Value Driven Data Science

[00:00:00] Dr Genevieve Hayes: Hello, and welcome to Value-Driven Data Science, where data professionals become strategic experts. I'm Dr. Genevieve Hayes, and today I'm joined by Lauren Pearl. Lauren is a business strategist, three-time founder, and CFO advisor who helps startup teams build data-driven businesses that thrive. She is also the resident startup finance expert at NYU Stern's Berkeley Center for Entrepreneurship, where she teaches financial modeling to founders and is the co-host of The Growth Minded CFO podcast.
[00:00:40] She previously appeared as a guest on this show back at episode 34. In this episode, we'll explore why building something valuable with AI doesn't necessarily make you valuable and apply the same strategic thinking companies use to protect their most defensible assets to figure out where real expertise still lives.
[00:01:02] Lauren, welcome to the show.
[00:01:04] Lauren Pearl: Thanks so much for having me again, Genevieve. We had so much fun last time. I'm really excited to have another conversation with you about AI.
[00:01:11] Dr Genevieve Hayes: I am excited about this too. I once met a data scientist who was so good at his job that he automated his way out of it. He systematically built automations to handle his team's work, and once those automations were in place, the team no longer needed as many people, so his role was ultimately made redundant.
[00:01:33] He built those automations because he believed they'd create business value for his employer, and they did. The problem was that the value ended up sitting in the code, not in him. And because his employer owned the code, he was the one considered dispensable, not the tools he'd built. This is an old story occurring well before the current AI wave, and it does have a genuinely happy ending in that this man ultimately built a business around code-based automation.
[00:02:03] But I was reminded of it recently, Lauren, when you shared a similar story from your own experiences with finance pros and generative AI, which you called the AI chicken nugget problem, which is a fantastic title I must say
[00:02:21] Lauren Pearl: Thank you very much.
[00:02:23] Dr Genevieve Hayes: so to begin with, what is the AI chicken nugget problem, and how does it affect those of us working with generative AI?
[00:02:31] Lauren Pearl: So the AI chicken nugget problem comes from a similar story to your friend actually, but it comes from the show The Wire which was a show that was heavily referenced in my MBA strategy course for its many valuable insights. The story they were telling in The Wire is two characters were eating chicken nuggets, and one of the characters says, "Man, these chicken nuggets are delicious.
[00:02:55] Whoever invented these must be a billionaire." And the other character laughs and says, "No way. He's not a billionaire. The guy who invented those nuggets is still probably a line cook working for minimum wage in a basement somewhere. The billionaire is the company the chicken nuggets were made for."
[00:03:13] It's this idea of looking at creating value in a company and being skeptical about who owns that value, because, yeah, the chicken nuggets are great, and if you made the chicken nuggets, you might think that makes you great. But as we can tell from the chicken nugget story and from you with your friend, that's not always the case.
[00:03:34] So this came up recently when I was looking at a flurry of activity among finance professionals building tools using AI. And there's a lot of excitement in the space right now, a lot of clubs teaching you how to use AI in your job, how to vibe code, pushing things forward to become an AI expert, and also a lot of talk around not falling behind and not losing your job to someone who does know AI.
[00:04:02] And in all of this, I saw, oh, this is the chicken nugget problem, that people are being sometimes sold this idea that they need to make nuggets, that they need to create these tools using AI to do their work faster and better and innovate at their companies. But what they're not really thinking about is if I do that for my company, what am I actually getting out of that?
[00:04:30] Am I really the value here, or is the value the tool? And sometimes I would say it's one, and sometimes I would say it's the other, but we should be thoughtful about which it really is gonna be.
[00:04:43] Dr Genevieve Hayes: So the core claim here is that value attaches to the asset, not the person who built the asset. But presumably that's when the skill needed to create it is common or easy to acquire. Does that sound right?
[00:04:56] Lauren Pearl: Yes, that's exactly right, and that's really key. So one of the things that makes AI particularly challenging in this context is the fact that building tools with AI is easy,
[00:05:12] Dr Genevieve Hayes: Yes
[00:05:12] Lauren Pearl: right? In the past, building a new tool that can do something fabulous has been a challenging skill, so for years, engineers have enjoyed higher salaries, lower unemployment because they can build things that nobody else can.
[00:05:28] And now suddenly AI is making building things, programming easier. And so the folk kind of understanding would be to think that, "Oh, wow, AI's easier. I can build things more easily. I should be able to get all that value that engineers could get." Great. But what's really happening is that when you make it easy in general, you're not just making it easy for you, it's easy for everyone else, and now this skill that used to be scarce and valuable deflates.
