John Shegerian: This episode of the Impact podcast is brought to you by Closed Loop Partners. Closed Loop Partners is a leading circular economy investor in the United States with an extensive network of Fortune 500 corporate investors, family offices, institutional investors, industry experts, and impact partners. Closed Loops’ platform spans the arc of capital from venture capital to private equity, bridging gaps and fostering synergies to scale the circular economy. To find Closed Loop Partners, please go to www.closedlooppartners.com.

John: Welcome to another edition of the Impact podcast, a very special edition. We’ve got Ash Fontana with us today. He’s the Managing Director of Zetta, but more important for this discussion, he’s also the author of this book “The AI-First Company”. Ash, welcome to the Impact podcast.

Ash Fontana: Thank you for having me.

John: You know, the subtitle to your book is “How to Compete and Win with Artificial Intelligence”. That’s what I’m really excited about. You’ll explain today. But before we even get to your new great book, share a little bit about your backstory, Ash. How do you even get here as an AI expert?

Ash: Yeah, sure. I mean, I think it all started with curiosity, being early curious about what’s going on inside this computer I’m using when I was pretty young like four or five years old and you just stop pulling them apart. You pull apart one bit then another bit, and you learn more and more. So I got really curious about these things. If you spend enough time getting curious about computers and then if you were around a decade or so ago, you started seeing something that was both potentially really useful but also pretty hard to do, and that was AI.

It was a time about 10-15 years ago that people will call in the Big Data era. There was lots and lots of data coming online and through all these devices, we started carrying around with ourselves, etc. There weren’t really a lot of good ways or there wasn’t really a good methodology around managing getting insights from that data. And so we started building analytics tools and whatnot in the industry. But really the only way to manage those huge amounts of data and understand what’s going on is to use another form of intelligence that’s not our own. That’s why I got into AI and also from the investing side, that’s from the technology side.

From the investing side, I grew up in a family of entrepreneurs and we’re always obsessed with obsessing over and obsessed with what makes a good business. And so I got really into investing, you know, analyzing businesses and spent enough time analyzing business and come to an opinion on it. You’re like, well, I want to make money out of it and more importantly, I want to see if I’m right. So let’s put some skin in the game. So you buy some shares. I got really into investing as well. And there’s a job that combines investing and technology and it’s called venture capital. So, I found my way into that job very secure less. I think from the investing perspective, once you realize that there’s something out there that’s potentially very useful but also very hard, you try to understand it and try to invest in it. So that’s what we did in seven or eight years ago. I just completely focused my career on that.

John: Investing, that’s an interesting topic. I want to touch on that for just a second. If you were to say to me, “John, my favorite investor over the last seven or eight years has been…”. Tell me your favorite investment and then why.

Ash: Favorite investment?

John: Yes.

Ash: It’s in a company that is very private, so I can’t say much about them. But the investment, I can speak about what they do and it’s a really cool investment because it goes to the power of AI and that it’s using AI to do something we can’t do very well, which is predicted demand and supply of electricity on all the major grids in the US. So, 24 hours ahead, how much energy are we going to use once we going to demand? Therefore, how much should we make so that we don’t just produce too much? It’s really cool because that’s a very complex system like the energy grid, the energy system. There’s also a lot of complexity around understanding what humans are going to do 24 hours ahead, how much energy they’re going to need and use. And you really need to use AI to bring all of those data sources that will form that prediction together.

The other reason, so it’s a really cool machine learning and Artificial Intelligence problem. It’s a very good problem to solve in terms of you do get compensated for it well if you do. But most importantly, it’s saving huge amount of CO2 from entering the atmosphere. You know, just getting a head a little bit and having slightly better predictions means we don’t produce more today. That’s the sort of leverage that AI gives us. It gives us leverage over time. It doesn’t give us leverage over physical space like a hammer does or even an Archimedean lever. It gives us a bit of leverage over our intellect, like it helps us calculate things and whatnot. But really the power of it is giving us leverage over time, seeing around the corner, making a really accurate prediction about tomorrow so that we can make a better decision today.

John: Is that kind of company that you invested in is one of the other benefits besides preventing CO2 emissions and also overproducing needs, also the prevention of brownouts? Is that also?

Ash: Yes, exactly. Yeah. And so, this is a really good question, because this is why this whole system was developed and mandated by the government in the 70s and 80s. This system of making predictions a day ahead and then consolidating everyone’s predictions and putting it back onto the grid to say this is how much you should produce tomorrow. So yeah, on the one side, it’s about making sure we produce enough power. On the other side, it’s about making sure we don’t produce too much power.

