#095 icetana AI | We Write AI That Keeps People Safe

Kevin Brown:
The cameras are nice because it gives us an idea about how many people we’re keeping safe. We've got stories of actually saving people's lives. These are the stories that we like.

Intro:
Welcome to the Midstage Startup Momentum podcast. Each week, we interview up-and-coming founders of some of the fastest-growing mid-stage startups across the world. Your host is Roland Siebelink, who will share some of his own experience helping startups scale from 10 to 1000 people in a few years. Here is Roland.

Roland Siebelink:
Hello and welcome to the MidStage Startup Momentum Podcast. This is Roland Siebelink, and I am a coach and ally to many of the fastest-growing startup companies around the world. And around the world, we take quite literally today because with us is Kevin Brown. He's the founder and CEO of icetana AI, dialing in all the way from Perth, Australia. Hello, Kevin, how are you?

Kevin Brown:
Good, Roland. Happy to be here.

Roland Siebelink:
I'm very happy to have you on our podcast as well. Often, I have founders that I just have a first conversation with, but Kevin, you and I go way back, right?

Kevin Brown:
Yes, we do.

Roland Siebelink:
Tell the audience the story, if you will.

Kevin Brown:
Okay. We had a growing startup that was moving into scale-up. We had a lot of growth challenges and Roland came in and helped to stabilize it, put structure in place, and using the scale-up methodology and some pretty deep industry experience, and a cheery personality, helped us take a business from probably about half a million dollars a week to about $1.4 million dollars a day.

And I think the team grew - I think probably when you started, we were probably about 200. 

Roland Siebelink:
Oh, no. You were 80 or something when I started. 

Kevin Brown:
It's funny, cause when I started, there were 70, and then I had to make it 30. And then it went up to 80. And then I think by the time you left, it was probably closer to 500. You were instrumental in the scale. 

Roland Siebelink:
I actually think they were over 1,000 when I left. That was just three years later.

Kevin Brown:
I use the scale-up methodology still. We use the strategic framework at icetana AI, we use ROCKS, we use the new leadership, team members, and the scale-up book. 

Roland Siebelink
As we're talking about it, Kevin, what would be your one-minute summary for founders? Why should they look into something like a scale-up methodology?

Kevin Brown:
Because I think that the skills that you need to create the embers or the flame of a business are very different from the skills that you need to take it to the next level. There's an entrepreneurial lack of structure that just works or doesn't. 

And then once you start dealing with scale, I think often those creative entrepreneurs struggle with the structure and need a bit of help around how to plan, how to scale, how to prioritize, and also how to think about the business like a machine - figuring out the flywheel, understanding the unit economics.

Roland Siebelink:
Let's move on and talk to you about your current company. You left to ultimately found icetana AI, is that right?

Kevin Brown:
Well, I didn't found it, Roland. This is the interesting part of it. I don't know if you know this, but before VGW, I was involved in another business that sold for a billion dollars. And VGW was the last - probably about $7 billion at the moment - and when I came out of VGW, I met the CEO, and the CEO was a caretaker CEO for icetana AI. It's actually been around for 15 years. 

Roland Siebelink:
You re-founded it, in a way.

Kevin Brown:
Yeah. I consider myself to be the founder at the moment. This was one of the reasons why the business didn't take off 15 years ago, because it didn't really have a founder. It came from a general-purpose anomaly detection algorithm that applied to video. It came through at the local university, got commercialized, and raised some money. And then they brought in a career CEO.

The career CEO was more sales, less tech. Fifteen years ago is like 150 years ago in AI. It was actually remarkable what they achieved. But by the time I came, the business was close to dying. And it was one of these ideas where you're like, “That seems like a good idea. Why is it not working?” 

And just the application. Maybe this is a good segue. We write AI software that keeps people safe across thousands of CCTV cameras. We specialize in self-learning AI. The problem that we fix is that security teams are often drinking from the fire hose of video. Our largest customer at the moment has 8,000 cameras - and this is across 17 high-end shopping malls in the Middle East. Every five hours, those cameras produce more video footage than the entire Netflix catalog. They have a legislative requirement to monitor all the cameras in real time.

And this was our first application 15 years ago of this emerging technology. When I joined, there were some cultural issues. The last dying moments of a business, there are either the fanatics or the people who can't get another job. And I'm sorry if I sound harsh here, but I had to sift out the fanatics and keep them on and change the culture. 

I changed the culture. I flattened the technology. Our mission was to retain this 8,000-camera customer. And this was for me. I was like, “If I can do this, this is a big business. If I can't, then I'll go and do something else.” And we did it.

Roland Siebelink:
Was it primarily a product and people challenge when you joined?

