board you put them, you tried to put the star in the circle and it didn't fit, and you tried it in the square and it didn't fit, and you found the star hole, and then you picked up the square. So, kind of, random trial and error, you learned, and you did that. And when you smacked yourself in the head with it, you go: ‘that didn't feel good.’ So, let's, so through experience you, you brought your experience to it, and you, and you eventually figured that out. So, that was trial and error, trying something you've never done before. So, I think that's part of the AI definition. That, if you're, if you're trying something that's never been done before, that, then how do you evaluate something you've never done before? Well, that's where it comes to. You have to have the ability to simulate the world independent of, of actual measurement, like self-driving cars. There's laws of physics. I know if I turn the wheel, the car is gonna go to the left, and if I, you know, turn the wheel to the right, the cars are on the right. I push the accelerator, it goes faster. So, you know those things, and then you can calculate with the laws of physics. So, now I can simulate car driving without actually driving a car, so you have car simulators.

Rob: Right. But that is machine learning, there’s intelligence happening –

Gary: That’s AI –

Rob: Because now it’s like: ‘Oh bus.’ If I go left and I hit a bus, that's probably not a good thing.

Gary: Yeah. So, the idea is, AI is, to accumulate long term rewards. I talked about, reinforcement learning is the technical term for the type that I consider AI. And that's, you're trying to accumulate long term reward. And the long-term reward is stay on the road, don't crash and kill yourself, don't run over people. And don't, you know, that's the long-term reward, that I got to repeatedly do that over months, years. And you can simulate that. You can, you can now evaluate the algorithms and let the algorithms learn what, you know, you have the laws of physics govern it. The game of chess, same thing. It's got a highly structured rules, and you can simulate without playing games. Same with go, you can simulate it with a computer. And so, basically, you can do more learning than human beings could ever possibly do, like in the examples of go and chess, you can branch out more games, right?

How Quickly Can One Create an AI Startup?

Rob: So, it took you, it took you quite some time to get, you know, like 13 years, to get from where you were to where you are. Is it realistic for a small, like, will it, I know, things have, you know, competing cycles, as you said, Moore's law and all that other good stuff. Is it realistic to create an artificial intelligence style company in less than a few years? Or does it take just a really long time? Because you have to build that machine, and you have to train it, and that just takes a really long time.

Gary: I think you could do it quickly. Like, I mean, you can take --

Rob: If you started Daisy again today, would it be a quicker machine to build?

Gary: Well, I think the problem, we had to learn about the businesses that we work with. So, that's the difference. So, you can build an AI platform to solve a problem that maybe doesn't require so much learning about the domain. But to understand retail, and understand the human dynamics that goes on in retail, and the challenge. I think that's what took years. I think, and I think our goal is to be a billion-dollar business. We, we want to compete against IBM, and apple, and Facebook, and Google. And these are the companies that are using some of this technology. And Uber and Alibaba. And we want to be the new entrant into this space and be a multi-billion-dollar business. So, you know, we've built a solution that we believe is enterprise class and can make the largest companies in the world double and triple their profitability. And I've been there and done that. So, we know we're on the path, and whether we succeed or not, who knows, but –

How the Daisy Platform Remains Competitive

Rob: So, does that mean, so unfortunately, as I said, it goes quickly, we're timing out. So, you know, in terms of, you've raised some capital to kind of blow out your market. Are you on this venture capital trajectory? Is that inevitable? If you're competing against those guys, you really need to be super well capitalized from sales. They have a huge R&D team, or they're like, you know, they're hiring the same people, presumably to build their cars, that you hire to build your sausage maker.

Gary: Yeah, I think, you know, we've gotten some, you know, some great backers who are helping us, and our goal is, if we become profitable, in this coming year, that we're going to start generating enough cash as a business that we may or may not need to go to a larger, kind of, venture capital round. I think the difference with us is IBM can't become the AI that we are, you know. We're gonna, we're better, faster, cheaper. We can be up and running with a new client in less than a month. And my goal is to drive that down to like a day or two. So, you can send me your terabytes of data, and we'll be up and running, and starting making decisions, and running your process within a couple of days. We’re an order of magnitude cheaper, you know, maybe two orders of magnitude cheaper than IBM, or Oracle, or the existing status quo companies. And the solutions are better because it's real business domain, we have a theory that's independent of just the historical data that companies collect. So, I think, and these companies have so much vested in their current models. IBM is not going to throw that all out to become the Uber of analytics, right?

Rob: Right. Not that they even could be if they did, necessarily –

Gary: Yeah. And so, these companies have so much inertia around their existing model, that you're not going to see an entrant come into this market from those existing companies. It'll be a new player, like us, somebody who will come in, like an Uber, like an Alibaba, like an Amazon, when they came in that totally changed the market. And then everyone else is scrambling to catch up, and they’re unable to, you know.

Rob: So, you're what's, you’re what’s keeping them up at night, basically?

Gary: I don't know. I don't know yet. Maybe the more we talk, and I don't know if they know about us, and they know about, you know, IBM certainly knows me a little bit, you know, from IBM Canada. I worked there and was on their, kind of, top 25 technical leaders. So, I'm sure there's a few people out there who look, but we haven't really hit the radar screen in a big way yet. We hope to do that over the coming months and years.

Concusion

Rob: Cool. Well, we are out of time. I'm really sorry.

Gary: No worries.

Rob: But thanks for coming down. So, if people want to check out Daisy, where do they go?

Gary: Go to www.daisyintelligence.com. We have a presence on LinkedIn and Twitter. You can see some social media blogging, and we have a YouTube channel. So, entertaining stuff. We're trying to be professional and entertaining. I don't like the dry stuff that's out there, you know.

Rob: Like our show, I get, I get it. Well, well thank you very much for coming on the show. And, you know, what's cool is that you actually wrote into us, which is kind of cool. So, if you guys, you know, you can actually talk to us, and you actually can be on the show if you're a legit startup. Isn't that crazy? So, thanks again for writing in. And if you do like the show, please do rate us on iTunes. Five stars, woo-hoo. You’ll make us rich, or not. But we'd appreciate it. Thanks to TWG for hosting the show, thanks to Nick Kuhn for producing it, and thanks to Gary for coming on the show. Check out daisyintelligence.com, and we'll see you next week.

Gary: Great. Thank you very much.