John: Ash, you’re here today and you’ve written this great book “The AI-First Company: How to Compete and Win with Artificial Intelligence”. And as I was sharing with you off the air a little bit and for our listeners who want to find Ash also and this great book, besides on Amazon and other great booksellers, you could go to the aifirstcompany.com or ai-first.org.

You know, I was sharing with you my October 19 story where I was sitting in a room in Armenia and there was a great guy that I’ve never seen in my life before talking about the application of AI to Medicine, Biotechnology, Healthcare, and that future. It was fascinating to me, but I just thought it was literally Star Wars. Five or six months later, we get hit with the pandemic and this gentleman turned out to be Noubar Afeyan, the Chairman of Moderna who solves one of the world’s greatest crises ever.

I’m a convert and I don’t have a lot of knowledge, so I love you to share your knowledge at different applications of AI and why leveraging AI to compete in the competitive business world that we are today is the way to go.

Ash: Yeah. Look, the two broad things AI can do is prediction, seeing around the corner, and automation. And if we just take that Moderna as our example, and I won’t go into too much detail about how they used AI to get this vaccine out the door because that’s a whole another conversation. There’s all these amazing ways AI contributed to the development of the vaccines that were lucky enough to have today. But broadly, the prediction side of it was, and lots of drug companies do this, they use AI to basically search across all these possible combinations to get a computer while we sleep because we can’t just sit there doing calculations all day long as quickly as a computer can.

Get a computer to ingest a bit of information from us. We know that the mechanism of action of the drug and the body is this, so we give it sort of a structure, get a lot of data from scientific reports, and whatever else, and research and predict which combination of molecules is going to be the most successful at getting this thing done, addressing this disease or whatever it is. So there’s like a prediction element, a way you can use AI for prediction in drug development.

Then there’s the automation element and that just really comes down to testing and production. So when you are running a bioreactor, using AI to figure out how to optimize the use of that bioreactor is really effective. On a production line, analyzing all the data coming out of the sensors on the production line. You know, the vials were a little bit big or a little bit small. They weren’t in the right shape or whatnot. You can tell all of these things using sensors from cameras to weight sensors and whatnot, then you can analyze it a super large scale to figure out, okay, whenever we have a breakdown on the line it’s because this thing happened five steps ahead. That’s the correlation that it can notice on a very large scale.

Look, just following on from Moderna as an example, it’s pretty clear AI is very useful to the big pharmaceutical companies in so many ways both on the prediction and the automation side. But this is true in so many industries. Look, you can just pick an industry and I can go on and on about all the potential applications of AI in the industry. The reality is and the reason I wrote this book is we’re only just halfway through what I call the AI-First century, and that is we’re at the point where a lot of the research has been done to build these things, to build these AIs. A lot of the infrastructure has been built by these very good companies like Google, and Amazon, and Microsoft to let us run these things at large scale. And a lot of the data is available because we’ve been collecting it for a little while now. A lot of companies just accidentally collect data. No matter what industry you’re in, you’ve got a big trove of data somewhere.

All of these things are sort of come together to let us make these AIs that learn from that data and learn on a computational substrate. So they learn on computers that can run on computers all day long while we sleep. They can learn some pretty interesting things that we don’t see or we don’t want. They can see things we can’t see. And so, we’re halfway through this century and we’re really at the point where applications of AI are very real and very broad. I want everyone to stop putting AI-first in the conversation, whether they’re in retail industry, manufacturing, healthcare, whatever industry they re in travel, hospitality, anything.

I want them to stop putting AI-first in the conversation so they can find these applications because the reality is AI is not going to find it by itself. We need to teach AI how to do things. We need to give it our expertise that we have in a certain industry. We know that if the concrete goes down three days later on a construction site, it blows out the schedule by three week, to just give a really simple example. And so if the camera sees the concrete go down on a certain date then it should feedback the information that the schedule is going to blow out. We need to give it that information so it can learn and help us make better decisions.

John: Ash, two things, going back to your investment thesis, when you invest, I assume that every company that you decide to invest in is also going to be leveraging the use of AI in some way, shape, or form to evolve thesis of that and the original mission of the company.

Ash: Yeah.

John: Got it.

Ash: Yeah, that’s right. You know, I specialize in finding companies that’s building their competitive advantage primarily through data and then compounding and making it stronger with AI. And so, they start with the data set and I analyze that data set. I try to understand, is it unique data? Is it valuable data? Is it the sort of data that was really hard to get, or is it the sort of data that actually someone else can either easily get or get data very similar to it? So, I try to understand, okay, do they have something valuable now and then can it feed an AI and get more and more valuable at the time? Because the thing about AI is they generate their own data over time.

