In this episode of The Parlor Room Presents: Hello AI, host and Harvard Business School Online Creative Director Chris Linnane speaks with HBS Lecturer Jake Cook about how artificial intelligence is changing the way people work. Cook shares practical advice for using AI effectively, explaining why it should be treated like an intern that requires direction, oversight, and critical thinking. He explores AI’s opportunities and limitations and why human judgment remains essential in an increasingly AI-powered world.
Resources
Catch up on previous episodes of The Parlor Room, featuring Jake Cook:
Compilation Episode (Part 1): Early Career Advice in the Age of AI
Compilation Episode (Part 4): How Mid-Career Professionals Can Lead and Grow with AI
You can also watch The Parlor Room Presents: Hello AI on YouTube.
Transcript
Editor's Note: The following was prepared by a machine algorithm and may not perfectly reflect the interview's audio file.
Jake Cook: You can make terrific time and go nowhere.
Voiceover: Jake Cook teaches marketing and AI at Harvard Business School.
Jake Cook: Treat it like an intern, and maybe one that partied really hard last night.
Voiceover: His work moves between the classroom and the real world.
Chris Linnane: What is it about this AI wave that feels fundamentally different and not just maybe more intense?
Jake Cook: What's different about this one was two things.
Voiceover: He's not just studying this moment in marketing and AI. He's living it.
Chris Linnane: What is it that excites you about AI?
Jake Cook: What has me excited is...
Voiceover: He's asking questions that don't have easy answers about judgment, priorities, and how leaders decide what's next.
Jake Cook: I've done this a long time. It's not that difficult. It shouldn't be that hard. And then when you say, "What biases do I have?" And wow, does it kind of hit you between the eyes.
Voiceover: Questions shaped by both theory and practice.
Jake Cook: Does it need to be made, and does it meet what the business is trying to achieve?
Voiceover: This is The Parlor Room Presents: Hello AI.
Jake Cook: The future's going to reward the playful and curious.
Chris Linnane: Over the last year, when you're working with students and companies, was there a moment where you paused and thought the way these people are approaching AI could be problematic?
Jake Cook: That's a fantastic question. I've seen that in the classroom and with clients. Initially, I think some folks have approached this and kind of treated it like Google, or we're just chatting on the internet. We've done that for 25, 30 years. What's the big deal? And this is changing exponentially. And so you can write something off 90 or 180 days ago and say, "Oh, this can't do that. " And then it can. So I think that's one thing is, early dismissals.
The second piece I've noticed was with my MBA students—they can take a break between their first and second year. So some have chosen to go do startups or whatever, and they're kind of coming back maybe three years later, and for whatever reasons, they weren't exposed to AI as much or gen AI and large language models. And I think the anxiety when they get on the classroom and to see that everybody else is fluent and feeling like they're a little bit—and we all kind of feel a little bit behind with this stuff in the pace of change.
But those students, I noticed, really felt like, "How to prompt? What's it doing? How do I trust? How do I build this kind of sixth sense of when to use it for certain tasks?" They seem to really have a lot of anxiety around that and feel like they were left behind. So I think that what can be down the road for certain organizations or even individuals, all of us have to kind of rethink our roles and how we create value. I think that the idea of constantly testing and what it couldn't do yesterday, it might be able to do tomorrow. And so there's a certain level of just being comfortable with the chaos.
Chris Linnane: Oh man. So for me, very uncomfortable with the generative AI because, like you're saying, we'll make things, and I'll be like, "Oh, that's pretty good." And then instantly, three months later, I'm so embarrassed about the thing I made three months ago.
Jake Cook: Yeah.
Chris Linnane: How do you get that mindset where you think, "I'm building, but everything I'm building is kind of like a draft because it could always be better."
Jake Cook: I think Ernest Hemingway had a great quote that the first draft of anything is garbage. And so I always think that idea. One of the things I think that helps a lot is pulling some ideas from the world of design thinking, or growth mindset, and Carol Dweck's work around this, but this idea that V1 doesn't have to be perfect, and it's okay to just make things. And so a big thing I really push with the students and with clients and our own team and my company is, "It's okay to play, and it's okay to fail."
And we can talk a little about little micro experiences I've kind of built for clients and students to fail in a small way, because from that failure and where you find where that is, you're kind of failing in private a litle bit on certain low-risk tasks and therefore you kind of know where it's going to break or what I can and can't trust on that side of it, which I think is a really good way to build some confidence in it and not the first version is polished and perfect.
Chris Linnane: Yeah. No, it's interesting because when you think about putting things out in the public, there's the drafting on your own and being like, "Oh, I could do it better." But I don't know why it always comes to my mind that the embarrassment of like a perm—like a male perm in the 80s—because my dad had one, and we found this old videotape where he was running down a hill with no shirt and a perm, and we're all like, "What is it?"
Jake Cook: And short shorts?
Chris Linnane: Yeah, or jeans, something. But it's such a specific time, and it has such a specific look to it. How do we know that the things we're putting out right now won't have that, like, laughability 10 years from now?
Jake Cook: That's a great question. And I think when you look at the image generation of AI, what it could do even a year ago, when you think in kind of 2024 and 2025, it was laughable, right? You'd have it, it had seven fingers, and it kind of looked like an alien, and it almost looked like the Men in Black-type of villain from that movie, and now it's way better. But some of these, like imagery, I think is a good way you can kind of sense this feels very kind of cartoonish, but now with some of these new models, the imagery is insane and the video as well, and some of that stuff.
But you make a good question around what will become kind of taste visually and in written forms, and how does that get created? Do we go through quicker, faster cycles now? There's also this kind of idea that large language models learn from lots of data, so everything's going to kind of converge to the mean, and I think some skeptics or cynics are more like, "Now we're just going to have a bunch of slop, the AI slop," and everything's kind of—yeah.
