Hello everyone. Thank you for being here. This is a conversation that's going to be dominated by insights into how AI operates in the real world, both from an investor perspective but also operationally in a fast-growing business. I'll start off with a question for both of you: where are you seeing genuine changes in business outcomes for your business thanks to AI? And as that becomes more common and accessible, what actually stays proprietary? Let's start with you, Christine.
Well, it's great to be here. If you're asking about how AI impacts outcomes for our firm, one thing that's interesting to point out is that there is so much talk, especially in our world for early-stage venture capital, about VCs investing in AI and disruption when the VCs themselves may or may not embrace it. For 500 Global, we are a venture capital firm with over two billion in AUM across many funds. This is our 16th year. I'm based in the Bay Area but our team is in 20 countries around the world. So we definitely have that perspective, seeing both our teams and many other teams from early-stage initial checks all the way to the companies that have grown and how they themselves have embraced AI. For us as a firm, it is something that we have really instituted in the last year or so: we want the team and expect the team to be AI fluent. It doesn't matter if you're an investment team member or on the finance team or other platforms within our group, you have to really be AI fluent. A lot of the impact comes in certainly low-hanging fruit, things that are administrative or back office. But as you get into data intelligence, so far to the extreme of does it actually help select investment opportunities, those are all areas where we are certainly leveraging it. Within our portfolio, it definitely ranges. But by and large, as a rule of thumb, companies if they are really not embracing it in terms of being AI-first or AI-native, then it's not to their advantage.
Less of an attraction. Same question for you, Rishi.
So operationally, first again, as a digital commercial bank, we've been running 10 years, lent cumulatively $21 billion over that period with less than five basis points of principal losses. We care a lot about lending well into the low-middle market, which is our customer base, and we've been using ML in the business to actually build our credit approach really since the get-go. And clearly with what's been possible with Gen AI, we've been able to evolve that pretty materially. So really at the core, I'd say AI infrastructure is affecting the core infrastructure of the bank. There's obviously the usage of AI to drive efficiencies across the business. An example of that would be some of the teams absorbed sort of over a 30% increase in volumes last year with flat headcount and have committed to doing pretty much the same again this year. So you're getting real efficiency gains. And then in terms of the customer experience, you're being able to deliver things quicker and actually deal with more edge cases than you otherwise were, and therefore be able to serve a wider set of customers in a more timely manner. So I'd sort of say three buckets. And in the infrastructure bucket, I'd also put the very conscious decision to move to actually probably buying a lot of SaaS componentry but buying much less SaaS applications and actually being able to develop those SaaS applications at incredible speed internally, which again was core to what we were doing. But we did have various applications which today you'd sort of say you probably don't need, you aren't as reliant on as you were previously in terms of the buy decision versus the build decision there.
Yeah. No, that's interesting because I was thinking about coming back to you, Christine, when I was thinking about this in terms of how you talked there about looking at companies or startups that are clearly AI fluent. How does that change how you think about founders? Because that's who you're really making a bet on every time you invest in any startup. What do you expect from the founder in this time that's changed the way these companies work?
I think some of the foundations of building a business, whether it's in this era or not, there are some fundamental things. But I think one that's especially important these days is the ability to navigate change and move quickly. Because technology is moving at a really fast clip, the velocity of which AI has evolved even in the last couple months. When I think about some of the early generative AI images, I was just looking at my phone and even a year ago, it pales in comparison to what is capable now. And I don't think people have anticipated just how quickly that sort of change can happen. So I think with early-stage founders especially, they do have the advantage they're building today, they're building AI-native, they don't have a legacy technology stack. So in some ways they can adapt to that. But with some of our more mature portfolio companies that haven't exited yet, if they have done a good job, they have really been able to reinvent not just internally their tech stack or their team organization, but also the offerings that they have. They potentially have an AI offering. Intercom, which is a company we invested in, one of the first investors in the company back in 2011, they have been very public, Owen the CEO, about how they reinvented the firm. Made a lot of really hard changes, but now it's for the better. Finn is kind of their agentic product that has been the fastest-growing revenue. And so I think Intercom's a great example. Canva is also one of our early investments, a great example of that in terms of really being able to leverage the technology that could disrupt them in theory.
Right. And this whole point about uncertainty, I guess it doesn't just apply to technology in itself. It's just uncertainty. It's always uncertain in macro terms. And so you're really looking for people who can adapt, move quickly.
Yeah. Move with the time.
Exactly. Coming back to you, Rishi, just thinking about AI and the edge that you've used AI for in terms of figuring out credits on potential clients. How has that sort of changed your approach? How's it sharpened? And just to sort of explain to people who might not know, how much you've really made a bet on that for OakNorth.
Sure. So, look, our credit model has fundamentally been the underpinning of how we've lent our $21 billion and lent several billion every single year. So the bet's pretty big in our view, in terms of how much capital we're deploying off the back of our approach. As I mentioned, ML has been built into the model since day one and the use of Gen AI has enabled us just to use our data sets to an even greater degree. Because with ML you still go from statistical regression analysis to machine learning. And then you go to Gen AI and Gen AI is just a totally different level to what you're able to do with more traditional ML-type techniques. So that's been absolutely critical. And the evolution with Gen AI, as I said, in terms of the data sets we're able to actually absorb and drive insights from, and drive those insights back to what does it mean for a single credit, which is really where the big challenge is. So both having the data sets, which we have one of the largest US proprietary data sets for US commercial lending data today, but then applying that through to actually what is the impact for a single underwriting decision. And also, somewhat uniquely at OakNorth, we actually re-underwrite our book every single month. And we drive that by different scenarios as well. So you take the current situation of the Iran conflict and you sort of say what would an extreme situation look like there, and that scenario, how would you use Gen AI to actually inform what are all the interrelationships between the energy shocks, inflation rates, etc. All of that, which obviously you'd parallel with previous periods to create your macroeconomic scenario, which you take down to an industry basis and then from an industry basis down to a company basis. And doing all of that is just non-trivial for it to make any sense and in a way have the confidence to lend on the back of it. So this does really get to the core of what we do.
Where's the human judgment in there? Where does that fit into that?
So for us it's always with a human in the loop. For sure, our average loan size is in the teens millions. So we are lending, it's not like we're lending $10,000 tickets which are purely statistically based. So having a human in the loop is absolutely critical. And all of this is about saying actually how do you reduce the amount of manual and increase the amount of machine. But in increasing the amount of machine, not to get the same result but to get a fundamentally better result because of the whole processing power and insight that you can actually drive from it.
Okay. So Christine, talk a little bit about hyperscaling. It was all the rage and this was the thing that so many investors thought about. Do you still think in that way? Do you think that's still a model that works in this current environment, just hyperscaling for investments?
