Justin Anderson outlines AI maturity stages and cost-control strategies for investors evaluating tech adoption.
Summary
Justin Anderson, Mawer's CTO, discusses how investors can assess a company's AI maturity by looking beyond press releases to actual token usage and cost structures. He categorizes companies into 'dinosaurs,' 'rat racers,' and 'natives,' noting that most are currently in the first two stages. Anderson emphasizes that minimizing tokens per task is critical to unlocking value via Jevons Paradox. He recommends investors interview management and request demos to verify if firms are optimizing both value and cost. The episode also touches on Mawer's internal 'Nova' harness and the broader economic impact of AI on consumer prices.
Episode sections
- 0:00 – 4:27
Introduction and AI Investor Tour Context
Speaker 0 introduces Justin Anderson, Mawer's CTO, who recently completed a speaking tour. They discuss the tour's timing relative to a previous podcast appearance and the market context of spring 2025, including tariffs and the 'build vs. buy' dynamic for LLMs.
- 4:27 – 7:15
AI Token Economics and Context Windows
Anderson explains that AI requests are broken into tokens, with costs directly tied to token count. Simple queries use few tokens, while complex tasks involving live searches or database lookups can use tens of thousands, driving up compute costs.
- 7:15 – 9:48
ROI Framework: Value vs. Cost
The speaker breaks down AI ROI into a numerator (value/productivity gains owned by business teams) and a denominator (cost/token usage owned by technology). He emphasizes that minimizing token spend per task is critical to unlocking more use cases via Jevons Paradox.
- 9:48 – 11:26
Orchestrator Culture and Mawer's Governance
Anderson defines 'orchestrator culture' as empowering non-technical staff to build AI applications directly, bypassing traditional slow development pipelines. He notes that governance is a major friction point and describes Mawer's end-to-end governance framework.
- 11:26 – 16:51
Mawer's 'Nova' Harness and Cost Reduction Strategies
Mawer built a custom harness called 'Nova' to efficiently route data requests and minimize unnecessary token usage. Anderson discusses using open-source models and fine-tuning to bake knowledge into model weights, reducing the need for expensive database calls and context window usage.
- 16:51 – 19:27
Company Maturity Stages: Dinosaurs, Rat Racers, and Natives
Anderson categorizes companies into three stages: 'AI dinosaurs' (resisting AI), 'AI rat racers' (chasing value but ignoring cost), and 'AI natives' (optimizing both value and cost). He states most companies are currently in the first two stages.
- 19:27 – 21:24
Evaluating Management Teams for AI Maturity
To assess if a company is AI native, Anderson advises investors to interview management, request demos, and ask specific questions about their cost structure, such as token usage dashboards and the use of open-source models or custom harnesses, rather than relying on press releases.
- 21:24 – 27:53
Economic Impact and Investor Identity Crisis
Anderson suggests that, historically, technology value accrues to consumers through lower prices, and he bets on a competitive AI market driven by open-source options. He also discusses the 'investor identity crisis,' comparing the disruption of AI to the Luddite movement, where human skills and identities are being reshaped by rapid technological change.
Claims worth verifying
Justin Anderson is Mawer's Chief Technology Officer.
Unidentified speaker (speaker_0) ·0:00
How to verify: Mawer company website or LinkedIn profile of Justin Anderson. For recent events, verify 2026-08-17 through 2026-11-17. Use another period only when the speaker explicitly names it.
Justin Anderson completed a speaking tour called 'The AI Investor' in the month prior to the podcast recording.
Unidentified speaker (speaker_0) ·0:00
How to verify: Mawer press releases or event calendar for late 2026/early 2026. For recent events, verify 2026-08-17 through 2026-11-17. Use another period only when the speaker explicitly names it.
A simple AI summary request might use 10 tokens, while a complex request with live search can use tens of thousands.
Unidentified speaker (speaker_2) ·4:27
How to verify: AI provider documentation (e.g., OpenAI pricing page) or API token counting tools. For recent events, verify 2026-08-17 through 2026-11-17. Use another period only when the speaker explicitly names it.
Mawer built a custom AI harness called 'Nova' to manage data access and token efficiency.
Unidentified speaker (speaker_2) ·13:25
How to verify: Mawer technical blog posts, whitepapers, or investor presentations detailing their AI infrastructure. For recent events, verify 2026-08-17 through 2026-11-17. Use another period only when the speaker explicitly names it.
Most companies are currently in the 'AI dinosaur' or 'AI rat racer' stage, not the 'AI native' stage.
Unidentified speaker (speaker_2) ·18:12
How to verify: Industry surveys on AI adoption (e.g., McKinsey, Gartner) from 2026-2026. For recent events, verify 2026-08-17 through 2026-11-17. Use another period only when the speaker explicitly names it.
