In this episode of The Gradient Ascent Podcast, CIBC’s Paul McKinlay and Daniel Lee sit down with George Babu and Gideon Hayden to explore what it takes to build enduring AI companies today, from cheaper experimentation and distributed teams to smarter go-to-market strategies. They also dig into a bigger question for Canada’s tech ecosystem: how do we keep top founders building at home when talent, capital and ambition move globally? Keep an eye out for Part 2: Why Enterprise AI Adoption Is Still So Hard, coming soon.
Daniel Lee: Today we've got two builders in the room coming at the same thing from different angles. George Babu has built hard things like robotics and AI. He's also co-founded and sold a Canadian company, to Salesforce, and then co-founded and sold an American one to Ocado, a British company. Gideon Hayden also co-founded a business straight out of university and sold it. But today he writes cheques at the Leaders Fund. Gideon recently went to Parliament to make the argument that we're losing too many of our best founders in Canada. And so on today's show, we're going to be talking about what's worth building today, what it costs, and whether or not the best of it stays in Canada. Welcome to the show, gentlemen.
George Babu: Thanks for having us.
Gideon Hayden: Thank you.
Paul McKinlay: We've all known each other for a long time, so this is kind of a
fun, friendly conversation. We're going to talk about a whole bunch of things. I'm going to kick an intro over to both of you guys. George, you've been an investor, more a founder, working on a new company now. We're going to talk a lot about it later, like give us a 15 seconds on what you're up to and a quick intro on you.
George Babu: Absolutely. The latest company I’m building is using AI and agents to help us build infrastructure faster.
Paul McKinlay: Crisp, I like it. And Gideon, also a founder but spent more time on the investor side of things and that's your full time gig now. Like tell us the 15 seconds on you know, what you guys invest in at Leaders. And then we got a bunch of things we're going to kick off with.
Gideon Hayden: Yeah, well firstly, thanks for having me and excited to be here. I’m one of the co-founders of the Leaders Fund. We started the fund about ten years ago. What myself and my two co-founders have in common is we're all operators, entrepreneurs, before we started the fund. And over the last ten years we've been investing a lot in enterprise software, in enterprise software, cybersecurity specifically, obviously doing a lot more in AI now and cloud infrastructure.
Paul McKinlay: Fun. Part of the connection too is, George, you used to work with your partners at Rypple
George Babu: That’s right. Yes.
Paul McKinlay: That Dave and your partners worked at together. So there's a big circle of friends.
George Babu: I worked with Steve DeBacco and David Stein.
Gideon Hayden: And George and I worked at OMERS for a little bit.
Paul McKinlay: That's right. Cool. We're going to get into it. And we’re talking about a couple of different topics. Daniel can you kick us off on the first one?
Daniel Lee: Yeah, absolutely. So, Gideon you've spent some time studying how and where our Canadian companies get built and funded and you took some of your research over to Parliament. Give us the punchline on the state of play right now. Is it easier or harder to build enduring companies here in Canada? Yeah.
Gideon Hayden: Yeah. Well, I guess this all kicked off about a year ago, when I'm sure all of us have been seeing the same kind of rhetoric online and people talking about the fact that we have this brain drain challenge and in this country. And I had seen that enough times where I got a little frustrated and said, “Well, what does the data actually say? You know, like everyone's talking about vibes and no one's talking about data.” And so we worked with one of our data partners to try and answer that question of the Canadian founded companies that that are being created today, how many of them could have been started in Canada? How many of them are starting in Canada? And it kind of started as just like a curious you know, off the side of my desk, let's see what the data says. And what was very interesting, I think it's pretty rare you see such a clear story being told in data right off the get go. So we received the data, we crunched it. And what it told us was from kind of 2015 to 2020, about 75% of Canadian founded start-ups were started here in Canada. start to basically flip, right? And it got to a point where in 2025, probably in the low 30% of Canadian founded companies were started here and a majority were started in the US. So, it got us asking questions around, you know, is that good? Is that bad? I mean, I think we can probably all agree we'd like more of those companies to be started here, but that led us down a path of just doing more research and putting out a report on that topic. So it's been an interesting journey for sure.