Kevin Brown:
It was a cultural and product challenge. The culture was terrible because it wasn't succeeding and everyone was bickering. One of the first things I had to do was put values in place, and a mission. There was no mission, there was no values. I put truth in as a value that still stands strong here because there was a real culture of not telling the truth. It was like, “He did it, he did it.”

It took me about six months to a year to figure out what was happening with the people and the technology. And I'm reasonably geeky - I've had a keen interest in AI since I was a youngster. I was like, “I feel like we can do a better job of this with new tech.” And sometimes, we had to wait for new tech to appear.

For the last five years, we've been building the tech. We've got strong growth in the Middle East. They've got lots of cameras. They've got legislative requirements to monitor them. They've got no problem replacing people with AI. 

Whereas in Australia, the security businesses here, the business model is based on head count. Giving them the tools to replace the headcount, it doesn't compute, or it didn't compute. But now these guys are coming - the local police, local council. And we don't even have a go-to market for Australia, Roland. Our go-to market is in the Middle East, it's in Japan, and we're just dipping our toe into the US at the moment. But the market came alive. The people are coming and saying, “Hey, does anyone have any AI around here?” 

Roland Siebelink:
Yeah, now people can imagine it. And they start identifying all those routine jobs that AI could most easily replace.

Kevin Brown:
This is it. This is the thing. And we've got a head start because we've been doing this for 15 years. We've built a handful of generalized machine learning algorithms that we've used 700 million hours of CCTV footage to generalize. We switch it on. It learns what normal looks like. It builds behavioral models across multiple dimensions for all the cameras, and then it reports on the unusual or interesting things. And after a week, it understands the flow, the weekly cycles. 

Roland Siebelink:
Essentially, it's threat detection, right? It can detect threats and things that are unusual that need to be looked into.

Kevin Brown:
Yeah, absolutely. 

We've got machine learning algorithms that take a full day's worth of footage and reduce it to maybe a percent of a percent. And then we double-check with the GPT. And as a consequence, we can have like one person monitor 2000-3000 cameras in real time.

Roland Siebelink
Wow. That's huge. That's huge savings. 

Roland Siebelink:
Let's talk a bit about your go-to-market. You were lucky to already have the customer with the 8,000 cameras when you came in. How have you been thinking about your go-to-market since? Have you grown the business? Have you specialized in some niches versus doing multiple verticals? How do you think about all that?

Kevin Brown:
First of all, enterprise software is much more difficult than B2C. The B2C - the time on VGW was often four weeks or six weeks until you made your money back. And you could test messages in the funnel.

Roland Siebelink:
That was also an exceptional distance model. That's not general for a B2C, I would say.

Kevin Brown:
Yeah, that's true. But enterprise software is more difficult, longer lead cycles. And when you're looking for product-market-fit, it's expensive to spread. We found a niche in the Middle East. We rebuilt the sales team as well, bringing in an experienced sales team.

Recently, we just got a purchase order in for $1.8 million. Our revenue is pretty low, it's like $2.2 million. We've doubled the revenue with one purchase order for a smart city in the Middle East. And when we do a good job of that, then there are another few deals in the pipeline that are three or four times larger, and there are a lot of cameras there. That's one go-to market. 

We've got partnerships in Japan. We're chatting at the moment with another distributor who has global reach in robotics. They see icetana AI as a key part of their consolidated security bundle. We've got that. We've got emerging interest in the US. People are bypassing the incumbents in the US and coming to us because we have the expertise. At the moment, we're putting the final touches on the go-to-market in the US. 

And Australia has woken up. I've got 10,000 camera deals in the pipeline from large organizations here who realize not only that AI can augment their existing security team, but also increase their security coverage. Some casinos here have got 4,000 cameras, and their security team can only look at 200 cameras. And even then, humans get distracted after 23 minutes.

It's great now that the business has come to us. I'm actually in the middle of building out. We're on a hiring spree just now. I'm hiring more engineers, more salespeople, more operations guys.

Roland Siebelink:
How big is your team now, Kevin?

Kevin Brown:
At the moment, there are only 21 of us. We're a pretty small team. The machine learning and product teams are eight deep. But we've got some pretty smart guys here. And also, the IP that we've got is in the generalized algorithms. And also, because enterprise software has such long sales cycles, it was important for me to keep a small team.

Roland Siebelink:
You don’t want to overextend.

Kevin Brown:
Yeah, and often when you're dealing with these large customers, you need to be ready to walk away and negotiate. I suppose everyone's interested in negotiating.

Roland Siebelink:
Okay, very cool. But your sales strategy is primarily based on distributors, it sounds like.

Kevin Brown:
Yeah, but sometimes the distributors need a bit - sometimes you need to take a leap, give it to the distributor, make the money for them, and then get things going. 

Roland Siebelink:
And then you mentioned, in America especially, people are going around the incumbents. Who do you see as incumbents, and are they your competitors, or are you disrupting them?

