Johny Fernandez: So the next time you buy a coffee, there’s probably a camera watching. Now, it’s recording everything, and most of the time nobody ever looks at it. Now, my guest thinks that those cameras are the most underused workers in the store. Now, his company turns them into AI that can spot an empty shelf, a slow drive-through, or a theft in progress. Now, last week, he raised more than $100 million to do it, with Goldman Sachs and Comcast on board. But first, a look at the company. The road from an idea to the opening bell starts right now.
There are millions of cameras watching businesses every day, but most of them have one job: record what happened after something goes wrong. Well, Mike Matta thinks AI can make them a lot smarter. He’s the CEO and co-founder of Solink, a company turning existing security cameras into what it calls AI vision intelligence. The idea: don’t just record the business, understand it.
Solink combines video with sales and other business data, using AI to flag things like an empty shelf, slow service, or potential theft. And it’s already finding scale. Solink says it manages half a million cameras across more than 35,000 locations in 65 countries, and now with more than $100 million in new funding.
Solink is betting this could become an entirely new software category. But can Solink turn cameras that businesses already own into AI that actually makes them more money? And can AI vision intelligence become the next big software market? From Startup to Stock Exchange, Solink CEO and co-founder Mike Matta is next.
All right. So joining us right now is Mike. Mike, thank you so much for joining us. I’m really excited to start this conversation. So let’s go ahead and start with cameras already on the wall in most stores. What are they actually doing all day?
Mike Matta: Yeah, it’s a great question. Maybe let me just zoom out a little bit. We [unclear: “started selling”] because we believe that there’s lots of data that sits in the physical world that just goes unharvested, untapped. When I first started, the opportunity was, I was working for a large telecom company in Canada. And what we did was we bought another company. We bought RadioShack in Canada, called Circuit City. And my first role was to go to all of these stores that we just acquired and understand what the shelves were like, what shoppers came in. And it was a lot of in-person work.
And the opportunity that we had was, how do we use the cameras that are there to understand what’s taking place? It was a lot of humans watching that footage, understanding and making decisions. And that’s typically what we’ve seen cameras do in the past: expose a feed, a human understands what the standard process is, and then acts on it. We’re in an age now where that’s radically transforming. You don’t need humans to watch 1% of the video. You can have AI watch all of it, and AI can help you make decisions as a business owner. Solink doesn’t collect any of that information. It’s really based on the owners, the practitioners that want to optimize their business, keep their employees safe, and give their guests the best treatment possible.
Johny Fernandez: So you started [unclear: “scaling”] back in 2009. You’re the CEO of a cloud video security company. When did you realize that the footage was worth a lot more than just security?
Mike Matta: Yeah. We initially started as a consulting company, but when we launched our product in 2016, 2017, our customers were primarily looking at the data not from a security perspective, because the data was sitting on their phone, in their pocket. They were able to pull it up and ask a hundred questions. And traditional entrepreneurs, if you go talk to these entrepreneurs, they have questions that bubble up every couple of minutes, and there’s lots of ways for them to get answers to those questions. We wanted to be that source.
So the initial customer that we had, we noticed that he was going to his donut cam all the time. And when we asked him, “Chris, why do you continue to go to that donut cam five, six times an hour?” he says, “Well, I know that when there’s donuts on the shelf, my order is going to be roughly about 17 times higher, or 17% higher, sorry.” And so that ability to take these ingrained knowledges that entrepreneurs have and express them through what’s happening in the physical world was the aha moment.
And what we realized is that everyone who walks into a business is typically there with intention. They’re there to transact or to buy or to look at product. But the reality is the number of people that walk in the door versus the number of people that leave with a transaction in place is not 100%. We understand that journey in the digital world, but the physical world has been very murky. And so what we really want to do is help enable entrepreneurs and practitioners to drive the outcome that they want to see in their business.
Johny Fernandez: So I want to talk about AI, because you call it AI vision intelligence. What does the camera understand now that it didn’t before?
