AI in Focus: Unlocking Innovation, Growth, and Investment Insights

Hi everyone, welcome to this wealthmanagement.com fast chat. I'm Shannon Rosick and today I am thrilled to be joined by Andrew Choi, Portfolio Manager and Senior Analyst at Parnassus Investments. In this conversation we'll dive into the transformative potential of artificial intelligence and its impact on investment strategies. From evaluating the current AI landscape to exploring its role in creating new revenue streams and driving innovation, we're going to cover it all. Plus we'll discuss why active management matters in this space and get Andrew's insights on the opportunities and challenges AI presents. So let's dive in. Andrew, thank you so much for joining me. Happy to be here, thanks for having me. So heavy-hitting question first, how are you evaluating the current AI landscape? It's a great question and I think, you know, a lot of ink has already been spilled on this but, you know, I think first and foremost, especially given the state of the industry today, it's really about having humility and thinking in a probabilistic sort of range of outcomes framework. So the space is incredibly dynamic, there's constant news flow coming out weekly with latest developments out of the AI research labs and a lot of these things can have very lasting profound implications. So there's a lot of capital flows, there's a lot of innovation and I think as a public stock investor, you not only have to be aware of all the opportunities but you have to think about the risks and this is especially true in the large cap space because you're investing behind, in many cases, the incumbents who have the most to lose. So, you know, how I think about approaching AI today, you kind of have to separate it into the various levels of debate so you don't miss the forest for the trees. So I'd say at the very highest level, you know, let's call this the forest, you have the question of frontier model scaling. So, you know, if you think about open AI, anthropic kind of, you know, metas, llama, a lot of these latest and greatest large language models, you know, there's a question around whether they will continue to scale at the same rate as everyone expects. And basically what this means is, you know, can we get more improvement in these models with the same amount of, you know, expected productivity gains from compute? So if you direct more computers at the problem, do we get better outcomes on the other side? Or will we hit a point of, you know, diminishing returns where we start to see plateauing productivity gains? And this has a lot of ramifications for where companies invest. And it also has broad implications for what the end state of the economy sort of will look like. So for example, if these frontier models continue to scale and achieve outputs and functionality that's so fundamentally valuable and different than what's currently possible today, they'll be able to achieve distribution in such a way that like the legacy information technology stack will be disrupted. And on the other hand, if progress on these models is more measured and there's more time and value in actually engaging this technology into specific consumer and enterprise applications, it's more likely that the models become sort of complements to the existing legacy IT stack. So that debate, which isn't resolved, right, will have constant news flow, you know, basically giving you evidence for one path versus the other. And then as you move down to the level of the trees, right, when you're actually picking individual stocks, right, we've always been in the camp that the old rules of, you know, competitive advantage and Porter's five forces apply. And broadly speaking, the areas that we focus, it's sort of the picks and shovels. So semiconductor companies like Nvidia and Broadcom, you've got in large infrastructure providers, more commonly known as like the hyperscale cloud providers, Amazon, Alphabet, and Microsoft are some examples. And then the third bucket where we spend a lot of our time is on, you know, companies that own specific workflows or have proprietary data. And often these companies are one in the same, but this covers areas like software and information services. So there's been a lot of volatility across all of these sectors, and we continue to see there's a lot of attractive opportunities across all three. But that's really the overarching framework, which is big picture, you know, where are we headed with the frontier models, right? What implications might that have for the existing IT stack? And then when we go down to the level of individual companies, which ones are most likely to be advantaged by, you know, the prevalence and the adoption of generative AI and AI in general. So hopefully that gives you a sense of the framework we use to kind of evaluate the current state of AI. I appreciate those insights. Incredibly interesting, but I'm curious because would you say that there's maybe more potential for AI to create new products or services, streamline expenses, or to open up new revenue streams, or is it potentially all of the above? We're in the camp of it's all of the above. And they're sort of related. I think if you think about, and, you know, we think about software and AI as being kind of a form of software, what it's effectively doing is helping with labor costs, right? Like tasks that used to be very laborious, right? Whether it's data entry or calculations, all software is doing is saving on labor costs and actually helping labor become more productive. And this is just another iteration of that. And I think there will be a lot of labor cost savings. And then the question of whether it will be revenue generating is more of a function of is there a business model? And is that business model, does it lend itself to like actually monetize the functionality, right? So is the functionality different? Are there switching costs? Is there a lot of value being created? So I think it's all of the above, but it starts with helping create value in the real world, whether it's, you know, you can actually do things quicker, you can do things more cheaply, or you can do new things that you weren't able to do in the past, which I think we're starting to see a little bit of that now. And it seems like a case really could be made that over time, AI models themselves might even become commodities and the data offering context for the models could determine which companies ultimately get that competitive edge. So