Anderson bets that the AI market will become competitive with declining prices due to open-source options.
Unidentified speaker (speaker_2) ·23:02
How to verify: Market data on AI model pricing and open-source model releases from 2026-2026. For recent events, verify 2026-08-17 through 2026-11-17. Use another period only when the speaker explicitly names it.
Do your own research
- Review the company's latest 10-K or 10-Q for 'Artificial Intelligence' or 'Machine Learning' risk factors and capital expenditure disclosures.
- Check investor presentations or earnings call transcripts for specific mentions of 'token usage,' 'compute costs,' or 'AI ROI' metrics.
- Verify if the company publishes any technical blogs or whitepapers detailing their AI infrastructure, model choices (open-source vs. proprietary), or cost-control strategies.
- Interview current or former employees to ask about the 'development pipeline' speed and whether non-technical staff are empowered to build AI tools.
- Look for evidence of 'fine-tuning' or 'model weights' optimization in technical reports, which indicates advanced cost management beyond simple API calls.
- Compare the company's AI-related revenue growth against its operating expenses to see if cost efficiencies are keeping pace with adoption.
- Search for news on 'open-source' AI partnerships or implementations, which may indicate a strategy to reduce dependency on expensive proprietary models.
- Assess management's ability to provide specific 'token usage dashboards' or cost breakdowns during investor days, as a lack of detail may signal superficial adoption.
Full transcriptexpand
- 0:00
- Unidentified speaker (speaker_0): Hi everyone. Back on the podcast today is Justin Anderson, Moore's Chief Technology Officer. Last month, Justin was on the road with a speaking tour called The AI Investor. And today he walks us through it. We start under the hood with what actually happens behind the scenes when you're using AI tools, and from there we talk about the economics, some things we've learned putting this to
- 0:20
- Unidentified speaker (speaker_0): work at Moore, and what it all means for investors themselves. With that, here's my conversation with Justin.
- 0:26
- Unidentified speaker (speaker_1): This podcast is for informational purposes only. Information relating to investment approaches or individual investments should not be construed as advice or endorsement. Any views expressed in this podcast are based upon the information available at the time and are subject to change.
- 0:43
- Unidentified speaker (speaker_0): Justin Anderson, welcome back to the podcast.
- 0:46
- Unidentified speaker (speaker_2): Andrew,
- 0:47
- Unidentified speaker (UNKNOWN): great to be
- 0:47
- Unidentified speaker (speaker_2): back.
- 0:48
- Unidentified speaker (speaker_0): It's great to have you. As I said in my introduction. to this episode, you've been out on the road doing a speaking tour called The AI Investor. I joined you on a bit of that tour, and you actually begin your talk by going back to the spring of 2025, which is actually right around the time that you were last on the podcast. So I wanted to start there.
- 1:08
- Unidentified speaker (speaker_0): Why did you choose to start back in the spring of 2025?
- 1:11
- Unidentified speaker (speaker_2): Well, your initial comment there about the Cross Canada Tour. I don't know if you've been on a WestJet flight recently. I'm sure you have. And they have the little safety briefing where they go through and they kind of take you through the cross country, and there's like the Toronto scene, and then there's the Alberta scene, and then there's the Vancouver scene. I'm on the plane and I'm like, I'm doing that same tour that they're doing in the safety briefing. So that was fun.
- 1:31
- Unidentified speaker (speaker_2): It was, it was a great way to see the country. I really enjoyed it. I know I got back to your homeland in in the Halifax area, the Nova Scotia, actually, my first time in Halifax and it was absolutely stunning. So really had a good time there. But to your question, so a year and a half ago, people at the time were talking a lot about tariffs and Liberation Day had just come out and there was a lot of concerns about
- 1:51
- Unidentified speaker (speaker_2): what what that meant for markets and so forth. And in hindsight, looking back, we've always got that benefit. But turned out that tariffs weren't really the thing that moved the needle in the market. It was a number of other things, which we also discussed on the podcast. And one of them that really sticks with me is that build versus buy tension that we talked about. I think we referred to it as build-in versus build
- 2:11
- Unidentified speaker (speaker_2): out. And the concept there was essentially what is happening with LLMs and how that's changing that dynamic where the build option is becoming a lot more attractive to companies internally as the cost to build comes down dramatically. So they can still build that couple. custom software, but they can now do it at a competitive and in some cases even cost efficient way
- 2:31
- Unidentified speaker (speaker_2): relative to some of the legacy vendors that they've been using. So that theme played out over the next year and a half. We heard about the SAS pocalypse. I think it hit home. People are going to be pretty familiar with that thesis playing out. It's led to this thought when we were doing the Cross Canada tour of this sort of thesis approach to investing is going to become increasingly important of looking for these
- 2:51
- Unidentified speaker (speaker_2): kinds of inflection points, buy versus build and that dynamic changing, especially as the LLMs and the AIs get really smart on historical data. Where they're not as smart is how the landscape is changing looking forward. So that was one of the main points of the presentation: just emphasizing where the humans have an edge relative to the AI in and that ability to look for
- 3:11
- Unidentified speaker (speaker_2): these sort of inflection points out there and from first principles type things. thinking of identify an investment opportunity and try to invest around that. So that was one thing we emphasized. And then the other piece in the podcast, I really listened to it recently. We talked in there about how at more we are going to act as if AI is a revolution for our business. And the
- 3:31
- Unidentified speaker (speaker_2): reason that we gave on that presentation or the podcast was that hey, there's an asymmetric payoff here. If we put some resources into this and it turns out that AI is a revolution, we're going to be in a really good position. And so we did that over the last year and a half, and we've really put a lot of emphasis on setting the foundations for how we can scale the use of AI at the firm more
- 3:50
- Unidentified speaker (speaker_2): economically. And we're starting to see the payoff of that. I was really excited to listen to that podcast because I was like, those are a couple themes that tie directly into these presentations and then some of the investments that we've been making.