Mike Matta: Great question. Traditionally, computer vision is not a new idea. It’s been around for decades, and you need to train a model to understand. You need to translate what we think of as a human, as, say, a person, in terms of shapes and pixels to a computer, and you feed it more shapes and pixels of what humans look like, and you ask it to define what a human is. Now, that reasoning of understanding a human waiting in line, and a human getting frustrated, a human sitting at a table, a human not enjoying their meal, all of that is what we interpret through the data with reasoning. We look at that and we don’t just see a human. We understand the intent that that human has.
And we believe that now, in the last couple of years, that opportunity, with not just computer vision but with vision language models, has really opened up. Traditionally, we’ve always talked about LLMs in the digital world. In the physical world, with video and images and visual properties, we’re going to start talking about [unclear: “the LMS”]. And that really gave us the opportunity to accelerate, not just giving access to humans to watch their businesses or see their businesses and react to them, but to have sidekicks and copilots to go in and drive an impact in their business.
Johny Fernandez: It’s truly fascinating, to be honest. So I want you to give us an example. Let’s say a Tim Hortons owner tied a full donut shelf to about a 32% lift in order value. How does the camera figure that out? Because that’s truly fascinating technology.
Mike Matta: Yeah. I mean, think about a general manager. When you have the best-performing stores that you have, generally it’s tied to two things. It’s tied to location. We all know location, location, location. It drives sales and performance. The second thing that we often hear from practitioners and operators is, “My people, my manager, my general manager, they’re great. They have the sixth sense of being able to understand what to do, when to do it, how to do it.” We believe that cameras can get all of that context. It sees everything, and the ability for that context to be taken in and understood, so that you can prioritize what needs to be actioned, is the opportunity that we think businesses in the physical world have.
An example would be, say you have customers coming in the door, and let’s say your shelves are empty. Let’s say your queues are long. What is the probability that those customers coming in the door are going to transact? Probably quite low. But understanding that your shelves are empty and your queue is long helps you get product out from the back of house to your storefront, and potentially manage your scheduling and have the right people working at the right time. These are some of the implications of being able to understand in real time, which, if you have a great manager, they kind of do as second nature. It’s really ingrained muscle memory of knowing when to go help out your cashiers, knowing when to go stock, take stock out onto the storefront. Those are things that great people will continue to do. But now they have that much more context to be able to do it by understanding what’s happening across my entire floor plan.
Johny Fernandez: So I want to do another example. Let’s say Take 5 Oil Change says that drive-offs fell about 78%. What is the system catching that a manager misses?
Mike Matta: Yeah. I think the biggest thing is the volume. When you’re in the physical world, it’s very chaotic. I would wager, if you tried to close your eyes and walk out of this building, even though you’ve walked these floors hundreds of times, your ability to do that just depends. It depends on where there’s stairs, where there are people. That chaotic nature of the physical world just isn’t easy to navigate. And what we’re doing is taking that information, understanding it, and turning it into structured outputs, outcomes that businesses care about.
For example, understanding your speed of service. When customers come in the door, they line up, they buy, they wait for their product. What if you order a product that takes three minutes to prepare, and I order a product that takes five minutes to prepare? Is that going to affect my queues? Absolutely. But at scale, they don’t understand that compounding effect, because of the chaotic nature. It’s very hard to measure. We used to be in a very deterministic world, where everything was measured in zeros and ones and you had to follow specific processes. Now there’s a reasoning-based model that can understand time and space and make decisions that humans would make if they were seeing it with their own two eyes.
Johny Fernandez: All right. So before we go on break, we have about a minute or so. More than 80% of retail sales have been offline. So why has the store been so much harder to measure than the website?