Chris Linnane: That's the biggest concern I have right now. So kind of building on that, you've seen in your career lots of different waves of technology pass through. What is it about this AI wave that feels fundamentally different and not just maybe more intense?
Jake Cook: One of the advantages of getting older is that history doesn't repeat, but it does rhyme, as Mark Twain says. I think what's different about this one, compared to say maybe the consumer internet or mobile, was two things. First of all, access is necessarily just dictated by the ability to pay. For example, you can get access to the world's best models for free in many cases. I think that's a big idea, like the democratization of access to some of this technology. Conversely, a lot of the models are trained off of data on the internet, and so if you're in different countries with different languages, there may not be as much training data. So there's things there that are still kind of a mirror of the real world. I think AI is really built on data, and data has biases and flaws in it, just like humans. So there's some things there that I think don't go away with AI as well.
The second thing, in addition to the access, is kind of the rollout, where every interface that we interact with, every screen is connected, almost everybody in the world is connected to a screen now, and the stuff's getting pushed like hourly almost. So if you log into Instagram or you go shopping on Amazon or—anything now has an AI layer put to it. And so when you go to your doctor's office, it's like, "Can we take your electronic medical records in your notes?" And like it's in your face everywhere. And while the technology's moving super fast, I don't think it necessarily means that humans are going to adopt it at the speed at which it's getting rolled out.
As humans, we tend to adopt innovations in a very kind of predictable bell curve. And so I think we'll continue to see a lot of friction around that for sure.
Chris Linnane: Now, you personally, looking at AI and where we are right now, what is it—I don't want to assume—what is it that excites you about AI, if it excites you? I think it probably does.
Jake Cook: It does. And I think it's interesting. I think there's certain folks who are, "It's going to be amazing, and it's idyllic and utopian." I think it's kind of like the internet. It's going to have some, like any new technology, even when we invented fire, there was some drawbacks, some people lit things on fire they probably shouldn't have as a result. But what has me excited is I think a lot of how people approach their careers and their craft could be stemmed with, maybe, we all have insecurities around things that I'm not, for example, "I'm not creative." Well, if you ask, "Why aren't you creative?" "Well, I was never really good at drawing." Oh, so you never had the feedback that you're good at taking a tool and creating an artifact visually, but you probably, in your mind, expressed, maybe you couldn't get it from your brain to your hand to draw and shade this charcoal drawing or whatever, but you knew what that should look like. You had some sort of a sense of what that could be, or you could iterate through that.
That's what I think is exciting about AI. So it provides this new interface for us to kind of express ourselves, and people will do nefarious things with that. And I mean, that's on a whole spectrum of things. But, I've seen even with students, one of the best things that comes out of working with them is you watch them kind of start the class, and they have anxieties, or they're nervous, or, "I'm a computer science undergrad, I'm not creative." And then we'll do some exercises, and they find out that turns out, lo and behold, it was an interface challenge and a tool challenge, but they had ideas, and they just couldn't. And conversely, "I'm not technical, I don't know how to do code and all that stuff." And all of a sudden, they're building an app. And I think that's what's really cool about it, similar to what the PC did.
Chris Linnane: Yeah. So it's interesting. So if you look at it that way, there are people who have this creativity that's kind of locked, and these tools help them unlock that creativity.
Jake Cook: I think they do.
Chris Linnane: What happens to the person who was very gifted at something, and do they have the same kind of lift from this, or is it more of a detractor now because everyone can do what they're doing or can get close to what they're doing?
Jake Cook: It's an excellent question. And in my opinion, I would say, I think what's interesting is, if you've done a craft for 10 years, if you're a world-class copywriter, you've got a sense of taste, trends, historical things. If you're working internally at a brand, you kind of know what's resonated, what hasn't. That's kind of, to somewhere you could be encoded in an AI if you can get that into a text format or whatever, and put that into AI. But I think making that next big jump creatively or moving your brand into an interesting direction, I still feel like there's a huge amount of opportunities for that. Also, if you've worked in creative, one of the hardest challenges is getting approval.
Chris Linnane: Yeah, definitely.
Jake Cook: And when you have, depending on the culture of, "Well, the CFO saw this Instagram ad, and they did it." And everybody, and I think I always use this phrase: It's easier to be a critic than a creator.
Chris Linnane: Oh, without a doubt, yes.
Jake Cook: Yeah. So it's way easier to stand up and throw the tomatoes in the stands down in the arena. And I think now you can say, "Whoa, well, why don't you take a swipe at it? I don't quite get you. Let's see your vision." And then you can kind of have people similar to grabbing a pencil at a whiteboard or something, I think that gives you a chance to kind of maybe communicate a little bit better. And oftentimes, it's the data that comes with running creative, too, is like what you'll see a lot of times in campaigns is, "Oh, that ad's going to crush." And turns out it's the shaky iPhone, user-generated content,
Chris Linnane: Yes.
Jake Cook: The customer's sweaty, but they're out there using the product in the field, and that just is amazing, and it's so authentic, and people love it.
Chris Linnane: Oh, so yeah, that actually brings me down this little path I'm wondering about. So just the volume—and we'll talk about this a little bit more—but the volume of content we can create right now, and I know you specialize in analyzing data and figuring out where we should go with different things.
So, say there's a company that has an ad, it seems like what we do now is 7,000 variations of an ad. And I wonder, and I might be wrong, but I imagine in my head a concert in a triangle with three different musical acts all facing the middle. And you're in the middle listening, and it seems like whoever's loudest wins now. They're the best musical act because they're the loudest, where it seems like now the other musical acts wouldn't be like, "Well, let's write a better song." They'd be like, "How do we get these amplifiers to be louder?" You know what I mean?