- 4:00
- Unidentified speaker (UNKNOWN): And
- 4:00
- Unidentified speaker (speaker_0): I want to get into some of that detail, but most people listening to this have used ChatGPT or something like that. Very few, and I would certainly put myself in this camp until I heard you talk about it, but very few have thought about what's actually happening when you type something in and you hit enter. So can you walk us through what's happening behind
- 4:20
- Unidentified speaker (speaker_0): the screen when we do that and ultimately why that matters, because I think that will be really useful context for the rest of the conversation.
- 4:27
- Unidentified speaker (speaker_2): Yeah, it's a great question, Andrew. And I'll take a couple minutes to answer it because I think it is really important to understand a little bit of the details on how these models function in order to start to think about how it's going to roll out into the industry over time and where we're focusing. So it'll set the kind of foundation for us. So listeners, if you can come along with us for the
- 4:47
- Unidentified speaker (speaker_2): journey. So when you ask a question, so we just talked about buy versus build, for example. So the listener might type into ChatGPT, summarize what buy versus build means. So what happens when they hit enter on that request? So the first thing that happens is the LLM model is going to break up that sentence into tokens. And essentially the way it does that is think about every word
- 5:07
- Unidentified speaker (speaker_2): is roughly one to two tokens each. Those tokens, the job of the model is to then predict what the next token should be. In this case, summarize what buy versus build means. It's going to predict the next token, which is essentially the next word. And once it predicts that next word, it's going to take all of those back and repeat the process. And it's going to keep doing that and eventually have a sentence
- 5:27
- Unidentified speaker (speaker_2): and eventually have a paragraph or a whole page. So that's essentially how they're working their token prediction machines. And when you're asked that question, in that case, it's a relatively small number of tokens, right? You're using like maybe 10 tokens on summarize what buy versus build means. But if you're going into more complicated questions, like if you ask, hey, give me an update on the latest
- 5:47
- Unidentified speaker (speaker_2): news in AI. Well, now you're asking a question where the model hasn't been trained on that news. Maybe the model was trained three months ago. And so in order to answer that more complicated question, you have to then do something like a Google search. You've You got to retrieve all the results from that the Google search gives you and then pass. that into that context window. So now you're going from 10 tokens, maybe you're
- 6:07
- Unidentified speaker (speaker_2): going to 100 or 1,000 tokens. And eventually you get into much more complicated operations. Maybe you're tapping databases. You can get into the thousands and the tens of thousands of tokens. And the thing with these models is the cost and essentially the compute that's required to process a request is directly related to how many tokens are involved in the request. So if you're using
- 6:27
- Unidentified speaker (speaker_2): 10,000 tokens versus 10 tokens, you're looking at orders of magnitude more expensive sort of task that you're asking it. So the game in this business is really to understand the token use and try to figure out ways to rein that in, bring it down so that you're minimizing your token use to get the maximum output you can. So that's the reason
- 6:46
- Unidentified speaker (speaker_2): we set the stage here is help understand the context window is really an important part of defining how AI is going to roll out through the industry. And the game that companies are going to try to crack is how do you drive down the token spend?
- 7:01
- Unidentified speaker (speaker_0): Good foundation there. If I'm an executive at a business, Justin, and I'm trying to think through
- 7:06
- Unidentified speaker (UNKNOWN): the
- 7:07
- Unidentified speaker (speaker_0): economics as you just laid them out, can you help me understand that side of things? Where should I be looking and is there a framework that I can apply?