Kevin Brown:
The competitors over there - we've got Ambient AI, and those guys raised about $54 million through Andreessen Horowitz. These three guys were in stealth from Stanford for three years. They've got a bottom-up approach. They've got a great product, but they've got a bottom-up approach, where they're more precise. We're more accurate and we come top down.

We don't specifically find fights. But we find people who have fallen over, people who are running. We're more likely to find things that you hadn't thought about or planned. Whereas the Ambient AI guys have a list of 35 specific use cases that they'll find. But we compete with them. We are more efficient on compute. We can fit 400 cameras on a server, and their tech requires a lot more grunt, so we scale better and we're self-learning.

We've got other guys like Spot.ai, they are small to medium. We started at the top end, which is probably what you shouldn't do. But they started at the bottom end, and they're making their way up. 

They're educating a lot of the market over there. This is wonderful for us to have these guys educate. And also, it's wonderful for me as a CEO, we're getting movement now. The investors are being educated on autonomous security and AI. It's all good.

Roland Siebelink:
How ambitious are you for icetana AI? How big do you want it to be in five to 10 years' time?

Kevin Brown:
Five to 10 years - this is maybe egotistical, but I'm gonna say this anyway - I've been involved now in two unicorn billion-plus businesses out of Perth, and the market for this is massive. We've only got 20,000 cameras on the license. And our aspirations - our near-term aspirations - 

is 100,000 cameras. 

We've been plugging away at that number. But now the go-to-market - reaching 100,000 cameras seems so plausible that we were looking at a million. And when we look at the million cameras - and again, this is not crazy. There are people, there are competitors who have four million cameras on their license, so the numbers are not wild. 

For me, that's where I'm heading. I'm heading to a million through 100,000, and all the numbers between there. When we do that, this becomes a significant business that again, comes out of little old Perth. And it's a software business. We need to be clever. We need to scale. That's where I expect it to be. I expect this to be pretty big.

Roland Siebelink:
That's amazing. And a great example to other founders of deciding on one widget. How do you measure the impact of the business in non-financial terms? And in your case, it's just “I count the cameras.” And you don't have to be confused about: is it customers, is it accounts, is it sites? Just pick one and go with it.

Kevin Brown:
It's funny. I was chatting to some remote video monitoring services in the US, and they don't count cameras; they count revenue. And my leadership team was like, “Hey, why don't we count revenue?” And I was like, “Because the abstraction makes things much easier to share all your numbers and your growth and your planning.” It's a strong scale-up move to find the unit economics.

Roland Siebelink:
It's that, and also, the unit economics is the real reason. But I would also say from our joint experience, it's also so much more inspiring for many people on the team, especially lower ranks, because nobody's going to get inspired by more dollars.

Kevin Brown:
The cameras is nice because it gives us an idea about how many people we’re keeping safe. We've got stories of actually saving people's lives. These are the stories that we like to hear and share. 

Roland Siebelink:
And I love the purpose that you've built there, about keeping people safe, and more and more stories there will inspire people to know they're doing something good in the world.

Kevin Brown:
Yeah, exactly. I sleep like a baby knowing that my AI is keeping people safe.

Roland Siebelink:
That's amazing. Well, Kevin, this has been a pleasure. Now that you've gotten yet one more company under your belt and you've got so much experience, I'm sure you talk to younger founders, people who come behind you every now and then. What's the typical advice you share with them?

Kevin Brown:
Just do things you enjoy. Don't do things that you don't. Also, from a founding perspective, because I've got one of the young guys here who is leaving to become an entrepreneur. Find a group of people that you want to help. Don't go looking for an idea. Go looking for a group of people that have a problem that you can fix. Keep doing that and everything will figure itself out.

Roland Siebelink:
Yeah, I think that's really good advice, because it immediately helps you understand, “What would the customer think about this?” And you start thinking in terms of pains to solve. Also, what is good enough? I love that approach. Really good.

Very cool. Well, thank you so much, Kevin Brown, the founder or refounder and CEO of icetana AI, dialing in all the way from Perth, Australia. Thank you. It's been a pleasure. Any last words for the audience?

Kevin Brown:
If you find yourself with a startup and you find yourself struggling with scale more than business, then give Roland a shout. He's helped us build a billion-dollar business, and there's a few more in there for him to help you out with.

Roland Siebelink:
Wow, thank you. Nobody's going to believe this, but this was completely unprompted. I really appreciate this, Kevin. Thank you so much. 

And for the audience, we will have another episode for you ready next week. Thank you, everyone.

Outro:
Like what you heard? Subscribe to this podcast and leave a review. Tune in next time for more hot startups and interviews with some of the highest-momentum startup founders in tech today.

/