Daniel Lee: You came up with a number of pretty interesting solutions in your research. Of all these fixes, you know, whether its capital gains or angel whether it’s capital gains or angel tax credits or Buy Red procurement, which one do you think actually moves the needle the most in the short term?
Gideon Hayden: It's probably important to say that if you're building a company here in Canada, you need to be elsewhere as well. Like you can't just, you know, build in a vacuum here. I don't blame any founder for going and starting their company elsewhere. Silicon Valley is a gravitational force. We're not we're not going to replicate that. But I think if we can create an environment for entrepreneurs that convinces more of them to build here, I think that's healthy. So I think it starts with at least matching or becoming more competitive with the regulatory regime that entrepreneurs face in the US versus here. An easy one is just long-term capital gains exemption. I think an angel investor tax credit could be really powerful if all of us could be taking some of our hard earned capital and putting that into early stage start-ups and creating more activity there. And that was, that was given back to us as a tax credit in some form. I think that could be really interesting. That first kind of that first angel round is one that is pretty hard for a lot of young founders to come by. So those two were pretty excited about. So those two we’re pretty excited about.
Paul McKinlay: That’s when like the ecosystem starts to have a flywheel is when you have like a big outcome, people make wealth in the sector and then they instinctually just like, yeah, I'm going to go back my former colleagues that are starting new businesses or whatever. Like that's one of the things you see in the Valley that is so powerful. To enable that here would be amazing, for sure.
Gideon Hayden: Yeah, 100%. I think it's just that culture of like, “Yeah, this guy's got this person has an idea, let's give them some cash and let's see what happens.” Right. We need we definitely need more of that.
Paul McKinlay: Yeah, totally.
Daniel Lee: Speaking of the back and forth between Canada and the Valley, George.
George Babu: Yes?
Daniel Lee: You've built on both sides of the border.
George Babu: I have.
Daniel Lee: First Rypple, here in Toronto, acquired by Salesforce. And then Kindred, you founded in San Francisco.
George Babu: Founded in Toronto. But I went down to San Francisco.
Daniel Lee: Sorry, then you relocated to San Francisco and ultimately got acquired by Ocado. And now you're back here in Toronto building Aire Labs.
George Babu: Yes.
Daniel Lee: What brings you back?
George Babu: Great question. I mean, family, love, right? I love my wife. She didn't like the US, so we had to come back here. Obviously, I had to think about I wanted to do another start-up. Can I do it in Canada? Should I stay in the US? Ultimately, I decided to do both. I ended up building a team that's distributed. So we have teams in the US, we have folks in Europe, and we have folks here. And the technology to do that easily now has gotten so good. So I was able to do that here. But that was one reason. And the other reason was, you know, to Gideon's point, you can't build something in Canada for only Canada. It has to be global.
Paul McKinlay: What about from a talent perspective? Like you've built and distributed, but one of the things that Canada is decently well known for obviously, is like AI talent amongst some of the schools here, whatever. Is that proven true and for your ability to build a team and put people together?
George Babu: Definitely. Like I mean we have since everything's distributed now, you can find the best people wherever they are. Right? So there's a lot of folks on our team are from the Valley and they've gone to different places. Right. So now we can hire them wherever they are. And some of them have become nomadic since Covid. Like, literally. My co-founder lives in Berlin sometimes, in Japan sometimes Costa Rica sometimes. And you can't tell, everything is happening 24/7. Right. It's amazing. So it has made it easy to get talent. We, I wanted to be close to the University of Toronto for that reason. Like I wanted to be able to interface with the business school and all the AI-related business strategy stuff they've been talking about. The legal school as well, and all the legal-related AI issues and the Vector Institute. So I've tapped into all those networks. Amazing. So that has been pretty awesome.
Paul McKinlay: Amazing.
Daniel Lee: You brought up AI. I mean, that's one of the things that's obviously different now, but building a start-up, you've got AI doing a lot of the grunt work, building a lot of the code.
George Babu: Yes.
Daniel Lee: If you were to have, well, you are starting a new company. What's different? What made building Rypple different from building Aire Labs today?