- 7:15
- Unidentified speaker (speaker_2): Yeah. So today you hear in the news all the time this comment around what's the ROI on these AI investments? And there's thesis out there that says there's limited to no ROI, and so that's going to cause a capex bubble. And there may be some potential truth to that or something to that story. But the bottom line is ROI is
- 7:35
- Unidentified speaker (speaker_2): really in focus for the industry today because it's going to define the risk. of a bubble. It's going to define the longevity of these technologies as they roll through the enterprise. You have to get value out of them. So then we think about okay, well, what is the ROI calculation? If you start to try to break that down into its component parts, ROI is a ratio. On the numerator of the ratio, it's the
- 7:55
- Unidentified speaker (speaker_2): value that you're getting from using the technology. So you can think of it of anytime you ask it to write a report or you ask it to explain something for you or to do some kind of job, you're getting some kind of value for it for your business. Some of those values become measurable things where you can see that you're doing more with less, or kind of a productivity improvement. You can look at, hey, maybe we're expanding our marketing reach or
- 8:15
- Unidentified speaker (speaker_2): our marketing engagement. So you can kind of measure some of these things. In many cases, they're very hard to measure. What's the value of being more prepared in a management meeting when you're trying to do an investment research? type of call? You can't really quantify that directly. It's going to express itself through better returns over time, hopefully. But it often becomes a qualitative
- 8:35
- Unidentified speaker (speaker_2): piece. But that's the numerator. It's really what's the value proposition of using the tokens? Then there's the denominator, which is the cost of using the tokens, right? The way we sort of break that down is to say, okay, how many tasks do you ask the AI to do? Multiply that by the number of tokens per task that you end up having to use to serve it. So that goes back to our
- 8:55
- Unidentified speaker (speaker_2): original comment about the context window being so important because if you're using lots of tokens, that's directly linearly related to the cost that you're paying. And then finally, there's the dollars per token, right? What are you spending per token consumption? And that's really a function of the model for the most part: of, hey, are you using the right model for the job? We talk about open source models as another.
- 9:15
- Unidentified speaker (speaker_2): lever that you can pull in order to drive down that dollar per token spend. But that's really the equation that we're talking about and something that we spent a lot of time during the tour emphasizing.
- 9:25
- Unidentified speaker (speaker_0): So don't maximize tokens, maximize the effective tasks associated with the use of those tokens.
- 9:31
- Unidentified speaker (speaker_2): Exactly.
- 9:32
- Unidentified speaker (speaker_0): One of the terms that you used throughout the talk, whether it was in the prepared remarks or in the QA and the lively discussion that we had in a lot of cases at these, you keep coming back to this term orchestrator culture. What do you mean by that? And why do you put it on par or even ahead of the technology?
- 9:48
- Unidentified speaker (speaker_2): So coming back to our ROI calculation, value versus cost, the orchestrator is, well, that's you and me. You've been building some applications yourself and some folks on your team. But the idea of the orchestrator is that we've gone with this technology into a world where everybody at the company now can actually be an
- 10:08
- Unidentified speaker (speaker_2): orchestrator. They can work directly with the technology to build little applications or have chat experiences that are very specific to their domain.
- 10:17
- Unidentified speaker (UNKNOWN): One
- 10:17
- Unidentified speaker (speaker_2): of the challenges is as organizations, we're not really designed for this world in a sense. We've been designed to be functional specialists where you want to build an app. application, you go to a business analyst and they help you write out the requirement for that application. We then put it through a governance process to see should we be spending firm resources on this. And then we put
- 10:37
- Unidentified speaker (speaker_2): it to a developer who tries to implement it and then it goes up and down the chain. And that becomes a pretty expensive process to sort of manage. And the idea of the orchestrators, you can collapse a lot of those steps into a single person who can now have the vision for the product. they want to build, and then they can build it, they can deploy it, they can share it and really leverage their knowledge base. But to get there
- 10:57
- Unidentified speaker (speaker_2): isn't easy, because as I said, organizations are organized and functional specialists. And so how do you get there? Well, it's a number of different things of incentivizing folks and driving the culture in a way that encourages people to build an orchestration culture, not a trivial task, something to spend some focus on.
- 11:15
- Unidentified speaker (UNKNOWN): I
- 11:15
- Unidentified speaker (speaker_0): mean, I'll put you on the spot there. Justin, you're the chief technology officer here at Moore. How are we applying this? Whether it's the technology or the orchestrator culture at Moore, and what are some of the things that you've learned along the way?
