How to make AI worth your time with Max Mullen

If you're like me, moments like this can feel exhausting. Your bosses are telling you to play around with AI, and it makes you want to punch them in the face. Someone told you to spend your Saturday setting up a new AI tool, and it makes you want to throw your computer across the room. I get it. Despite working in the tech industry, I am actually a late adopter. I did not run out and buy the first iPhone. I don't pick up every piece of software that Twitter or the venture capital ecosystem is hyping. I have this strongly held belief that great technology should be easy to use. Great products don't need a user manual. I often say that great technology should come to you. You shouldn't have to go to it. And what I mean by that is simple. Great products make it really obvious why they're going to make your life better. I've been around long enough to see a few real revolutions in tech, things like the internet, the smartphone, social media. But I have also lived through plenty of hype cycles that turned out to mostly be noise. And to be honest, when ChatGPT launched and suddenly every conversation in Silicon Valley turned into an AI conversation, I was skeptical. When venture capital websites rebranded overnight and every startup pitch had AI in the title, I was honestly a little jaded. There were two reasons. The first is that AI and automation research have been around for decades. Chatbots have been hyped before. Responsive AI has been a thing since Siri came out in 2011. And big tech has been automating all sorts of things from resume screening to user support for what feels like forever. So I genuinely did not understand what made this moment different. The other thing is that we had just come off a very loud, very expensive hype cycle around Web3 and crypto that just did not live up to its promise for most people. But then something shifted for me. I kept playing around with like basic tools that anyone can access, ChatGPT, Gemini, Claude. And at first it felt like fancy Google. It was helpful. It was interesting, but it did not feel revolutionary. And then one day I tried something different. I uploaded a bunch of documents that I'd written in the past, blog posts and essays and notes and emails. And I asked it to help me organize an idea for a new piece of writing. And suddenly I had this robot collaborator. It produced a really solid first draft of a blog post, something that might have taken me hours. It did in minutes, something that I had previously experimented with paying ghostwriters to do. This robot did for 20 bucks. In one moment, it had done three things. It made my life massively easier. It saved me a bunch of time and it replaced someone's job. That was the moment that everything changed for me. I couldn't help but become a believer, not because the writing was perfect and not because I want AI to replace everyone's jobs. But because I could suddenly see the shape of what this technology might become. Slowly, over time, I've come to the conclusion that AI is going to be as big and disruptive as the introduction of the internet. But I also think that we are very, very, very, very early in this cycle. And I think most people that are talking about AI don't know what the fuck they're talking about. So now I am on a quest. I want to understand what's real and what's hype. I don't want the VC hype machine. I don't want the scare tactics that say everyone is ahead and you're behind. I actually want the use cases. I want the real tangible changes that make a difference in my life. The stuff that will help me understand what this technology is actually going to do to our work and our lives. So I'm going to do what I do when I want to learn something. Every now and then, I'm going to spend an episode asking some very smart people some very dumb questions. And I'm going to make them explain this technology to me. And either they're going to convince me that it is worth spending my Saturdays learning this stuff or I'm going to tell you that maybe it's too early for the rest of us. My goal is simple. I want to understand what the average person, not the engineer, not the venture capitalist, not the hyper-technical early adopter, what you and me should actually be doing right now. What's real, what's useful, and what's actually worth our Saturdays. I'm Molly Graham and this is WorkLife, where we untangle the messy human side of work. Today we're going to talk to Max Mullen. Max is the co-founder of Instacart. He led product in the early days, building software that coordinated millions of real-world actions. Groceries, shoppers, deliveries, all moving through one system. He's also someone I trust enough to ask all the dumb questions. These days, he invests in very early-stage companies, often just ideas, which means he sees a lot of new technology before most of us do. And yes, technically, that makes him a venture capitalist, which might sound like exactly the kind of AI hype machine I just told you I'm trying to avoid. But the reason I like talking to Max is that he is not breathless about technology. He's a builder. He's practical. He's spent most of his career turning messy, real-world problems into software that actually works. And he's been using AI tools inside his work and his life long before most of us started playing with ChatGPT. Max is firmly in the camp that AI is going to reshape almost everything about how we work. So today, I want to push him on it. What does he actually see happening? Max Mullen, welcome to WorkLife. Thank you for having me, Molly. I'm glad you're here. I've been looking forward to this conversation. Me too. So I am curious to just start with, like, Max, do