Jake Cook: Yeah.
Chris Linnane: So, how do we think about this, where is the louder always going to win, or is there a chance for the person or the group or the company that's quieter but did a really good thing, can they compete nowadays in this world?
Jake Cook: Yeah, that's an excellent question. I would say, "What's timeless?" And I think about what's kind of timely and what's timeless in kind of two dimensions.
And I think what's timeless is brands that resonate with a customer that have authenticity or whatever that may be. And at the end of the day, if you look at Google's IPO letter, the founders say, "Our job is to basically put the best results in front of the user." In 2004, they write this. A lot of the reasons Meta and Facebook won initially with social was like, they just did a better job of helping you connect and find—so that idea of like the algorithms will serve what will get the engagement and there's a dark side to that a little bit too as we're seeing, but you can't actually buy your way to the front and have it hold your place in line. So running ads and screaming and yelling, that'll work. And I've seen a lot of brands do that, but sooner or later, whatever, it could be a month, it could be a week, it could be a year or two.
But sooner or later, if you don't have some sort of durability and some stories and some of these timeless components that make brands go, with products and services that deliver on that promise, it actually happens faster and unravels faster on the internet than it did before.
Chris Linnane: Oh, wow. Okay. Just as a scenario, you've got a company that puts out three web banners, and the variations are like, "Buy now," "Try it," and "Learn more." And that's the change between them. Are we literally just looking at which one gets more clicks, or are there other factors we're looking at?
Jake Cook: Yeah. And pull me back if I get a little too nerdy on some of this stuff.
Chris Linnane: I'd love to hear the nerdy explanation.
Jake Cook: I think the data science in me, scientists would kind of think more about this idea of statistical confidence. Have we shown enough? And you can get this idea, what they call local minimum maximum. We're testing this call to action around "Buy now," blah, blah, blah, like something really small, but all our competitors could be testing the same thing, and you can get some confidence, but you're not really finding what's going to break out and get you big, big orders of return. So you might have to be like, "Okay, we're going to do, of those three, one of them is going to be radical, like not, 'Buy now,' but like 'Don't buy this.'"
Chris Linnane: Yeah, sure.
Jake Cook: Using the Patagonia idea from Black Friday, Cyber Monday a few years ago, you might do something super radical like that. And that's where something, where the AI maybe isn't giving you those clues when it's doing some iterative brainstorming. Maybe it's learning from historical data, but there's nothing in the data that shows doing something totally at 90 degrees to what you've done. I think that's the part where the human can kind of come in, test some stuff, and then, and who knows, it may or may not work, but the scientific method of like, "Okay, we have some ideas, we're going to take a hypothesis, we're going to test it." But if all the hypotheses are really close and tightly knit, you're not necessarily having any sort of breakthroughs on that side of it.
Chris Linnane: When you see teams struggle with AI early on, what questions do they usually start with that send them in the wrong direction?
Jake Cook: I think this kind of comes down to a couple of pieces primarily related to culture. So, if the culture doesn't reward experimentation or it's okay to fail, if they sort of expect perfection, which no one wants to ship sloppy work, right? But inherently, AI is probabilistic versus deterministic, meaning that there's a problem at its core. The way it's working is it's kind of guessing what's going to come, whether it's a pixel or a word, it's just making a probability of a guess, which means inherently it's going to be wrong, just like human beings are wrong. And so culturally, if they're penalized for that or they give it a task that's too broad and complex, and so it's almost like they'll hand something, maybe some approaching, I always say, treat it like an intern, and maybe one who partied really hard last night.
Chris Linnane: Oh, okay. Yeah. So, limited expectations.
Jake Cook: Yeah, limited expectations. It might be a little bit fuzzy, but it wants to please you in a very big, big way. And so when you hand it a task like, "Come up with our go-to-market strategy for next year and what products we should develop," you would never in a million years deploy capital off of an intern's recommendation on that, right? But we kind of treat it the same way, "Okay, well, tell me what to do." And then what happens internally with cultures is some people are really fearful of their jobs. "This is going to replace me." And when it fails, they cheer. "I told you it couldn't work; you need me."
And so that can be kind of this false negative with the AI, or a little bit of like they're still kind of building a sandcastle on the beach, hoping the wave's not going to come in. And so culture, I think, celebrating failure, building small micro experiments, and modeling that. One of our clients, the chief data officer, is wonderfully brilliant, very, very innovative, man, super great to work with, and he started going through in their all-hands and just being like, "Okay, I'm going to walk you through how I think we should do this. Here's some data, let's go through...let's do it live." And when people kind of see leaders or their managers kind of doing this and failing, I think that's huge.
Chris Linnane: It's a great signal that there's space for me to do this as well.
Jake Cook: There's space for me to do it. Yeah.
Chris Linnane: And when you are coming up with a plan for something to test with AI, how do you know? I know it's case by case for the most part, but I think the biggest fear would be, "I didn't plan this correctly, and I won't get to the result I want because I didn't build a framework that would capture this." Or the other version of it is, "I took too big of a swing." So how do you know, I mean, it's case by case, but what's an example of the right fit for your test?
Jake Cook: So I think about it, if I were to try and shift a culture to adopting this approach with the mind, a lot of them are going to be fearful, nervous, but the way you can kind of trick them into using it is like, "What's one part of your job do you hate?" Have them list it. Okay. We're going to have a Google Doc, or we start a document or an email chain like, "Oh, I hate doing campaign analysis. A lot of market with the creative stuff. I don't want to deal with the data stuff. It's super boring for lots of reasons."