- 11:26
- Unidentified speaker (UNKNOWN): You
- 11:26
- Unidentified speaker (speaker_2): may have heard of Jevin's paradox. The idea of Jevin's paradox is that as the cost of some sort of technology comes down and the price comes down, essentially that deliver the technology, you might think that the technology companies company makes less money from that over time. In many cases, it's not the case. It actually makes the market much bigger and drives up the ultimate
- 11:46
- Unidentified speaker (speaker_2): value that they're able to capture. The philosophy that I see with these technologies is very similar, where there's an opportunity for us in technology to drive down the cost of using the technology. So if we can get our dollar per token and our tokens per task. down to cents on the dollar, that'll
- 12:06
- Unidentified speaker (speaker_2): open up all sorts of different use cases that our business users can tackle. Because the business users are really more in charge of the numerator part of that ROI calculation, of finding the opportunities because they have the domain knowledge. They've been serving clients, they've been doing investment research, they've been marketing, they've been in finance, and they have that specialist knowledge of their field and they're
- 12:26
- Unidentified speaker (speaker_2): best suited. to figure out how to use AI to help drive their field forward. And our job in technology is that denominator is to make sure that the cost for them is as low as possible without compromising quality so that they can maximize the number of use cases per Jevins paradox that makes sense. So that's something we've been very laser focused on. In terms
- 12:46
- Unidentified speaker (speaker_2): of how do you do both of those, we could do a whole podcast on the value side of it. It's things. like holding hackathons and having a weekly AI moment where you're showcasing things that people are working on. It's really important that the leadership drives it, the leadership of the company takes command and figures out how to drive the value use cases for
- 13:06
- Unidentified speaker (speaker_2): their respective departments. And then on the denominator, it's a number of things, but some of the ways that you can tackle it are open source models. Open source models gives you a lever. of driving down your dollars per token in particular. Something we've done at more is we've designed and built our own harness. We call it Nova. And the job of this harness is to serve that
- 13:25
- Unidentified speaker (speaker_2): context window, those tokens as efficiently as possible, so that when our users are doing tasks, they're using exactly the tokens that are necessary and no more in order to capture that Jevons paradox cycle.
- 13:37
- Unidentified speaker (UNKNOWN): So
- 13:37
- Unidentified speaker (speaker_2): those are some of the ways that we we're trying to tackle these ROI calculation at more.
- 13:42
- Unidentified speaker (speaker_0): The harness, this would be akin to it's almost like an air traffic controller or a stoplight directional light on a train track or something. It takes what you're asking and just routes it to the most efficient use on the other end.
- 13:55
- Unidentified speaker (speaker_2): That's it. So our harness now has access to essentially all of the data at our company.
- 14:00
- Unidentified speaker (UNKNOWN): And
- 14:00
- Unidentified speaker (speaker_2): it's all governed under who should have access to. what data. And so when someone's asking a question, it's determining does this user have access to it and which data do I need in order to serve the request? And to do that at lightning speed and only pull the data that's necessary to serve the request. So it's handling all those data operations in the background and trying to do it as efficiently as
- 14:20
- Unidentified speaker (speaker_2): possible. The other thing you kind of get into as we talk about ways that you can drive down your token spend might be interesting for listeners. is you eventually get into this place where you might want to start actually fine tuning and do post training on your own models. That sounds kind of complicated, but the reason you want to go down that path from a cost perspective, it ties back to that token
- 14:40
- Unidentified speaker (speaker_2): use that we opened with. As you have more token consumption, we realize that, hey, what if I moved some of the knowledge from my databases that I have to pull and put into the context window? What if I took that out of the database and instead baked it into my model weights? In which case I wouldn't have to do those database calls and I wouldn't have to pass it into the context window. I could
- 15:00
- Unidentified speaker (speaker_2): save an order of magnitude of tokens. So that's something that I think more advanced companies, we call them the AI natives, are going to get into more and more is they're going to start figuring out ways to get that cost advantage relative to their peers of even fine-tuning models to drive down. their token use.
- 15:16
- Unidentified speaker (speaker_0): Interesting. And we're going to come back to that AI native term shortly because I want to ask you more about that. But just before we move on, if you're not an AI native firm currently, what's more important at this stage if you aspire to be? Is it to just get in there and use the technology and not really worry about cost? Is it to apply
- 15:35
- Unidentified speaker (speaker_0): a cost conscious framework to it at the same time? Are you Or trying to balance those?
- 15:40
- Unidentified speaker (speaker_2): The number one thing I heard when we went around and did this tour, constantly the feedback was hey, we want to go get more into this stuff, but our governance, our compliance, there's a lot of hesitation to bring the technology in. And so I think if I was putting a word on it, I think governance is something that is a major friction point for people to
- 16:00
- Unidentified speaker (speaker_2): get into this technology and for good reason, right? This technology. has a lot of power. People recognize that. And so they want to appropriately govern it because you're dealing with your business and you don't want to put your business at risk. So there's a sense that, hey, let's take this gradual and let's do this right. We are more, we've spent about a year of time on really refining our own governance process. It's a it's
- 16:20
- Unidentified speaker (speaker_2): quite a boring topic, but we like to be boring, make money. So it's right up our alley. But it's really, hey, how do you set up your data foundation? foundations, your data governance levels, your app governance? And then eventually you get into the node governance. And that's really the controlling how people are accessing your data. And then eventually AI governance itself. And so we've built an entire governance framework end-to-end
- 16:40
- Unidentified speaker (speaker_2): that is really taking into account the unique risk profile that AI brings.