you consider yourself an early adopter? I think so, yes. I mean, I do love trying products early and often. I mean, I early adopted Facebook when it first came out on college campuses. And then I tried it over many years. And I think through looking at the way it evolved as a product, I think I understood something about that company and about how people use social media. So when something new comes out, you're like, let's try it out. Yeah, I'm excited to try it. And as an investor, I want to kind of know what's going on with people and what interesting things they're trying out there in the world. Like, not just early adopting each thing once, but I try to early adopt whole categories of things at a time and sort of say, okay, here's how four or five different people are thinking about the same thing. What did it teach you to be so early on Facebook and then watch it evolve? There's just something different about playing with a product once and then playing with it again in six months and then again in six months and seeing how it evolves, right? And like great products evolve pretty rapidly. It almost sounds like you like being like part of the story, watching a product grow and change. I think if you want to know what the frontier of technology is going to do next, you kind of have to be in today's, you know, in today's moment, trying the riskiest, most interesting, most fringe things that you can. And those things often look a little bit weird and contrarian today. Okay, so I have a question for you because obviously today I want to talk about AI and this moment that we're in that has, you know, a lot of hype and a lot of conversation around it. Part of what happened for me with AI was there was this just massive hype cycle around crypto and Web3. And I actually like, it was so loud that I like tried to buy in at some point. Like I really tried and then I feel like it, in my opinion, didn't really deliver on the sort of like hype, though I'd love to hear your opinion. But then like AI followed so quickly after that that it actually made me jaded. Will you just like tell me a little bit about, like your experience of like those two hype cycles? Yeah, and what's really interesting is there's a bit of an overlap between some of the people that were involved in each of those cycles and sort of pull me at once. I also didn't get too involved in crypto and Web3. And I look at any technology or any company through this lens of like, well, what is the problem that it really solves in a consumer's life if it's a consumer technology? And I could never find the problem that Web3 or crypto was gonna solve for me. I don't mind my bank. I didn't have a problem sending money and I don't send money internationally. So it wasn't really a big problem that, you know, that the crypto solved for me in terms of the money transfer. And I thought blockchain as a technology was very interesting, but it was also so early. And then I also just saw a lot of bad behavior and I said, I don't really want to be involved with that. And I don't think I fully understand it. And there are a lot of people that specialize in crypto. And so that'll just be an area that I just don't invest in, was how I thought about it. Interesting. So you actually like tried it, played around with it and then like opted out? Yeah, I followed Bitcoin very closely and still do. And, you know, Coinbase was in the same Y Combinator class as Instacart. And I thought it was fascinating, but it just didn't, it never for me solved a real problem. And meanwhile, there were other more interesting technologies that I think were solving real problems. Yeah. Well, talk a little bit about with AI, was there a moment that made it real for you where you personally realized that it was different than crypto? And what was that moment? Yeah. I mean, the first use case for me was writing. I could ask A little bit of elbow grease to validate that this AI had magically developed kind of what a travel agent might have spent a week doing, but it turns out that it was a very good itinerary, and it is now the itinerary that we will be traveling on. Interestingly, though, at that point, I said, well, now I need to make a bunch of hotel bookings, and I need to find out some things that are not available on the websites of the hotels, so I need to call Japan. And so I ended up actually hiring a travel agent to book the hotels for me. And so this is this line, right? What can you do with AI, and where does it stop and a human has to get involved? And I think a lot of people kind of assume that AI can either do nothing or everything, but the reality is it's somewhere in the 80 or 90 percentile range, and you eventually or someone eventually has to get involved and do the last mile, at least today. You're kind of talking about one of my challenges with it, and I think a lot of people, where it's like, you start to ask it to do something, but you don't entirely know whether to trust the information. You know, how do you handle this new relationship with a computer where you're like, I know you can do these things, but are you actually good at them? Well, AI used to hallucinate a lot a couple of years ago. What does that mean? When people say hallucinate, what does that mean? It means it makes stuff up. It will confidently state facts that it has no idea if they're true or not. And this is obviously a bug and not a feature. And so the AI labs went about doing what they call reinforcement learning. They went and gave it a bunch of examples of it doing this, and it gave it a bunch of examples of better responses, and they sort of worked this problem out of the system. And so in most cases, for most of the normal things you'd