Okay, well, as a team, on Friday, we're going to have a lunch and learn. We're going to come in, and we're going to see if we can analyze what worked on Meta over the last year. Something that's kind of large and onerous, we get a CSV data out of Meta, and we're trying to kind of figure it out. And then we start to live in a real-time test, "Okay, here's a CSV. Which campaigns worked for click-through rate or impressions?" Okay, blah, blah, blah. And I think from that, people can say like, " Oh gosh, I actually don't want to get really good in Excel and the tables and whatnot." And they start to see what's possible. And then I think when they start to see, like, "Oh, it's an enabler." Just like in the old days, as I remember when it went from the library to Google, and like, "Oh, I don't have to go to the cranky librarian and get a lecture on the Dewey decimal system?"
Chris Linnane: There's a big improvement there, for sure.
Jake Cook: And that kind of idea, "Now we save all this time," and now what we—and so there's a certain level of like, " Hey, we're going to give you time back to do some of the stuff you like to do."
Chris Linnane: Yeah. Now that just reminded me, I had an extensive set of Encyclopedia Britannica in my house, and it was like every time a book report came out, or any kind of historical report was like, go to that same shelf. I get my stuff. And I can't say I feel bad for the Encyclopedia Britannica people when Google came around. So I just got to keep looking forward because, I'll tell you, Jake, I do get a little nervous about where everything's going right now, but I did enjoy Google over those 27 editions of Encyclopedia Britannica.
Jake Cook: And what's interesting, just as an aside, I was doing a presentation, a workshop for a company on how to kind of all-hands the entire company and how to think about this. And so for fun, I just pulled headlines from newspapers, and I was like, "Is this current or is this AI? When do you think this happened?" And I pulled headlines from 100 years ago when we kind of electrified the Northeast here in the US, and you see almost exactly the same type of headlines.
Chris Linnane: Really?
Jake Cook: "This is the downfall of society. "This is going to wreck the youth." I mean, on and on and on and on. "If people can have light at night, they're going to do all sorts of bad things because they have more visibility."
Chris Linnane: Is that one of them?
Jake Cook: It was totally. Yeah. They're hilarious if you go read them because now we kind of laugh and we think like, "Oh gosh, how ridiculous." But there are things like, "Crime's going to rise because criminals will have access to doing all these things in near daylight." And so that part is, I think it's kind of a funny tongue-in-cheek way, but when we see a new technology as a species, we always assume the worst, and there will be some bad things with it. I think sometimes in the AI conversation, there's a lot of this utopian-like, "Oh no, it is disruptive." It is going to be. We're already seeing that.
Chris Linnane: "People will commit crimes if they have light at night" is one of the funniest things ever. I'm going to have to remember that. I enjoy those kinds of things a lot, so that's very funny. So you describe AI adoption as something that teams need to grow into over time. And say that's going well, what does that look like? What does the growing look like?
Jake Cook: I think that, again, the cultural drives all this stuff, in my opinion. That idea of let's assume that we're celebrating failure, we're playing with new models, we're not becoming entrenched. And one thing I see a lot is like, well, again, that didn't work. Well, that didn't work three years, six months ago. It's not linear, it's exponential. That's a big difference to approach things with. So I think they're revisiting tasks consistently on that, maybe that was beyond the reach of AI, and they're kind of finding that jagged frontier. They're also building that kind of cultural sense of what's possible and what their team or their brand or, more importantly, their data can kind of guide them on.
And that's where you got to find the edges of what's possible. It's not unlike training for a marathon. If you go out and try and do a long 15-mile training run, I've never run a marathon for these reasons, but I did train for a little bit for a half-marathon, and you just don't start off running nine, 10, 12 miles; you start off with two or three, and then you kind of move yourself into the hard pieces of that. And I think that's huge.
The other thing, back to adoption, is finding people, their peers are going to drive some of those changes more than top down because every leader is saying like, "Oh great, I can remove all this labor from the P&L. I can fire you, and I can fire you." And there's that kind of concern and fear, and quite frankly, given this interesting environment we're moving in, the cost of capital, there are things to reduce labor and things like that too.
And I think when people can kind of say, "Hey, your peers are going to help you." And "Okay, I need to do that." And that's one thing I really push students in particular, is like, you're going to kind of come out of your MBA program, and people are going to expect you to lead with some sort of fluency around that. And a lot of what you find when you get out in school, and you have to lead and get handed teams, is like, it's not as easy as the business case you read.
Chris Linnane: Yeah, it can't be.
Jake Cook: Yeah. It's a lot messier and a lot harder. And so I think that idea of modeling things, getting people, and then finding people that really want to lean into that. And I think Professor Avery's got some stuff around personal branding and branding in general, and like, is that part of your personal brand? Like, "Hey, I'm curious, I'm playful, I'm not afraid of these tools." That's a good way to kind of build my personal brand. I can then be seen as a peer and help others. I think there's a lot that can be done on that for sure.
Chris Linnane: So, just constant flexibility, it sounds like.
Jake Cook: I think so. I think the future's going to reward the playful and curious.
Chris Linnane: It always has. So what changes when AI stops helping people execute their work, and it starts helping them or starts helping shape their decisions?
Jake Cook: I think what's really fun about AI is, again, once you kind of understand how to prompt her to work with it a little bit, and you, as I say, kind of kick it out of grandmother mode because every grandmother will tell you their grandkids are smart and good-looking and talented and athletic, and AI will kind of do that to you. And once you kind of teach it to promptly, "No, no, no. Give me the brutal PE coach."
Chris Linnane: Yeah, no, the tough one.
Jake Cook: The tough one, yeah. I think when you do that, then you can have it kind of use it as a thought partner to critique decisions. And that's what I personally really enjoy is like, "Where's the bias I have in this?" I'll go to teach a topic, and I think, "Okay, this is super search engine optimization, SEO. I've done this a long time. It's not that difficult. It shouldn't be that hard." And then when you say, "What biases do I have? What decisions should I make to unfold a lesson plan or explain this thing to it?" And wow, does it kind of hit you between the eyes.