- 16:45
- Unidentified speaker (speaker_0): Justin, in your talk, you sort companies into three groups. Can you walk us through those?
- 16:51
- Unidentified speaker (UNKNOWN): You
- 16:51
- Unidentified speaker (speaker_2): kind of got at this on your previous question, asking kind of about the evolution of how do you bring someone in who's very new to AI or hasn't used it and evolve them. A little demeaning, but it's it's not meant to be. It's just to define these groups. You start with the AI dinosaurs. That's the idea that someone who's maybe resisting it or maybe they have never used
- 17:11
- Unidentified speaker (speaker_2): it.
- 17:12
- Unidentified speaker (UNKNOWN): We
- 17:12
- Unidentified speaker (speaker_2): all start as dinosaurs. You start using the technology. get all excited about it, and you evolve into what I call another somewhat derogatory comment, which is the AI rat racer. The idea there is if you go back to that ROI calculation, the AI rat racer is very much focused on the top line value. They're creative, they're coming up with all sorts of great ways to use the
- 17:32
- Unidentified speaker (speaker_2): technology and different use cases. Maybe they're even engaging in a little bit of AI slop. You know, don't just hand it all over to the AI. You've got to be thinking as and working as partners. And so that's part of learning that orchestrator culture to call back your previous question. But that's the next phase is the AI rat racer spinning on the value that you're trying to create with this thing. And you're not spending a lot of time thinking about
- 17:52
- Unidentified speaker (speaker_2): Jevin's paradox or the cost side of things. And eventually the bill comes in and you look at the bill and you say, Well, I was doing a lot of cool experimentation, but I'm not sure it was worth this much money. And that's where a lot of the rat racers are today, is they're starting to question that. And they're starting to say, okay, experimentation time is over. Now how are we actually adding value here? And that leads you to the
- 18:12
- Unidentified speaker (speaker_2): next phase, which is also thinking about that numerator, value creation. That's the AI native, but also considering the cost side and being obsessive about the cost, realizing that driving cost down to five cents in on the dollar means you open up Jevin's Jevon's paradox, you open up all sorts of new use cases, and you start to actually drive a very positive RY. And so I think
- 18:32
- Unidentified speaker (speaker_2): how companies are going to almost self-sort into those buckets, the vast majority today are probably in the dinosaur or the Rat Racer bucket. So if you're one of those, don't feel too bad. You're with most of the crowd. That's where the world is today. But to get to that AI native, I think it does mean thinking a lot about some of these issues. We talked about fine-tuning models. harness using open
- 18:52
- Unidentified speaker (speaker_2): source models to drive down the cost structure of the technology.
- 18:55
- Unidentified speaker (speaker_0): If we shift gears a little bit and step back and talk about what we do at Moore or any investment management firm as our core business, a big part of our job is to assess the management teams of our investments or our potential investments. And seemingly every management team out there says they're all in on AI right now. We're
- 19:15
- Unidentified speaker (speaker_0): even talking a little bit about that right now, right? If you put your equity analyst hat back on for a few minutes, how would you assess or actually tell a dinosaur from a native from a rat racer at this point?
- 19:27
- Unidentified speaker (speaker_2): I always love the opportunity to put that hat back on. So thank you for the offer. Let me take a shot. I actually bring this one up because I think it's a really good example of the role of the human in investing. And it there's many roles, but one of them is to figure out these types of questions. We talked about these different buckets of going from dyno to native. How do
- 19:47
- Unidentified speaker (speaker_2): you get at that? So I think it really is a people question. And you're not going to be able to ask an LLM, is this company AI native? Like it's not going to tell you that. It's going to give you whatever they've curated in their press the releases, and you're not going to get that information. The only way you're going to get it is to hit the ground, in my opinion. and go and talk to these companies and interview them and get into the weeds of
- 20:07
- Unidentified speaker (speaker_2): how are they maximizing the value that they're getting out of these technologies? Like what are the different use cases? Have them demo it to you. Show me like what are the different ways that you're using the technology. And that's only going to get you to Rat Racer. It's like, oh, you're this guy's using it a lot. That's great. I'm On learning. Then you want to start getting into hey, do you have an open source model? How do you use your open source models? Do you have your own
- 20:27
- Unidentified speaker (speaker_2): harness? And you may not want to ask it in a leading way like I'm doing right now, you know, because they might be like, of course we got a harness. Figure out in a more neutral way to see if they're going after the cost side of the ROI calculation. So those would be some of the techniques I'd be really going at is trying to figure out, do they take the cost side seriously? Are they able to articulate the role? Like, can they show you a dashboard that shows all
- 20:47
- Unidentified speaker (speaker_2): the token use that they have throughout the company and their dollar per token spend? and how they've been systematically driving it down? Can they articulate all of that and get down into the weeds there? I think that's the mechanism that we would use to try to sort that out. And I don't see how you could ask that from an AI because I don't think it's going to come across unless you talk to the people.