ask an AI to do, it rarely hallucinates today. And I think this is an interesting point, Molly, because people may have tried AI some amount of time ago, and it didn't work for their use case, or it did something egregious, something that you'd fire an employee for if they did it and lied to you. But they haven't tried it for that same thing again recently. And I have this thing I call the six-week rule, which is that if you tried to do something with AI more than six weeks ago and it didn't work, forget about that. You have to try it again because every day, every month, the models are getting better, and features and apps that you're using AI through are getting better. And the thing that you couldn't do six weeks ago, I'm pretty sure is probably possible today. And if it doesn't work today, you should try again in six weeks. And AI is moving so quickly, we might have to reduce this to like the six-day rule. Like, again, things are happening so rapidly. In a week or two, some of the things we'll say on this podcast will be wrong because the new features have come out and new things are possible. I love that rule, first of all. I think it's such a good, I guess, like an indicator of how fast this technology is growing and changing and getting better. But it's also kind of exhausting, like as a normal person, to be like, I have to keep trying. Do you know what I mean? Like, I feel like a lot of people tried, to your point, like a lot of people tried it a year ago and then like wrote it off because it couldn't do what they wanted it to do. Why should I keep trying? Well, I think there's a couple of reasons. I mean, it can be really useful and save you a bunch of time in your personal life or in your job. I mean, it can do all of the mundane things that you're doing that take up time. It can do them usually instantly. You could send it an email that you got from your school and say, you know, help me put these things on my calendar. You could have it write a performance review of somebody that's really thoughtful and kind from just a few bullet points. You can really save time by using AI. And so that's one reason to use it. And the other thing that I'd give you as a reason why I think the average person can really get up to speed quickly is that all of us are learning on the fly. You and I are maybe one or two days of playing with AI away from being experts. And the half-life of that expertise is very short because again, there'll be new paradigms and new models that are coming out every month. And so it is a little bit exhausting to keep up, but there's very little work that you have to do to become an expert. Like the best experts at using AI, they learned a month or two ago. So nobody's too far ahead of you. That is really interesting to me because I also think that, you know, you talked about sort of like the history of computing and like that obviously didn't used to be true, right? To become, like to get a computer to do something complicated, you used to have to study for four years and get a CS degree. I mean, part of what you're saying is that both the ability to talk to a computer is more accessible and to get it to do things you want, but also that the expertise is more accessible. Yeah. And one of the things that has changed in the last couple of years is that some people, people like me or even people who are less of an early adopter, have used ChatGPT or their preferred AI assistant enough times that it's collected this memory of what you need help with and who you are and what you do for a living. And now it has this context, right? And when you ask it a question, you don't no longer need to brief it on all the details. You just ask it a very simple question and it gives you a much better answer because it knows you. And so I would, you know, I would encourage most people to pick one of these platforms and just stick with it. And then it'll get to know you over time and that'll make, you know, the process of prompting and getting a great answer much easier. It's interesting because like when you, I know that like, I'm thinking of my mom here. Hi, mom. Like you say, I give it my data and it remembers things about me and it gets to know me. And she says, that sounds scary and I don't want it to know. I don't want a computer to have all this information about me. Like, I'm curious, like, do you have a reaction to that kind of hesitancy? I totally understand that. And I started off skeptical and hesitant as well, but I started to build a little bit of trust with, particularly with ChatGPT. And I'll give you, you know, a personal use case. I, I got a bunch of labs back from my doctor. And there's, you know, lots of numbers in different ranges and different units that they measure things in. And I wanted it interpreted. And so I uploaded this to ChatGPT. This was a moment of trust. I said, you know, well, here we go. You know, it's now going to know like my blood work. And I, and I asked it to just give me advice as though it was a medical professional. And it was able to kind of clearly break things down. And, you know, it didn't obviously prescribe me any medications or tell me I was healthy or not healthy, but it gave me the questions that I should ask my doctor. And so then it made the, you know, phone call I had with my doctor much more productive so that we were sort of both looking at the same, at the same sort of hypothesis of what was going on. You know what, pause there for a second. Let's just list, like, if you, if somebody was like hesitant or reluctant, what are like the three places you would be like, start by trying this? Like, like you said the six week rule. So it's like, you kind of have to pick one or two places