Chris Linnane: Yeah.
Jake Cook: So I think that's a huge one on kind of shaping decisions. If decisions are driven by data, we can then, whether we mean to or not, we can exclude data we don't want. And I call this the Apple Watch problem. You're supposed to take 10,000 steps, but did it change my behavior? The data's telling me I took 5,000 steps. "Well, maybe tomorrow I'll get to 10,000," right? So what do we ignore in the data? But when you basically take that data in and say, "Hey, am I being active for this many calories I need to burn a day, or how much results I need to get for revenue for this campaign?" That's where AI can really help you make decisions and say, "Hey, you're choosing to ignore some inputs here."
Chris Linnane: Yeah. And you said the Apple Watch thing, I could feel my...I'm like, I'm not reaching out on steps at all. So I love your terminology of the grandmother mode because that's the biggest challenge I have, because I mostly use things like ChatGPT and Gemini. It's like brainstorming.
Jake Cook: Yes.
Chris Linnane: I have the best ideas in the world.
Jake Cook: Oh, I'm brilliant. I mean, to the AI.
Chris Linnane: And it took me a couple of months to realize I'm not a genius. This thing's just telling me what I want to hear. But it's so interesting. And I think the PE teacher balance is good because that's where the reality hits. I had this PE teacher who was a former, I think, he's World War II Army veteran. So we used to play softball sometimes.
Jake Cook: Oh, yes.
Chris Linnane: And for anyone who doesn't know what that is, it's like baseball. And third base is a position that's closest to the batter, and it's kind of scary, so the ball can come at you really fast. He invented a new position, called Reflex, that played in front of the third baseman, which basically meant you had no time to defend yourself at any point. And that was a reality check where it was like, "If you act out in class, there was a real ramification." So that was not grandmama. And to add to that, he pitched every game and didn't wear a glove. So what was amazing, he's this really fit old man, you'd hit a line drive on him, he'd catch it in his bare hand, and you'd see his face absorb the pain. And I was like, "This guy's the real deal." He scared the heck out of me.
Jake Cook: He'd take a bite of it like an apple and throw it to third base.
Chris Linnane: Throw it back. But yeah, we need more of my gym teacher in these models to kind of be like, "Not a good idea, move on."
Jake Cook: Oh, it's fascinating. I think that's one thing too is when I'm downloading or teaching how to actually act, and people really love to watch you use some of these AI tools, and that always like, it shocks me. They start laughing because I'll just be like, "Give it to me straight, quit being nice. What are the flaws?" The other thing I think helps with decision-making is if you're in a domain, or maybe there's people who are thought leaders or certain academics or something, like one of I use a lot is I'm a huge fan of Charlie Munger from Berkshire Hathaway and he's got a book called Poor Charlie's Almanack, which is kind of a collection of all his talks he gave. And another one is Danny Kahneman, who wrote Thinking, Fast and Slow. And so you can say, "Act like Charlie Munger and Danny Kahneman, what biases do I have in my thinking and logic here?" And it will start to talk like Grouchy Charlie Munger back to you.
And so kind of taking on personas of people that, "Okay, if I could have a conversation and pitch this idea or this initiative around this big decision or capital deployment or whatever, and Charlie Munger was or Warren Buffett or whoever, my CEO or whoever's in the room, what would she say?" And it's like, wow, it's great.
Chris Linnane: That's a great way to frame it. That's more useful and probably a little more business savvy than my gym teacher, who would just tell you, "Absorb the pain."
Jake Cook: I never passed the pull-up test. I could do one pull-up all the way through, and in front of the whole class, I'll never forget that and just be like, "Oh, Cook, you better try harder next year."
Chris Linnane: Oh my God. And did you ever have to climb the rope to the top?
Jake Cook: Oh, totally.
Chris Linnane: I intentionally acted like it was tough because what are you going to do when you get up there? It's petrifying.
Jake Cook: Exactly. Yeah.
Chris Linnane: So I'd go one notch and be like, "Oh, I can't do it." Maybe I could have, but I knew well enough to not—once you get up to the rafters, that's the dumbest place you could be as a jock. That's what
Jake Cook: That's what I thought. Yeah.
Chris Linnane: Yeah. Can we play a game?
Jake Cook: Absolutely.
Chris Linnane: I like to play a game to try to make things a little more concrete. I'm going to give you a few, I don't know, a few scenarios that AI could potentially do in marketing or a human could do, and I'd like you to tell me which person's going to do it. Is it going to be AI? Is it going to be human? Or, is it going to be a hybrid? Okay. And to make this more fun, for the people who are watching on video, I have paddles representing each scenario. So I have the human, I have the AI, and I have the hybrid. And again, I'm not trying to brag, but I went to art school.
Jake Cook: I was going to say that's...I honestly thought these were stickers you applied based on that, so, excellent.
Chris Linnane: Okay. No, that's just amazing Sharpie drawings.
Jake Cook: It's very good.
Chris Linnane: The School of the Museum of Fine Arts in Boston.
Jake Cook: I love it.
Chris Linnane: Thank you. So you can hold it up, but we'll still say what it is, for the people who are just listening to this. Okay, perfect. All right. So our first scenario: Creative ideas, generating, getting new ad concepts based on past campaign performance?
Jake Cook: I'm going to say it's going to be a combination of robots and humans.
Chris Linnane: Okay. All right. So what's the human responsible for in this?