what do you think about this possibility and how does it factor into your current investment decisions? Yeah, that's, it's a great question. And as you know, as I alluded to earlier, it's, you know, it's a, it's a debate that isn't actually fully resolved, right? So, you know, the, the weather AI models will be commodities or not, still is very much contingent upon the next generation and the next iteration and, and who's left at the end of the day, competing for kind of a frontier model that's able to generate kind of the latest and greatest output. As of right now, we have multiple different players. And with varying levels of different kinds of business models, like some, some large companies have their own internal models that they're giving away for free, others are charging for it. And so that, that debate is something that we're, we're obviously tracking. But what we know now is that the, the age of using a lot of synthetic data or manmade data to continue to get better and better, better models, we've clearly seen some plateauing in that space. And what we're seeing on the other side of it is companies that have, you know, incumbency in, in workflows and a lot of proprietary data are actually seeing better results with some of the, the models that they're, you know, deploying into certain workloads. Salesforce is a great example. You know, they, they released a product that's now generally available called AgentForce. And, you know, the, the types of feedback that we're initially seeing from that type of product is very positive. And I think part of the reason why it's so positive is it's baked into intuitively into workflows that a lot of users are used to. And at the same time, there's a lot of proprietary data that is really difficult to replicate, right? Like every customer interaction you've ever had for the past 20 years and all of your product information that's bespoke, like all of those things help feed the model. And if you, if you believe the world is going more towards that route where that type of workflow integration and proprietary data is the differentiator, you will see, you know, some level of commoditization of AI models because it's less about the actual foundational model. And it's more about how it's applied in practice. So again, the debate isn't fully resolved, but what we're seeing signs that the model in the models are actually relatively well-served for the, for the initial use cases today, which would imply that over time it might not be a differentiating factor. But again, that could all, that could all change. And speaking of debates, we have to talk active and passive management. Why is active management an important consideration when investing in AI? I think if you step back, like anytime there's a large technology shift with sort of manic market reactions where terminal values are like debated on like a weekly basis and company valuations are marked up and down at a pretty aggressive rate based on like a quarterly earnings report. It's a really attractive and interesting time for active managers because the volatility obviously can create opportunities and sort of the manic short-termism on some of these like incremental data points, right. Where, you know, you see one quarter and you immediately you know, the entire narrative around a company can shift quarter to quarter on whether it's an AI beneficiary or not. I think active managers who are, you know, able to take a longer term duration and perspective on some of these things, think from first principles and, you know, do math on valuation and are able to weather some of that volatility or lean into it when it benefits them. I think it's an interesting time. And I think AI as a secular shift, generative AI as a secular shift has kind of created those opportunities. You know, we've seen it before with like web, with mobile, with cloud. I think now it's become, you know, you could even look at 2024, right? Like the stock performance of certain software companies, you know, swinging by 50% up and down. And these are large cap, well-covered companies all based on, you know, whether they grew nine or 10%, right? Like these are the types of times where it actually pays, I think, to have duration and to be able to look different than the broader index. No, that makes sense. And as somebody who is following this industry extremely closely, what excites you the most about AI? And on the other side of the coin, what concerns you the most? I mean, it's, and maybe this is, maybe everyone feels this way, whenever there's a technology shift, but it does feel like we are advancing society very, very rapidly at an exponential pace. I mean, I've been covering the industry since 2018, just AI. And back then, you know, NVIDIA's data center business was like not even, you know, it was, we were doing some of the parts where the whole company was worth less than a hundred billion. So a lot's changed, but the advancements that are happening, truly it's very exciting, right? Like I'm excited to see how we can improve the human condition. We can further democratize sort of knowledge and opportunity. But on the flip side, you know, Parnassus really works with our portfolio companies to make sure that all of this is being done in a responsible and thoughtful way. And what that means is just transparency of like, you know, how AI is being developed, what type of like, what are guardrails for where these models can like make decisions and how data rolls up and feeds into training. And, you know, whether there's appropriate governance in place to ensure that like rules are being followed, right. And this benefits shareholders too, because over time, you know, this will be increasingly an area of focus. So, you know, I'm very excited about the potential benefits, but, you know, of course there's a certain, you know, dull anxiety over, you know, the guardrails because, you know, anytime there's a new, anytime there's new innovation in the financial industry or in technology, there's always kind of like, you know, a rush to see what it can do. And then, and then after the fact there's regulation. So we'll likely go through a similar cycle here. Absolutely. And it seems like we're very much scratching the surface with all of it. So obviously more to come, as you said, it changes on a weekly basis and I don't expect that to slow down anytime soon, but Andrew, incredible perspective and insights. Really appreciate the conversation. Thank you again for joining me. Thanks for having me.