Once you sort of say feed it data, it trains on that data. It says, all right, I’ve seen these thousand images, I know what a cat looks like. And then you feed at the next image and it says, it’s got a cat in it or not and then you correct it. Sometimes it’s wrong, and human can correct it. When you correct it, it gets better. It learns from that. So, it gets better and better over time, and this is the power of AI. Once you get it going, once you put the effort into collecting the right data set, training the model and building a system to monitor it and corrected over time, it will get better and better with every turn of the crank, so to speak.

That’s really cool, and that’s why our focus on these companies because if you get it right at the beginning, if you put AI first when you’re thinking about the design of the product, when you’re thinking about the people you hire, where you allocate capital into acquiring unique data, if you get it right at the start and then let it run, in one year’s time you’re not a little bit better than the competition, you’re ten times better faster cheaper than the competition. It compounds so quickly and you get really far ahead really quickly. That’s why I like to invest in these companies because they have a defensible position in the market and they can win for decades.

John: You know, we have a lot of young entrepreneurs and also just other entrepreneurs like myself that will watch this episode and listen to our show. Again, it’s Ash Fontana, the AI-First Company. I’m embarrassed to tell you Ash that I have used AI here with tremendous results but only one application. We use it as part of our optical sorting and it’s become a massive tool for us to create more “liberated commodities” because it’s a self-learning tool, like you said, with one data set of pulling pollutants and other deterrents out of our commodity process through the optical sorter, and it works tremendously. But we talked about Moderna a little bit, talk about some other examples too of companies that are winning today putting AI first.

Ash: Yeah, that’s a great example, actually. And they’re very many variations of that which is put a camera somewhere, ask it to recognize one thing, but very reliably over and over again. In our sorting for recycling, find all the plastic bottles in this pile. Or in a supermarket, is there passed on the shelf at the moment? Was this self need to be restocked in a warehouse? Is this part of the warehouse full or empty? Is there someone walking behind the forklift right now? Lots of examples like that. So, I can go on and on and give you examples, but the point is so many applications of computer vision like that.

A company I can give a really concrete example on is called Focal Systems. And what they do is they help retailers, supermarkets, figure out if the shelves are full. That’s it. At the moment, the way that’s done is someone walks around with a clipboard at 4 a.m. every night and figures out if the shelves are full or not. The problem with that is: one, it’s really tedious; two, they have to wait till everyone’s out at the store to do it properly and efficiently; and three, it’s too late. The shelf might have been empty at 10 a.m. and they don’t know until much, much later, and all the sales that could have happened in the meantime were missed. And not just all the sales of like one thing that ran out. The pasta that ran out. The rice that ran out. It’s the case that if someone walks into a supermarket, fills up their basket a little bit, gets to the thing they really came to the supermarket for like a bag of rice and it’s not there, they just leave the bus on the floor and walk out. You lose all the whole basket. You don’t just lose. It’s really weird the people behave this way, but they do. You don’t just lose the sale of the rice, you lose the wholesale, so to speak.

And so, what they do is they put cameras all over these stores and they’re just constantly looking at the shelves, and they’re figuring out what’s in stock from what’s not in stock, and then alerting the people at the back of the store when something runs out of stock to restock. And that’s really helpful. It’s a really hard problem to solve though because there are lots of products in a supermarket. There are lots of things that happen in supermarkets. People mix things up and move things around, leave their kids sitting on shelves, whatever. And so, it’s a sort of a hard problem to solve, but once you solve it, it’s really valuable and got really tangible results both in terms of making sure that these grocers can sell what they need to sell, but people get what they want when they walk into the store.

And then there’s all these other applications of this, some of which we’ve got a taste of by looking at these new Amazon stores which is once you know what all the products look like on a shelf, well, actually you can see people pick them up off the shelf and then just automatically charge them for it as they walk out the store. They don’t even have to go to the checkout. Or if you do want them to go through checkout just to double-check, you can make that whole self-checkout process a lot more efficient. We’ve all been through a self-checkout and had to find the code for red apples and it takes forever and you’re pressing all these buttons and the machines talking at you and it’s not a great experience. But with really good vision systems, you don’t have to do that. You can just walk straight out or just one click confirm off you go.

So yeah, there are lots of other ways you can leverage that initial data set around identified products and the models that are built on that data set to do really cool stuff like auto-checkout of the supermarket.