Jake Cook: I think the human may have context on which campaigns have worked historically from the data or what they've seen. Some of that stuff is not encoded in the data, per se. So if you get the return on ad spend or conversion rate or whatever, you maybe didn't know that, "We didn't run a lot of ads because we're out of stock on this product." So why did this product? "Oh, we were blowing it out." So there's context that the AI won't have where the human can be important on that. However, if you had to generate, again, "Give me five wildly different creative ideas for spring around these themes, and it drops on this date around this holiday," that is where the AI can quickly boot you up on something as well.
Chris Linnane: Perfect. All right. Scenario two: Creative choices. Little slides, I think. Yes. So, deciding which ad concepts will actually go live. And feel free to play along at home while we think these things.
Jake Cook: Oh, this is an excellent one. Let's assume I have full creative autonomy. I don't have to get approval from anybody. Then I would say, using a data set, I'd have the human ultimately make the call on that as well.
Chris Linnane: Okay. All right.
Jake Cook: There's some context here. I'm hesitating because I'm like, "Is this a new product? Is this a new brand? Where are we at?" And kind of a trajectory on that.
Chris Linnane: Yeah.
Jake Cook: But I think that's a part where the human ultimately...
Chris Linnane: And what would make you uncomfortable about AI potentially being involved there?
Jake Cook: I think what's really interesting is we look at, say, paid advertising, for example, where we can run lots of variations. We can now start to really use a scientific method. There's a little bit of, you're kind of trusting these ad platforms to tell you that the impressions are what they're showing these ads. And there's a little bit that kind of happens black box and is getting increasingly more so. But the idea of, if you want to come up with five campaigns, think about in print or TV. I mean, that could take you six, 12 months to run a campaign and get the data, and it's pretty leaky as it is. Now, with some of these things, we can do it in hours or days, which is really fun. So you can find kind of the signal and the noise quicker and run that at scale.
Chris Linnane: All right. You ready for number three?
Jake Cook: Yes.
Chris Linnane: Optimization: Deciding which ads to scale up and which ones to pause.
Jake Cook: I'm going to go...
Chris Linnane: Hybrid again.
Jake Cook: Hybrid again.
Chris Linnane: Okay.
Jake Cook: I think, again, you can run a bunch of stuff, see and test it. We are right on the cusp of doing a lot of creative in these ad platforms, but they're not quite there yet. Meaning that they'll have the alien or the seven fingers or it misspells your brand name, and even though how could you misspell that? So we're not quite there where you could just set it and forget it. Spin up creative and copy and all that kind of stuff. So you still need that oversight.
Chris Linnane: I feel like I could turn on my computer now when they would be there, though. That's what I'm nervous about. It's happening so fast that who knows? We're not there yet. I agree. But maybe it's a couple of hours away or a few months away. I don't know.
Jake Cook: Yeah, we're pretty close. And then this gets into how good is the data you're going to feed back into these platforms. For example, do you have good margin data, or do these types of ads drive high-value customers? And maybe it performs exceptionally well, but it gets you customers who buy and return all the time, or they don't stick around. So that's again, additional context and data will help clue the intern in.
Chris Linnane: Yeah. All right. Next scenario, scenario four: Recommending when a campaign should be shut down.
Jake Cook: I actually kind of like the robot.
Chris Linnane: Really? Yeah. Okay. I wasn't expecting that. So tell me why.
Jake Cook: And you mentioned recommend. So it didn't automatically do that, which I think is great. But back to the bias, if you create something, you get an emotional attachment, I call this the "Ikea Effect," where you move that terrible bookcase for 10 years because you spent four hours on a Saturday putting it together. I think you can build that connection to ads or campaigns or creative experiences to it, and you're like, "Yeah, but it was so brilliant and ah." And then yeah, "It's not converting." Or, "It's giving us bad customers." Or, "Man, the cost to get a click on this thing is insane." So I think the robot can be a little bit more objective and say the binary like, "Nope, didn't work, and these are the ones we should kill."
Chris Linnane: That makes total sense. I think about it from worlds I've been in, where making film or making songs, and they have that thing called rough cut love, where you get so used to the version you used for so long that when you actually film it, or you actually go to the studio and record it, you're like, "This sounds horrible." It's like, "No, it sounds great. You're just so used to the old engine." So yeah, the robot would be able to come in and be like, "Listen, I've analyzed this, and the studio version is much better."
Jake Cook: Yeah, exactly. Yeah.
Chris Linnane: Interesting. Okay. Fifth one: Recommending a shift in budget from retention to new customer acquisition.
Jake Cook: I'm going to go with a robot on that one as well.
Chris Linnane: A robot again. So, no need for a human in that world, or if they are, is it very small?
Jake Cook: I think some of that can be a lot of analysis and a lot of times in campaign managing. I mean, there's a feedback loop from getting the data to analyzing it to making the decision, right? Yeah. The Apple Watch problem. So I think the robot can analyze it, make the decisions, and then go take the walk for you, if you will, on that side of it, which I think is really interesting. Now you bring up a really key point, I think, around marketing is not just acquisition but retention. And in my work and industry, a lot of the brands we work with have gone on to be acquired by exits and things like that. And a lot of this comes into how good are these customers? How did you acquire them? How risky are those channels you acquired them through? And then, as a result, how much capital are we going to have to put into these things to continue to hit these growth curves and things like that?
And so I think that a lot of the hype in marketing acquisition, but there's so much—I can go on for hours about retention and how you can build really good stable businesses on retention—and I call them a lot of the canaries in the coal mine. Maybe the 2022 cohort of customers, it was like amazing because it was post-pandemic, but '23 and '24 and now '25 are not that great, but everybody remembers '22. And so this is, I think, a part where that data and that customer data, because we probably don't remember those customers from four or five years ago, or how good were they? Or, it's locked away in an ERP dataset. We can't necessarily...well, robots can get to all that now and help us make some decisions on that.