George Babu: At Rypple, I was responsible for the operations. So I had to do all the work to get the offices, get the equipment, get this, get that, HR legal, blah, blah, blah. Same thing at Kindred. So fast forward to today. First of all, the infrastructure for building a company from an operations perspective has gotten so easy, right? So we use an AI native services firm that does all our finance, HR, legal stuff, founded by Eric Ries. Right. So I have outsourced all of that. I don't touch any of that anymore. I don't need offices anymore because I have WeWork everywhere around the world. We can have offices everywhere in the world for all of our employees everywhere. Right. And then in the operations now, recently, like I'd say in the last 3 to 6 months, what you've been able to do with Opus on the go-to-market side. Right. Everyone talks about the coding side, but what's happening on the go-to-market side is fascinating, right?
Paul McKinlay: Interesting.
George Babu: When we started the company, a lot of our discovery was recorded a lot of our discovery was recorded because the transcription software companies were first coming out like Fathom and Otter and all these companies. like Fathom and Otter and all these companies. So we have hundreds of conversations now transcribed already. Anytime I want to build a new product, I can just go and ask AI to go look at all these 200 transcripts and find information. Right? That's made it a lot easier. And then, now you have all these presentation tools like Gamma and Claude's presentation tools. I can customise messaging for a particular customer in a particular segment without having like a 30-40-person team. I can do it with one person. Right? So what has become possible with a small team of really good people is phenomenal. So all that means is like, I can run way more experiments on much less capital until I find that thing that then can be venture fuelled, right? And then you can keep doing that again and again and again. And ultimately that's how you build an enduring company. You just keep making these bets.
Paul McKinlay: How does that change how you think about funding a business today and how you think about funding or investing in business in terms of like size of round and how that gets allocated too, right? Like, does it take, you know, as way less capital or is it the same amount of capital but it just gets you farther because you're spending it on different things, like how do you think about that both from a builder and an investor perspective?
George Babu: At Kindred we had a lot of money early and I think that we did make a few mistakes because we had so much money.
Paul McKinlay: Yeah.
George Babu: So this time around we did not take a lot of capital. So we could learn the hard things the hard way. So I think you have options, right? You can take very little capital, have a lot of control and do a lot of experiments until you feel confident and then you take a lot of capital. If you take a lot of capital, you better go very fast and very far, because now the bar has gotten much higher, right? You know, the series A bar, the seed bars are significantly higher than they used to be. Right. Like 3 to 5 million to seed is like, “Okay, you couldn't do better?”
Paul McKinlay: Right.
George Babu: So if you're going to take a lot of money, you better go fast and very far. I don't think it changes that you still need a lot of capital to build an enduring company. I think it's when you time it. Right. So I liked very much when we were at OMERS, we backed Shopify, right? And I think... I love their early start founding story. They had not a lot of capital in the early days because people didn't believe that you needed another e-commerce start-up. So they were very lean and so they had to build their culture, build their processes, find where the real product-market fit was, and then they could raise a ton of money and go and just dominate the market.
Paul McKinlay: Throw fuel on the fire.
George Babu: Correct. We're seeing this in Anthropic. We're seeing this in OpenAI. We're seeing this with every company. Right. So I think you still need a lot capital it's just when you need it, and how many experiments you can run to map out the terrain before you take that big chunk of capital.
Paul McKinlay: Interesting.
Gideon Hayden: I think the allocation of that capital is something that's changed too. I agree with George. We have not seen the cost of building come down at all. If anything it's gone up. But when we look at our teams But when we look at our teams who are typically very R&D heavy. Many of them are not. Their hiring plans are, I don't want to say less ambitious, but they're less aggressive because they're now able to equip their engineers and their team with these new tools that are available to them such that they can be that much more productive. So we're actually seeing spend move from purely like hiring heads in R&D to here's heads in R&D, plus a certain amount of token spend per engineer. And it's shifting towards whether you use Claude or Codex or whatever they may use. But the overall number has not, really changed if anything. What these tools effectively allow you to do is buy time. It's like you can have 20 engineers running in parallel for however long you want if you want to spend that much money. You know effectively, there's an infinite number of engineers that you could be running in parallel. And so obviously we're compute-constrained. But hypothetically we could be there. So for the companies that are at that point where they're confident, we want to raise a ton of money because we see this opportunity. We want to attack it very aggressively. It feels like they can consume a ton of capital right now, so.