Mike Matta: Yeah, great question. Those are the exact same questions that we’re pulling on. And I think traditionally it’s been, again, because of the chaotic nature of that environment. It’s very hard to understand, and we’ve tried to measure it at certain points in certain places. For example, the point of sale. Traditionally, retailers [unclear: “call orders or traffic”], traffic equals orders. So the number of people that walk in your door equals the number of people that transact at your point of sale. But that’s not true. When traffic dips year over year, we call it potentially a macroeconomic impact. It’s not necessarily macro. It could be inside of your store. And so we help you convert every single buyer that comes into your store. How do we maximize that conversion impact that you have at the POS?
Johny Fernandez: Samsara puts sensors and AI dash cams in trucks, warehouses and worksites. Now, the physical economy never had the data that the digital one takes for granted. Samsara listed right here at the New York Stock Exchange in December of 2021, and today it’s valued at about $25 billion. In its last quarter, revenue grew 30% to $508 million, and annual recurring revenue topped $2.1 billion.
Now, here’s the contrast. Some cameras ride in the cab of a truck. Solink’s are already hanging on the wall of a store. It doesn’t sell the camera at all. Same bet, different floor. And that’s AI that can turn the physical world into data, and investors will pay for it. So we’re back with Mike Matta, co-founder and CEO of Solink. Mike, we were having a very interesting conversation about some of the results that you guys have. You raised more than $100 million, led by [unclear: “Virtue Capital”] and Goldman Sachs. So where does the money go?
Mike Matta: Great question. We’ve gone through this transition of infrastructure to software, and now we’re in a transition around intelligence, and we’re pointing all of our arrows there. Our biggest bet is around transitioning the software and the data that we collect over to an intelligence layer. The reality is, we collect over 25 times the amount of data that YouTube ingests. That’s a lot of data that humans can never look at. We need to be able to enable AI in the physical world so that our customers, our practitioners, can take advantage of it.
It starts in a security layer, but it evolves into the operational one very quickly. And more and more of the conversations that we’re having are with operationally minded folks that are focused on customer service, they’re focused on increasing top-line revenue. And that’s a lot of the conversation that we’re having these days.
Johny Fernandez: Now, that’s so fascinating. So I want to talk about Comcast Ventures. Comcast Ventures is new on the cap table, and Comcast will resell Solink to restaurants and [unclear: “realtors”]. So what does that distribution specifically mean for you?
Mike Matta: Yeah, I mean, look, we work with a lot of different channel partners. Comcast is an important partnership for us because it opens up not just the security angle but the opportunity to speak with IT practitioners and operations buyers who are really focused on, how do I take all of this information and codify it with AI, instead of having standard operating procedures where I need to have people do time and motion studies? You’ll recall when we started this segment, I told you about my early forays of how this idea came to light. I was doing a time and motion study at a store, trying to understand the display, the customers that are walking in, the behavior. That opportunity is now happening with operations and with the different executives that we speak to, both on the IT and operations side.
Johny Fernandez: So part of the plan is hardware, so your AI agents can run inside the store. And I think the question that people have about this is, why not keep it all in the cloud?
Mike Matta: Yeah, great question. To my point earlier, we ingest 25 times the amount of data that YouTube does. It’s a lot of data. And to push that up to the cloud, one, you lose control over it. There are latency implications. We do some of that today, but the idea is to take these models and push them down to the edge. There’s some of that that’s happening today. But that’s where a lot of this $100 million bet is going to go, is how we continue to accelerate our technology footprint.
Johny Fernandez: So this is really interesting, because a lot of workers hear that AI is watching the store and they automatically think surveillance. They think replacement. What do you tell the person behind the counter?
Mike Matta: Yeah, that’s a great question. That’s not really the business that we’re in. We’re not in the business of watching and pinpointing specific bad actors. The reality is, we’re looking at the actions that are being taken by lots of people, versus the individual actors of what’s happening. Now, obviously, we have security practitioners that want to reduce risk and crime and losses in their store, and so they’re looking at very specific actions that are being taken, and when they can react. However, we’re not introducing any new data. It’s very different than when certain vendors place a new camera, where they’re acquiring a new set of data. The data that we use was already there. Not only that, we don’t apply any biometrics. We don’t do anything to reveal the individual person.