Chris Linnane: I have one more question, because this is going so well, as an optional question, but I think I'm going to go for it.
Jake Cook: Fire away.
Chris Linnane: But before I do that, you did bring up retention, and you could go on for hours about retention. I don't want to leave that yet. Tell me something about retention that you think people need to know right now as it relates to AI and marketing.
Jake Cook: So when you know your best customers, which you'll find, oftentimes, is they are very, very small in terms of a percentage of the overall customer count. A good friend, kind of a hero of mine, a professor at Wharton named Peter Fader, and his colleague, Dan McCarthy, at Maryland. We've got kind of a three-part series we're working on for HBR on this for a case. But basically, these clues of these tiny customers are so important to watch because two percent can drive 20 percent of your margin.
And we've seen this pop up when these companies get ready to be sold, and they'll do some audits, or we're helping kind of put together the due diligence for it, and you're like, "Okay, what are those two percent doing? Which campaign do they come from? Do they open our emails? Did they subscribe to SMS? Oh, what products are they buying?" Those are such valuable things to know because 20 percent, so if you lose a couple of tenths of a percent of a customer base, who cares?
Well, actually, that's going to ripple through and cost swings of three to five percent, maybe, on your EBITDA or your margin. So, these little tiny customer groups that are very, very valuable, I think keeping eyes on those and building the data around them, that's what's really going to set people apart as we move into kind of this agentic commerce and some of these new worlds we're moving into. The data around that gives you a nice moat, which I know we're going to talk about moving in a little bit.
Chris Linnane: Yeah, yeah, we're going to get there. That's great. Thanks for that context. I feel like a game show host right now, being like, "Hey, we've got time for one more question." By the way, who's your favorite game show host of all time?
Jake Cook: I think it's Bill Hader from Saturday Night Live.
Chris Linnane: That works.
Jake Cook: Yeah. Alex Trebek, because the way he was also really good in the real world, I thought.
Chris Linnane: Yeah. I think he would get mad sometimes, and you'd understand that he'd get mad because he's doing five shows a day, and there's got to be things that would drive you crazy. The one I like, look-wise, for anyone in the United States, you'd know this. And if you don't, I'll try to describe it. It's Bob Barker from The Price Is Right.
Jake Cook: Oh, definitely.
Chris Linnane: I don't know anything about him, but the look was amazing because he had that long, thin microphone.
Jake Cook: Super long, thin microphone.
Chris Linnane: Yeah. So his hand would basically be at his belt, and the microphone would be all the way up to his mouth. It was just a very, very strange, casual, weird presentation. All right, so I'll start our last one. Okay, this is going to be either human, AI, or hybrid: Deciding what the marketing team should focus on next month.
Jake Cook: I think it's going to be a hybrid.
Chris Linnane: Okay.
Jake Cook: I think the ideas, maybe the robots are surfacing some stuff, or it's helping, if people are dealing with cuts and staff, or cuts and budgets, a lot of us are going through some stuff, some cycles right now with the current economic environment, having an analyst, maybe you lost, now, you never had an analyst. Well, now you kind of have this robot that can kind of do some of that work for you, and then what's working. This is where I think it's really great. It's, "I got this strategic plan, it was presented to the board. It was approved on da, da, da, da, da, da. How do I translate that?" Oh, well, and marketing and finance typically don't get along. We've got kind of the analytical, the bean counters on one end, and you kind of have the dreamers, the creative people that are out chasing butterflies.
And those are kind of the stereotypes that exist on both sides. And I think this is where designing campaigns that are going to produce value like revenue, but margin, or good customers. I think this is where the AI has kind of let us build some bridges across the domains.
Chris Linnane: Interesting. So it's the glue between the two groups or the translator between the two groups.
Jake Cook: Yeah. Back to this idea of like, if you think of it as a way to kind of supplement maybe the weaknesses or your strengths and weaknesses, but maybe areas you're not super familiar with, like you can say, "Hey, I'm a CFO. What do I struggle with as entrusting my marketing team and approving capital? I'm a CEO." And that can kind of give you empathy for what their worlds are like. And if you have a deck or report or a strategic plan, you can put that in, and data governance issues aside, we'll put that aside for a second. There's a whole thing around how companies should handle that. But I think that lets you really unpack what you might've heard in a meeting or what you got in an email or in an all-hands. If you're an individual contributor or a leader, like specifically, what does this mean?
Here's some ideas, would this resonate with what the plan was, going back and forth? And then you can walk into a finance meeting and be booted up and say, "Well, if I understand correctly, we have to hit this EBITDA goal over the next three years. Is that correct?" And then I think the finance people are kind of like, "Wait, what?"
Chris Linnane: "Wow, you're speaking our language." Okay, great. So yeah. That's a great way to look at it. So, as AI starts influencing real marketing decisions, how do we know who's really actually responsible? How do we keep clarity about who's responsible for those decisions if AI is influencing them so heavily?
Jake Cook: Yes. I think there's a really interesting world that we're moving into. This gets in a little more like agents or agentic AI, you might've heard of, around how do we let these robots do things? And I think that the analogy from the physical world is, if you're going to hire somebody, you have a job description, right? You're responsible for these things. Your annual review or quarterly review is going to be coming down to these, hitting these goals or KPIs or whatever. I think, okay, from that, what does a human need to be able to do on that?
If you had an AI doing some of this, what would it look like? And so to give you an example, a lot of people would just throw AI at something kind of messy and big without breaking a sub-task down and say, so say we want it to generate a daily analysis of which campaigns are performing. Well, does that mean email campaigns or search campaigns, or okay, you wouldn't give this to an intern, right?