[00:05:58] Now this skill is no longer a point of differentiation, And so the value then does not become the skill of vibe coding, because vibe coding is easy. Everyone can do it. The value now is whatever the tools you create through vibe coding can do for the org.
[00:06:14] Of course, there's still some distinction. If you are an engineer and you are a good programmer, arguably you still have rare skills, and using AI as a programmer is definitely gonna get you a better result than using AI to build software if you've never programmed a thing in your life.
[00:06:30] But vibe coding itself is not a rare skill, there's a million different courses that'll teach you how to do it, that promise you skills within a day, within an hour, within two hours. Nothing you can learn in two hours is going to be a skill that completely distinguishes you from everybody else at your job.
[00:06:47] So there's that, that it's just not as hard of a skill to learn as other skills might be. And secondly, we have a collapsing of the learning curve because AI companies want us to use their software. So it stands to reason that over time, the skill required to vibe code to work with AI is just gonna go down and down.
[00:07:08] And so even devoting time to learning how to program with something with AI or use AI in your day-to-day life, as AI software companies continue to develop their product, those skills probably also get less necessary because the software gets easier and easier.
[00:07:27] Dr Genevieve Hayes: I agree with you that AI is making things easier, but that also devalue- values them. It's commoditizing skills. But it doesn't necessarily mean that you don't need those skills. So just because everyone can use AI now doesn't mean that you have an excuse not to learn how to use AI.
[00:07:47] Lauren Pearl: I agree with that
[00:07:47] I think also we need to call out exceptions of context as well. So when we're talking about the chicken nuggets problem, we're talking about an employer-employee situation where when you innovate, you don't own the output. But I do think and this is again, like the other thing to pull out of like a business academic context is this is also related very much to like the resource-based view is what they'll call it in an academic setting.
[00:08:14] Who owns the resource? Where does it really live?
[00:08:17] Dr Genevieve Hayes: That was what I was gonna ask you next,
[00:08:18] Lauren Pearl: Yeah. That's where that idea of where is the value where that comes from.
[00:08:22] Dr Genevieve Hayes: So I was gonna ask, does your argument about the AI chicken nugget problem still apply if you're an entrepreneur versus as an employee?
[00:08:29] Lauren Pearl: Yeah, it's totally different, right? Because if AI enables you to create something rare or to put into practice something that folks will really enjoy that you can get out there, then it's different. You get the nuggets. You own the nuggets. Now, of course, if you're using AI to create something, there's the argument of that there's a little bit of a narrower moat that someone can just come along and do the exact same thing.
[00:08:57] So there's kinda gotta be something else that's special about it beyond just that AI got you there faster that makes that competition problem not a problem for you. But if you are creating things and have more of an entrepreneurial mindset, then using AI to think about what can you create faster, testing thing faster, building things faster.
[00:09:19] In my own company, there's so many things that I use AI for now cost me less to do. But the key here is it's not just that it's faster for me, it costs me less, not the company less. And that's kinda where the distinction lies. I'm getting all of those advantages by adopting AI in those ways
[00:09:39] Dr Genevieve Hayes: So obviously, if you're getting the advantage yourself, you're gonna want to adopt AI. And even if you're an employee, you still wanna have AI skills because you don't wanna fall behind the pack. Although granted, you can learn AI skills at an increasingly fast rate
[00:09:56] but if these AI skills are becoming increasingly easy to obtain, and the skills that AI enables, such as software development, are becoming increasingly easy to obtain, what skills should people be focusing on acquiring instead?
[00:10:13] Lauren Pearl: There are so many skills beyond just learning how to use AI to do closing your books faster. I'm in the finance context, and there's all of these AI applications. Closing your books faster is one of the biggest things that folks talk about. Skills. So I would think about In this context what is critical and what is rare?
[00:10:33] So critical, I want to think about in order to use AI, how do I do it safely? Anything that runs under the realm of safety is a thing that I would focus on, because even if you're bad at it temporarily, that could have really bad outcomes. So for me, when I'm working in AI, I work with a lot of security companies, and so I've been doing a lot of research to AI safety and how CFOs should be thinking about risk when it comes to AI and how risk has changed from living in an AI world so I think getting smart in that area, especially if you're a leader of a company, is extremely necessary because you're just in such danger now and there isn't yet a best practice.