- 21:04
- Unidentified speaker (speaker_0): Excellent point. And we're going to have our colleague Jeff Moe on the podcast in the coming weeks. And I'll put that exact same question to him to see what he says because he sits in a portfolio manager position. currently. I did want to stay on the investment side of things just briefly, but if we think about the economics, generally speaking, when a technology like this takes hold and you have an
- 21:24
- Unidentified speaker (speaker_0): industry that maybe the whole industry gets more productive. In the past, a lot of that usually ends up going to customers through lower prices, for example, coming back to some of the things that you were talking about earlier. Is that what you expect to happen here? as well?
- 21:39
- Unidentified speaker (speaker_2): It's another one where I firmly got that equity analyst hat on, so I appreciate it. The history of innovation and technology is such that it in most cases, if you looked at the base rates, the value accrues ultimately to the consumer. The technology goes up, people start adopting it, and eventually you get like the airlines Lines today. They don't make a lot of profit, but the consumers
- 21:59
- Unidentified speaker (speaker_2): have never been better off. They get tons of access and try. travel and whatnot. There are cases where that's not that doesn't hold. You look at the internet phase and the internet companies, and those businesses were able to carry a lot more of the value. Of course, consumers also got a lot of the value, but I think the producer in that case was able to retain more of the economics for various reasons around
- 22:19
- Unidentified speaker (speaker_2): concentration of services and that sort of thing.
- 22:22
- Unidentified speaker (UNKNOWN): In
- 22:22
- Unidentified speaker (speaker_2): the case of AI, it's a bit of a jump ball. There are a small number of companies. that are really leading this from the value stack. If you look at the labs themselves, the frontiers, as well as the value chain of producers for the chips and so forth, you could imagine a world where they are able to retain more of the economics.
- 22:42
- Unidentified speaker (speaker_2): And as they build more expertise and concentrate that expertise, competition is unable to bring its power to bear. It can also imagine a world where no, like it it becomes very competitive, and increasingly you see a lot of these open source capabilities coming out that are keeping the frontiers very competitive. I'm hopeful that that's the way the world plays out, that you have leading frontier models and
- 23:02
- Unidentified speaker (speaker_2): then you have lots of healthy open source options, because I think that'll keep it very competitive and very good for the consumer of these products and for prices and so forth. If you ask asked me to wager, I would bet on the latter. I think the competitive environment and therefore the prices declining and so forth is likely the way that this will play out. But there's lots of lobbying going on and there's lots of
- 23:22
- Unidentified speaker (speaker_2): efforts to shut down the open source options and so forth. So it remains to be determined how this market will play out.
- 23:28
- Unidentified speaker (UNKNOWN): If
- 23:28
- Unidentified speaker (speaker_0): you had your equity analyst hat on there for a bit and you were kind of looking outward, I want to look inward now as an investor. And there was a slide tucked into your appendix of your presentation. called the investor identity crisis. Can you talk about what you mean by that?
- 23:43
- Unidentified speaker (speaker_2): Yeah, I mean, this one, we faced this in the technology side of the house about a year ago or roughly when we were doing the podcast with you. Things were changing so quickly. In the case of software development, the feedback loops in that function are really tight. And so the technology, the AI, has evolved so quickly to the point that it is now writing
- 24:03
- Unidentified speaker (speaker_2): essentially all of the code for your AI native type businesses. And that happened almost within a year, right? So it's such a rapid change that we saw in that field. And it's been very difficult, right? In many respects, it's exciting to see all the productivity potential. But at the same time, you've got folks who maybe they spent their career developing software and writing code. And
- 24:23
- Unidentified speaker (speaker_2): you're telling them: hey, like the fact that you know JavaScript or you're one of the best users of JavaScript. we don't necessarily value that as much. I remember priding myself as being one of the strongest Excel users out there. I almost went to a New York competition on it. And nowadays I think it's probably not that valuable because the AI is going to know its way around that spreadsheet better than I could ever dream of. That's difficult
- 24:42
- Unidentified speaker (speaker_2): for humans, right? Because we built these skills and it changes. The privilege of the present is to judge the past. We always look at the past and say, oh, those guys, they weren't so smart. And one group in particular that I often look down at was the Luddites. It's like, okay, here are the Luddites, they're anti technology, they're breaking stuff and they're going into the factories. They're resisting the industrial
- 25:02
- Unidentified speaker (speaker_2): revolution. They're resisting progress. What's the problem? It's terrible. But you start to realize just how this goes to the heart of what it is to be a human, right? Is like your identity. And these people, they had their identities and their identities. were being shaken. And they were being told that the future is going to be different. And so I think we're not in a dissimilar situation today where, like
- 25:22
- Unidentified speaker (speaker_2): the Luddites, we see the progress and it's exciting what we can do and how productive we can build. And I don't think anyone wants to go back to 90% of people working in the farms. So that probably isn't a good idea. But at the same time, it's hard. The change is difficult and it shakes you to your core. And you're seeing this in mathematics. And you're starting to see it more and more in investing, where it's really
- 25:42
- Unidentified speaker (speaker_2): shaking up, particularly that analyst role, of how it is that you aggregate information and organize that and deploy it to a human decision maker. So it's definitely something that we're thinking a lot about is like what is our investor identity and how do we evolve that with this technology? But being very compassionate and careful to be very kind in our opinions and realize that
- 26:02
- Unidentified speaker (speaker_2): not everyone's going to have the same views on these things and we got to move towards this together kind of carefully and not too judgmentally.