Chris Linnane: Sure.
Jake Cook: We'd say, "I want you to go out there and look at what performed well the last 24 hours on Instagram for this campaign group." Well, that's something that you can kind of test and get an agent to go out and check. And once you get that, send me a Slack or an email, and I want this at 8:00 a.m. every morning before I start my day. And then once you get, "Oh wow, okay, now give me commentary on that." And so you kind of take these small steps where you can kind of expand it out, where you're starting to get some good data or at least good outputs from these things, where you feel like you can trust, but starting really, really small.
Chris Linnane: Yeah. The temptation would be, "Go for the big thing," because it's all coming back in the same amount of time, roughly. I put this question out, I get a bunch of information, but biting off more than you can chew, you kind of lose track of what you're even doing in the first place.
Jake Cook: Yeah. Keith Rabois, who was a COO at Square and LinkedIn, he's got a great talk from 12 years ago from a class that Sam Altman taught at Stanford about how to start a startup. And he gave this great example on operations of how do you hire and promote and let people do things. It's a really simple example. We said, "I had this task where I would have them try and deliver smoothies to my engineers at 9:00 p.m. I didn't want to eat pizza because pizza, you get sleepy, it's late at night, but I want fresh smoothies, not melted at 9:00 p.m." And he goes, "And I would give this to interns to see who could solve it, and nobody could solve it."
Then they had somebody walk in and just boom, "Yeah, I can do that. " And he's like, "Yeah, right. I've seen all people try." It's a great little story, but I think it illustrates this idea of pick a task and then once someone proves they can do it, goes, "Once I saw this person be able to deliver that at 9:00 p.m., then I was like, okay, well now we're going to hand you more and more types of things." But the simplest task, that seems simple on the surface, has some complexity to it, but defining those boundary conditions on what the task and well done means, I think, is really important for AI and agents.
Chris Linnane: A lot of the conversations in AI and marketing are about speed, and we talked about this a little bit before around the edges, but if we think of it in that framework, what kind of opportunities are we missing?
Jake Cook: I think you can make terrific time and go nowhere. Back to what's timeless versus timely. I think doing a job that doesn't need to be done really efficiently is an easy thing to do. And as Charlie Munger would say, I think he calls this, "twaddling," kind of just wasting time and thinking you're doing something. So I think that idea of, if you think of AI as a way to kind of get efficiency, but like efficiency toward what? And AI can be a thought partner, it can help question what are biases, what should we be doing in this business, or within this brand? What are we trying to achieve? It can be a really good thought partner.
One thing I'm constantly shocked about working with either startups or an industry, sometimes you ask, "What are we really trying to do here?" Not in the next two weeks or even two quarters, but over the next two, three, four years, because to create value, it takes compounding. Good, consistent, resilient value creation just takes time. And what are we compounding, and what's that rate of compounding look like?
And I think that is, "Okay, we can grow seven percent this year," work with compound annual top-line growth rate. Cool. "We're okay with moving...what do you think about margin? Can we move margin at the top-line?" Oh, now we're getting a little bit difficult. Okay, well, what does this look like? Well, AI can speed a lot of that stuff up, but if we haven't defined some of those ways in which we define good work and if we're creating value, then I think, again, you can make terrific time and go nowhere.
Chris Linnane: Yeah. Well said. That's exactly what I think I get concerned with a lot, is we'll churn out a lot of stuff and be like, "Did any of the stuff need to be made in the first place?"
Jake Cook: I think that's a really good first principle of the question. Does it need to be made? Does it meet what the business is trying to achieve from its goals? Are those goals realistic? Because we're seeing this kind of, in my opinion, this...I saw this a lot with the internet, with people like, "Well, we just get this stuff online, and we're just going to print money." It's an ATM, right? It's like, "No, no, no, no." It's good old-fashioned economics. I mean, just putting something on the new distribution channel on the internet or more efficient ads doesn't necessarily mean you're going to build a great business, or it's going to blow up and grow.
Chris Linnane: If access to these models just becomes more and more available, people are all able to see the same things, what happens to differentiators? How do you stand out in that environment?
Jake Cook: Nobody really quite exactly knows what some of this will look like, but if you take this idea from machine learning and computer science, it takes lots of data and knowing it's a probability, it's going to kind of converge on what the most likely answer is. And so when we think about creating new, exciting things, and if we just completely rely on the robot without the human, then I think you have this risk of your brand, your idea that'll kind of converge to the means. So the old joke is, "AI won't replace humans, but humans with AI will." I think that's kind of what you're going to continue to see. And so, how do you get something unique and special? Well, what does this stuff live off of? Well, it lives off of data, right? I mean, really, and so if you're using a model that's connected to the same data everybody else has, what's your differentiator?
Well, we can kind of fine-tune the radio dials, if you will, or the base and trouble, if you will, on a large language model. So there's some of that, which you can kind of think about, but I think, and what we're working with a lot with our clients is, what's your data strategy? Are you building data that nobody else can get to? I.E., are you having data around your customers? This could be customer service logs, this could be user reviews. I mean, there's a plethora of data, and that's what's really fun about AI, is all that data used to be locked away in a database or an ERP system or in something that was hard to summarize and take action on. And now AI lets that kind of become new. So I think that's a really exciting area where you can start to have a proprietary data set around these customers.
And then if you slice that by our best customers, as we were talking, our best customers say this. They came in from these campaigns, they buy these products on their first visit. That I think gives you a way to continue to refine and stick out. And then you have some confidence that you can build creative and campaigns that aren't just going to be like everybody else.
Chris Linnane: So, keeping the human in the mix somehow, it's going to work. We're going to be able to keep humans in the mix in the future.
Jake Cook: I think we're not going to get away from the timeless principles of some of those things.