[00:11:16] The AI security knowledge is also rare but that unfortunately in this situation as a consumer of it is really frustrating, 'cause you would hope that it would be really transparent and really heavily researched and out in the open all the time and it's not yet.
[00:11:30] So if you're gonna nerd out anywhere on AI, I would choose security of AI, which is what I've chosen. But then there's the stuff around AI, outside of AI. As a data scientist, I think there are skills that have always been important that are gonna become even more important.
[00:11:47] Judgment, I would say, is the biggest one, but that's a really hard thing to say because you can't just say build judgment." There's so many skills involved in judgment. As someone who's not a data scientist, I'm not sure I know the activities that will go into building that skill for yourself, but I can describe why it's necessary.
[00:12:06] And I think that's because having worked with AI in a technical setting, it moves so quickly and can calculate things that look right but are wrong in so many different contexts, in writing, in analysis. Sometimes you look at it and think, "Oh, it's so brilliant," and then other times you realize that, wow, it wrote out this entire analysis and I thought it was something really clever, but actually when I look at the sentence, it's not really saying much of anything, or it's actually saying the opposite of what I think is true.
[00:12:35] I think as we're using AI more often, having your own judgment to recognize that is super, super important because otherwise the more you use AI, the more quality gets threatened, unless you have the ability audit the findings and say, "This is trustworthy, this is not," or, "This requires more investigation to see if it's trustworthy," and guiding that
[00:13:02] Dr Genevieve Hayes: It sounds like what you've just described is depth of understanding in your specific area of expertise
[00:13:07] Lauren Pearl: I think, yeah, it could be. Which is also hard if you're using AI constantly to shortcut your research, then it's tougher to build those skills
[00:13:16] Dr Genevieve Hayes: I agree with you. You hear people saying, "What's the point of university because it's not teaching you practical skills?" And I agree that universities should help prepare people for jobs, but there is also value in understanding the theory behind something.
[00:13:32] Because, for example, when I'm looking at AI models, I'll go back to thinking "Okay, what did I learn about machine learning back when I was studying this at university? What are the fundamentals of machine learning? Okay, this is how a machine learning model works. This is what its strengths are.
[00:13:51] This is what its weaknesses are." Those allow me to make judgments about how these AI models are working. So just getting , those foundations strong helped me in my job. I would imagine if someone was working in, the healthcare industry, just understanding how the healthcare world works would help them to get that judgment about whether something looks reasonable or not
[00:14:17] Lauren Pearl: Yes, I think that's true.
[00:14:19] Dr Genevieve Hayes: So it sounds like if you're trying to succeed in the AI age, yes, learn about AI, but the other skills that you wanna target are good judgment about AI outputs so that you can evaluate them, and also it's what Cal Newport described, rare and valuable skills, because you wanna have that moat whereby it's harder for people to catch up with you, so therefore there will be a premium paid for those skills.
[00:14:49] Lauren Pearl: Totally. I think that's right. You're looking for those skills that are still rare, and if you have those rare skills, it also means whatever you do with AI is going to be so much more powerful because you're leveraging the stuff that, rare skills are the things that AI's gonna be bad at because AI needs a whole bunch of infinite data to build the pattern match on those things.
[00:15:09] So you're more likely to build something around a wealth of information you have that others don't, which all makes it harder to replicate if you're in a context.
[00:15:18] Dr Genevieve Hayes: Yeah. And I just, th-thinking about, rare and valuable skills, so the way you were phrasing that, it reminded me exactly of Cal Newport's book So Good They Can't Ignore You. I assume you've read that. Y-you haven't. You basically just replicated the central argument of Cal Newport's So Good They Can't Ignore You.
[00:15:36] So he's basically saying that if you wanna build a career, you should focus on rare and valuable skills because of exactly the reasons that you just outlined. And one of the examples he gives in his book, which was written well before AI, was someone who does a 12-week , course, and becomes a yoga instructor.
[00:15:56] If it only takes a 12-week course in order to become a yoga instructor, then in the space of just 12 weeks, anyone can become a yoga instructor. Whereas if it takes 12 years to become a nuclear physicist then that's gonna be far more valuable because most people aren't willing to devote 12 years to learning something, whereas just about anyone, if they really wanted to, could devote 12 weeks
[00:16:23] Lauren Pearl: It's a really good point. It also makes me think . If you're at the beginning of your career and you're trying to decide what skills won't get taken by AI, two things I would be thinking about. One thing is that we have no idea really what jobs are going to be completely taken by AI and which jobs will remain.