Johny Fernandez: So I want to ask you a question about what you have said, because you’ve said that this could be a $1 trillion software company. So the big question is, who else is racing you for it?
Mike Matta: Yeah, great question. Look, I think right now a lot of the folks that are looking at the video market in particular are really focused on the segment around security. And if you take the security market in and of itself, the spend that happens between hardware and technology versus the spend that happens on labor and people is disjointed. For every dollar that gets spent on technology, there’s about 8 to $9 that gets spent on people. Look at, for example, security guards. Security guards is a $400 billion market annually, where people are standing around waiting for bad things to happen. And the reality is that humans just can’t process that much information as quickly as we want them to. Bad things may happen, but it’s less than 1% of the time. And so that opportunity, I believe, for an agent or an AI system to pick up instantly on what’s happening and recover from it is a lot faster and a lot more accurate. It’s also a lot cheaper than having folks that are standing around waiting for something bad to happen.
Johny Fernandez: Right. [unclear: “I’m sorry about it.”]
Mike Matta: Yeah, that’s half of it. The half is security. The other half of it, there’s half a trillion dollars on the operational side. Rather than the 1% of when bad things happen, there’s the 99% of when the business is operating smoothly. How do you optimize for that? How do you identify some of the gaps? That’s the other half a trillion dollars. And so when you add it all up, there’s half a trillion dollars for security, there’s half a trillion dollars for operations. And candidly, I think that’s actually a low estimate. I think at its peak, what we’re going to see is the opportunity for this to be an even bigger market.
Johny Fernandez: It’s so fascinating. So I want to tie this question into the next one, because security is top of mind, and we know that the holiday rush is just a few weeks away. So what do your customers want cameras watching for on the busiest days of the year? Obviously security is top of mind, but what else?
Mike Matta: Yeah, I think, again, there are three key things, or four key things rather, that our customers care about. One is converting every patron that walks into their store into an order. And there’s a messy middle there called the store. Do we have the right product? Are we checking customers out fast enough? That’s the first one. The second one is, are we maximizing basket size? When we greet customers, when we work with them within the store and we service them, what is the opportunity to increase the basket size? The third is being able to take some of that [unclear: “row”] work that happens with every associate behind the scenes and maximize people time. You have employees and associates that are great at working with people. Optimize for that. Don’t put them behind the scenes on a computer, counting inventory. A lot of the AI can do that work.
So those are really the big focus points that we’re helping practitioners solve. And especially in this holiday rush, those are going to be the key things that drive it. Do you have maximum transactions, the conversions that are happening? Are you optimizing for basket size? And are we optimizing for guest experience?
Johny Fernandez: All right, Mike. So to wrap this up, everyone in this chair always gets asked about the bell. I saw you walking in, and you were taking pictures. You were just enamored by the floor. So without a date, what does Solink have to do to prove that it belongs on this floor?
Mike Matta: Yeah. Look, I think every company that raises capital needs to be able to return liquidity. We’re preparing for it today. We’re preparing for the opportunity to do that. The timing of that, however, is going to be at some point in time. We believe there’s a big market here. This is a market that is worth capturing, and we’re building a company that’s going to be here for a very long time. So we’re aspirational. We absolutely want to do this, and want to be on that podium, ringing the bell. It will happen in the years to come.
Johny Fernandez: So let’s talk about the future. What’s the plan? What do the next two, five, ten years look like for you guys?
Mike Matta: Yeah, I think this transition from an infrastructure market to a software market to an intelligence market is brand new. We think about intelligence in the digital world. It also is very new, and it has grown very rapidly. I think in the next three years, AI for the physical world is going to be a topic that we hear a lot about, talk a lot about, and I believe Solink will be at the top of that.
Johny Fernandez: Well, Mike, I definitely want to thank you for your time. It’s definitely fascinating to see what you guys are doing. And obviously AI is everywhere, security is top of mind, so I love to see what you guys are doing and what the future holds.