- 26:08
- Unidentified speaker (speaker_0): That's a really excellent point. And I was laughing a little bit at your Excel competition comment just because I'm not probably as good as you in Excel, but I'm certainly among the best in my peer group in terms of outside of more, just in my friend group. And I was always asked to kind of help out on Excel and now can't show off those skills as much. Justin. Maybe for
- 26:28
- Unidentified speaker (speaker_0): a period of time when I know how to use AI within Excel until they catch on. Maybe a last question for you before we let you go. Uh what have you changed your mind about since you were last on the podcast a year and a half ago?
- 26:40
- Unidentified speaker (speaker_2): I think the technology is moving so quickly. Maybe if you asked me this question naively, I might have said something like, I would think that humans would keep pace probably a little more. Something that has surprised me a little bit is just how quickly the tech has outstripped the ability for the humans to stay up to pace. And I think that's probably just my own inexperience or naivete in
- 27:00
- Unidentified speaker (speaker_2): realizing that it does take more time for us to adjust than some of the base rates that I was kind of assuming.
- 27:06
- Unidentified speaker (UNKNOWN): All
- 27:06
- Unidentified speaker (speaker_0): right. Well, Justin, over the last few weeks you've jetted across the country and back talking about this. You've now sat down here and walked us through it all the way from the technical to the financial to the cultural, and I think what it all means, both for businesses and investors alike. So I appreciate you taking the time, and I know our listeners do too. So thank you, Justin.
- 27:24
- Unidentified speaker (speaker_2): Thanks, Andrew. Appreciate the curiosity and the chance to be with you again.
- 27:29
- Unidentified speaker (UNKNOWN): Hey
- 27:29
- Unidentified speaker (speaker_3): everyone, Andrew here again. To subscribe to the Art of Boring podcast, go to more.com. That's M-A-W-E-R.com ma-w-er-r.com forward slash podcast or wherever you
- 27:38
- Unidentified speaker (UNKNOWN): can.
- 27:39
- Unidentified speaker (speaker_3): download your podcasts. If you enjoyed this episode, leave a review on iTunes, which will help more people discover the be boring make money philosophy. Thanks for listening.
More from Art of Boring
AI Build-Out: What We Know, What We Don't, and What's in the Stock | EP 236
Episode discussion: Petkovic discusses AI supply constraints, enterprise adoption, and political risks in data center build-out
Speakers’ claims: Irena Petkovic joins the Art of Boring podcast to analyze the current state of the AI build-out following recent market volatility. She argues that hyperscaler growth is now bottlenecked by data center construction speed rather than demand, citing revenue acceleration from Amazon AWS and Google Cloud. Petkovic highlights significant enterprise adoption data from Anthropic's S-1, noting that Fortune 500 spending on AI remains a tiny fraction of total expenses. The discussion also covers the increasing reliance on external financing for infrastructure, potential political headwinds in key states, and the structural shift in internet traffic caused by AI agents bypassing traditional ad models.
Global Small Cap: A New Lead Manager on the Missing Premium, Local Businesses, and Talent Density | EP 235
Episode discussion: Ching Chang on Small Cap Premium Disappearance and Local Business Opportunities
Speakers’ claims: Ching Chang discusses his evolution from value investing to focusing on quality compounding companies, citing his experience at Ontario Teachers' Pension Plan. He attributes the disappearance of the small cap premium to companies staying private longer and large-cap tech firms scaling without capital constraints. Chang views small caps as diversifiers with local business focus rather than expecting a return of the historical premium. He emphasizes the need for regional overlays to manage macro risks and evaluates management teams based on customer focus and talent density.