[00:16:41] So that's the one. And people who are predicting, they're making a prediction, they're making a guess, and we don't really know exactly how they to shift. Ob-obviously you can pick something that doesn't appear to be in complete takeover territory, but most jobs are probably going to be impacted.
[00:16:57] Point two, which I think is more important, would be I still think picking something that you're really passionate about that like you just never would get sick of, that you're totally a nerd about, that like really hypes you up is probably a very good bet because of what you were describing with that doing something for 12 years rather than 12 weeks.
[00:17:17] It is most likely the case that other people will not be able to replicate the thing you have devoted an inordinate amount of time and passion and excitement around. You are likely to over-index on something that you really love and that you can really be obsessed about. It's great if it's a weird thing, right?
[00:17:36] When we're talking about rare skills. But from what I've noticed from AI outputs, AI hasn't been able to produce results that are equivalent to what I can do when I'm doing something that's really in my sphere of passion. There's a couple of folks who have talked about this around using AI for different tasks, and I'm trying to remember who gave this quote, but I think it was from a study where they were showing that for people who were non-experts, they overestimated how good AI was in fields where they were not experts.
[00:18:11] And when you are an expert using AI, folks will say that AI is actually not very good in their area. So this idea that you perceive AI to be really good in the thing you don't know how to do. And it makes sense, I just used AI the other day to create a video game, and that was so cool.
[00:18:26] You and I are both nerds. I was so excited. It was like an old school, Nintendo style, block characters like 16-bit style. And I was like tickled. I was like, "Oh my gosh, I can program a video game. This is amazing." AI is great at programming video games. But my friend who works for a video game company would probably have a totally different impression of its output and its ability because they're working in the top point, .000001% of video game developers.
[00:18:54] They're gonna look at AI and be like, "No, this is crappy. It makes terrible stuff." So again, it's going to change a lot of things, but still for those folks, if you're willing to get to that like .00001% of people in a given field, it strikes me that there's still going to be value to bring there just by nature of how AI works regressing to the mean.
[00:19:18] I could be wrong, but that's my running theory.
[00:19:20] Dr Genevieve Hayes: I actually get my best ideas from asking AI to solve a problem, seeing the crappy solution AI gives me, and then saying no, this is stupid. This is wrong because of A, B, C, D, E." And then I look at my account of why this is wrong, and it's actually, that's exactly the answer that I wanted.
[00:19:40] And I'm doing it, but I can't get that answer until I see AI's crappy answer and be screaming at it as to why it's stupid
[00:19:48] Lauren Pearl: That's the academic in you. You can tell you're a PhD because you're like, "Ah, academic discourse. All I need is a dummy and then I can spout brilliance."
[00:19:56] Dr Genevieve Hayes: Yes, exactly. I need someone that I can have an argument with for eight hours straight, and I could not do this to another human being and yet AI is great because, You can spend all day telling it why it's wrong, and it says, "Yes, you are right.
[00:20:13] I am wrong."
[00:20:15] Lauren Pearl: This is very funny because actually, so I have the opposite experience when I'm dealing with AI and it's getting something wrong because my experience is more as a colleague or an employer or a teacher. And what I've noticed when I'm working with like an employee, I'll have the experience of moving someone from intern all the way to like a senior analyst.
[00:20:38] And with a human, when they're an intern, they make dumb mistakes, and they don't have a lot of judgment at first, but you can build that over time. It's this organic experience. They try, they fail, they get back up, they try again. You get a little bit of improvement. You reward them. They try something that you discover it's in their area of passion.
[00:20:57] They get a little more confidence. They're better at everything. It's this whole experience that I'm very familiar with, and, there's obviously some dips, but there's a general trend of overall improvement. And doing the same thing working with AI, I'll often get frustrated because you don't have that sense of linear improvement trend.
[00:21:16] The models continue to change under your feet. You can build skills, but skills are not a human brain being plastic and just absorbing things itself. You can train it to write properly, and then the model shifts, and then suddenly it writes a completely different way. You have a certain kind of instinct, a certain way that it works, and then it changes.
[00:21:34] It doesn't learn in the same way, and for me, that's a frustrating experience, but I'm glad that for you it really works.
[00:21:42] Dr Genevieve Hayes: Oh, it works really well.
[00:21:43] So we've just discussed how rare and valuable skills take years to build. But what's one thing our listeners can start doing today to start developing those skills that will ultimately lead to becoming rare and valuable in the AI era?