John: Ash, you know, I was lucky. The guy who brought me the optical sorting and AI technology was an investor, a long-time friend of mine, an investor in the company. But take a guy like me. I’m 58 years old, running a very big and growing brand. I want to use AI more and more because I’m convinced vis-à-vis your book and other thought leaders like you and the Moderna application, the one you just said about, the Amazon and the supermarkets. But how do I attract the talent that can then help inform me where we should be using AI in a company like ours and where can other entrepreneurs find that kind of talent as well?

Ash: Yeah, that’s a great question. I think a lot of people have the perception that you need to be able to afford and also just get a whole bunch of PhDs to work on this stuff. And the reality is, that’s not true. It’s not true for a bunch of reasons. One, a lot of the big companies, Amazon, Google, Microsoft, offers some really good tools that work really well out of the box to help you build your own models. They give really nice interfaces to let you build this stuff. So that’s a good thing to leverage to start with. The second thing is that a lot of AI starts as and the reality is a lot of AI starts as a pretty simple statistical model. It doesn’t have to start as some being machine learning model or network of models to start with. A lot of people can do statistics. A lot of people that don’t have PhDs have done a little bit of statistics here and there. So, that’s another thing.

The third thing is, you can find these people that can actually work at a very high level with machine learning in all sorts of disciplines. It turns out that you study a lot of statistics and probability and Mathematics and other fields, in Biology, in Geology, in physics. You study all of this stuff. So, it’s not the case that you have to just get this year’s top Computer Science grad to work for you to build the stuff. You can get 20 years ago fourth place Geology grad from whatever University and say, “Hey, there’s this new technology, you can probably work on it because you’ve got a really strong background in Geostatistics or whatever else.”

I go through this in the book, where to look for talent, what to do to support them, how to put a team together, how to manage on the one hand a bit of research into what could be done, and then the engineering, you’re getting it into the real world. The punch line is, you can probably use a lot of people that are already working for you today to build this stuff with the combination of their solid quantitative background and the tools available to them from a lot of the big companies.

John: Ash, one last question before I let you go. As an entrepreneur we’re taught to build moats around our company. Moats. Differentiators. But you have a different analysis of why loops are more important than moats. Can you explain your definition of loops, not moats for our for our listeners and viewers?

Ash: Yeah, for sure. As you point out, for years we’ve used this concept of a moat, and that’s a very static concept. It’s like I enjoy a 20-year advantage over my competitors because I’ve got a pattern that last 20 years. That’s a good concept and makes sense. It’s a good metaphor for competitive advantage, but it just doesn’t really map well to what AI gives you. The sort of competitive advantage that really helps you generate and build over time. There are two reasons. One, as I was sort of saying before, every time you run the AI system, you get a bit of feedback. You say, okay, this is something that’s in this image. This is something I’ve noticed through the optical sorter that there’s a contaminant in this image or not. As a human or as something else involved in the system, you say yes, no. And so, every time you run that, you get a bit of feedback under the output of the AI. And the next time it runs, the AI incorporates that to make a better prediction or a better identification of an object or something like that. Run the object recognition system better.

So, it gets better every time it loops around. It’s going around, and it’s increasing in its utility every time. So it’s sort of not a static concept, it’s a widening moat over time and it’s widening with every single turn of the crank. That’s why the loops a bit of a better concept than moats. Again, to get to the whole point here, AI-first companies have this amazing competitive advantage that are stronger than all the other ones really I think that we see day-to-day. Stronger than brands. Stronger than just having a patent over a compound or whatnot. And that’s why they’re the only trillion-dollar companies out there today.

If you look at companies like Google that got this loop going really early on that were really intentional about putting AI first in their products and the design of their products and in their strategy, that really deliberately collected a lot of valuable data often by giving stuff away for free. They give away Google Maps for free to most of us and they collect a lot of really valuable data from that. So, they’re really very deliberate about it. And today, you know what, they’re a trillion dollar company because they have the best system and no one can catch up to them in terms of developing the best ad serving system.

John: Good reason. It’s Ash Fontana. You can find him at ai-first.org. You can find his great book “The AI-First Company: How to Compete and Win with Artificial Intelligence” on amazon.com, on his own website, Barnes & Noble, and other great bookstores. Read this book. It could change your business and your life forever. Ash, thank you for joining us on the Impact podcast. Obviously, you’re making a great impact. You’re making the world a better place. You’re always welcome back here. We’re really grateful for your time today.

Ash: Thank you so much, John. Thank you for hosting this and getting these ideas in front of people and inspiring them.

John: This edition of the Impact podcast is brought to you by the Marketing Masters. The Marketing Masters is a boutique marketing agency offering website development and digital marketing services to small and medium businesses across America. For more information on how they can help you grow your business online, please visit the marketingmasters.com.