Paul McKinlay: I mean, it becomes way more efficient with like the code gen tools and things like that to actually build product. But then there's the whole go-to-market thing where you still have to do and build a customer base, which is which is probably the more expensive part of company building, frankly, than the product side, right?
George Babu: Correct. Like when companies mature, right? Like you end up seeing like 30% is R&D, 70% is a go-to-market sales and marketing expenses. Right. We've been talking a lot about agentic coding tools. I'm actually more interested in the go-to-market tools because that is going to dramatically change the cost set right. And you don't have to be as precise in the go-to-market. You can take advantage of the probabilistic nature of the agents because go-to-market is probabilistic.
Paul McKinlay: Yeah. Interesting. Maybe just switch gears a little bit. But around like cost of build and also cost to run. So we'll talk with you George, first about like, you know, as you're thinking about building Aire and the cost to your end customers like how does it change, like what you build, how you build it both for you guys to run these internally, but then also like one step forward, like how does what is it going to cost your customer in the end to run that platform? Part of enterprise buying enterprise software is like you have some predictability to like what that line item is. AI can kind of throw a wrench in that to some extent. So how do you think about the cost of build and involves, like, how that translates to the customer?
George Babu: You know, the Moore's law of technology, things will just get cheaper and cheaper and cheaper. So we are assuming that right. I think some time ago, Alex Kolicich of 8VC, he wrote this really good piece about the economics of models. So that's been in the back of our mind, right. There's no budget to expiration in the company. So we can do whatever expiration we want. And when you're building that v1 of the product nobody cares right. Customers don't care and we'll eat the cost if we have to. But absolutely worried about the long-term economics of running each action. Right?
Paul McKinlay: Right.
George Babu: So we have built, and many companies are doing this, we’ve built the architecture to be model agnostic. Right? So when the Anthropics and OpenAIs become, I think they will get commoditized. Right. We can switch over to another model. So or if they start jacking up the bills, we can switch to another model. We've been seeing that happening recently. In the last few months our bills have gotten quite aggressive. So we started we kind of accelerated our plans to start using multi-model workflows. Right, so, cost is a big thing, right? Like right now everyone's exploring and no one minds the bills. But once you operationalize something you don't want to pay that bill. So I think the vendors that optimize So I think the vendors that optimise from that from the beginning are going to be in a better position than those that don't.
Daniel Lee: Yeah.
Paul McKinlay: Gideon, how do you think about it from underwriting a new investment? In some ways like token costs and that kind of thing changes, sort of like how you think about a steady state sort of like how you think about a steady state gross margin in a business or what the economics look like. Is that a factor for you guys, or do you think about like, this is such a big platform shift, like we’re just going to build customers. Who cares about the gross margins for now? How do you guys think about that on a new investment?
Gideon Hayden: Yeah, I think, you know, we invest at the early stage: seed, Series A occasionally at the Series B, right? So seed, Series A we’re less focussed on margin profile. I think we need to believe that long term this can be a sustainable business on a unit economics basis. But we're far more concerned with are they building something that customers cannot live without. Right. And their customers, whatever industry they may be in. I think at the end of the day an end customer is only going to buy something if they think it's delivering ROI, right? Like, they think that they're getting value out of this thing. And implicit in that is there some cost, cost return calculation that's happening in their head. We're at this point in the market where we're seeing a lot of companies buying for the wrong reasons. They're like, customers buying for the wrong reasons. So, there's a board level...
Paul McKinlay: Go buy AI tools.
Gideon Hayden: Yeah, it's like we got to do something with AI. And these teams are feeling the pressure. So they're going out and they're finding some vendor and they're saying, “Hey, now we do AI.” But I think, you know, there's going to be this rationalization period that will come when companies are going to say like, “Okay, this is cool, but is this actually delivering value?” It's pretty hard to actually go and build something valuable yourself with agents or whatever. And it's pretty difficult to achieve that last mile value.
Daniel Lee: Thanks for joining us, guys.
Gideon Hayden: Thanks for having us.
Paul McKinlay: Thanks for being on the podcast.
Gideon Hayden: When are we on next?
Paul McKinlay: Episode two. There we go