[00:22:02] Lauren Pearl: So I think one is thinking about where your passion lies and thinking about where are the places where you want to have a little bit more obsession. And there's a lot of folks that are getting very excited about vibe coding in general, and that's interesting.
[00:22:18] But if you're interested in rare skills, maybe pair that vibe coding with the rare skill. So thinking about what it is that you wanna develop outside the realm of AI, and you can use AI to accelerate you there or build something that does something there. But just focusing on, "I'm going to learn how to vibe code, and I'm just going to vibe code a bunch of tools " is just not gonna get you typically, I think with how this world is headed.
[00:22:45] In terms of rare skills, I think too, I am really angling towards the touching grass sort of opportunities. So this is a great time to be building relationships, to doing things in real life, to building people skills. AI actually has some great ideas for, like, how to send emails and how to think about people situations.
[00:23:06] It's been trained on a lot of emotion stuff, and that's cool, but it doesn't rival just in-person experiences and getting that social interaction and thinking about working with people finally, actually, I think collaboration, I'll stick on that as the last point I would go searching for skills.
[00:23:22] As folks get more obsessed with AI, there's this sort of almost meme-level trend of people that then just they sit with the AI, and they get bags under their eyes, and they're just vibe coding 24/7, and they're just sitting there with all of their Mac Minis and not talking to their family or eating or sleeping or anything like this.
[00:23:42] And yeah, if you have a specific project and you really wanna get it done, I get it. Building with AI can feel like you're flying. It can feel amazing. And it's really important to still think about collaborating. AI gives us the opportunity to build as an individual things that used to take 20-person teams, which is cool, but in the same way our skills in judgment can atrophy if we're outsourcing all of our judgment to AI.
[00:24:08] I think our skills in collaboration and working as a team and just coordinating all of that, staying happy, staying motivated, creating more product from two people than just two individual people those skills are going to be really important to yourself and also to getting beyond yourself.
[00:24:27] Just you and AI isn't actually a broad enough insight because AI's not a dive deep expert on anything, so you need other dive deep experts to truly round out your knowledge. So I would say that's the third bucket where I'd go looking at skills.
[00:24:41] Dr Genevieve Hayes: For listeners who wanna get in contact with you, Lauren, what can they do?
[00:24:44] Lauren Pearl: They can do a couple things. You can connect with me on LinkedIn. I'm Lauren Elizabeth Pearl on LinkedIn to distinguish me from all the other Lauren Pearls. You can also sign up for my newsletter. I have a newsletter called The Daily CFO, which is actually where the AI chicken nuggets article first was published.
[00:25:02] So you can find other crazy insights like that on the newsletter. And actually coming soon, I'm starting to experiment with an AI security newsletter. So I told you that I was nerding out a little bit on AI security, and noticed that for finance professionals like me, there wasn't really a resource that gave me a snapshot of, like, where we're at with things.
[00:25:23] Have there been breaches? Is there new security protocols for certain platforms that I use? Has OWASP put out new guidance on anything? And so I decided for my own benefit to create a kind of a snapshot that I could read on a regular basis, and I wanna share it with everyone else, so everyone else has an idea of what's the perspective.
[00:25:43] Again, if we have this newsletter and we read it week after week, we will build this rare skill of understanding what the heck is going on with AI security.
[00:25:53] Dr Genevieve Hayes: I love people who do things like that 'cause it saves me having to do the research myself
[00:25:58] Lauren Pearl: so that one's not out yet. It's still in the works, but you can watch on LinkedIn. I'm basically posting the experiments of trying to build it in different ways, and you can take a peek there and when there's a time to sign up, that's where I'll share it.
[00:26:10] Dr Genevieve Hayes: And that's it for today's episode of Value-Driven Data Science. But if you want more from Lauren, next week you can catch our special reverse interview episode where we'll explore why large language models struggle with some tasks and excel at others, with Lauren taking over as host. And if you found today's episode useful and think others could benefit, please leave us a rating and review on your favorite podcast platform.
[00:26:37] That way, we'll be able to reach more data scientists just like you. Thanks for joining us today, Lauren
[00:26:44] Lauren Pearl: Thank you so much for having me. This was very fun. I'm excited to interview you.
[00:26:49] Dr Genevieve Hayes: And for those in the audience, thanks for listening. I'm Dr. Genevieve Hayes, and this has been Value-Driven Data Science.