to start. What are like a couple of the places that you like commonly recommend people starting? Probably not your health data. I think you could, I think you could start with something really simple. Like, here's an email that I wrote. How could I have stated this in a clearer way? Or here's a performance review that I just got. What are three things I could do tomorrow to become a better employee and get a better performance rating next time? And then I would even throw in one more entertaining one, which is, it's really fun to use AI with your kids. So I would always start by saying, I'm with my eight year old son. He has a science question. And then I'll let him ask the question. And then we're going to get an answer that's appropriate for an eight year old. And then if we turn around and say, okay, now I want you to give me the adult answer. It will go ahead and do that. So it's very flexible, very personalized. And you just have to ask it for what you want and it'll, it'll generally be able to do it. So he talks to the robots? Sometimes when we let him, you know, and we haven't even gotten into the idea of vibe coding, Molly, but this idea that you can, you can talk to a computer and it can build you a whole app customized for you. And so my eight year old who again, can't spell, but basically just has an iPad. He can, using the dictation feature on the iPad, he can dictate what he wants his website to look like. And then an AI tool built him an entire website. And I essentially just sat there and showed to wield this new technology. And then secondly, you start to get to know its flaws and you realize, you know, this isn't superhuman, right? It makes mistakes. It really requires human oversight. And I think that's the way it's going to be for some time. And you really have to think about different tasks as well, right? There's things we do in our jobs that are mundane and repetitive, and we want to automate those things, whether that's using a spreadsheet instead of a calculator or whether that's using an AI agent instead of a spreadsheet. You don't want to do that boring work, right? You want to do the strategic work, the human work, the thinking work, and you kind of want to outsource the other work. And so I think both can be true. We can embrace AI. It can help us a lot. It can give us lots of leverage. And you could still imagine a company having tons and tons of productive employees that are now even more productive as a result of embracing this technology. So I don't subscribe to this idea that AI is just going to, you know, ruin every job. I don't think that's going to happen. Yeah. Yeah. OK, Max, last question. I'm curious, and this may date us or whatever, but what do you think is never going to change? I think that there's some things we love to consume from other humans because they're human. Like, I want to buy pottery that's handmade, and I want to meet the person who made it. I want to understand the story behind an artist and why they painted a painting. These are just not things that, even though robots can do them better or faster or cheaper, they're just not things I want to be done by robots. And so I think art and certain crafts, you know, carpentry, these are things that are bespoke and human in their nature. That's one category of things that I don't think will change very much. And then I think there's other things, like where you really just want a human involved. Like, I want to be coached by a human. I want to be coached by an empathetic person that understands me, not really by an AI. I want, you know, a personal trainer that's able to look at my form and spot me on heavier weights and, you know, and really understand me and meet me and hold me accountable. AI and robots are just not really going to do that anytime soon. So I think there's a bunch of interesting ways where being human is actually more valuable than being faster or better or being computerized. I love it. Max, thank you so much for this. I learned a ton today, as I always do when I talk to you. Thanks for having me, Molly. Okay, that was such a fun conversation with Max. I'm glad you got to meet him. And I thought there was a few really great takeaways from that. And, you know, I do just want to say again, like all this talk around AI can be tiring. But I think, you know, part of what I heard Max say is, number one, this technology is getting better so fast. It is important to keep giving it a try. So his six-week rule of just like every six weeks, it's like you're playing with a different product. Like it's like almost like they launched a whole new thing to play with. I think if you treat it as if it is changing that fast, it means that if you tried it and it failed at that, it's worth trying again. And I mean, the most powerful point that he made, I think, is that this expertise is just not that far off. That the thing that differentiates the people that are experts at AI from the people that are new and beginners is actually just the playing around. The people that are willing to give, you know, these tools multiple chances that are playing with it every six weeks, that are treating it as a relationship, as a conversation around what can you do for me today? What can you do for me this month that you couldn't do last month? And that idea of like, you could be an expert if you just made some time to play around with it is really cool. I've never actually heard anyone say it quite that way before. And it really does actually resonate with my own experience. Like I was such a skeptic for so long. I think I was so burnt out from previous hype cycles. But it pretty quickly, once I found a use case that I could get, like that just clicked for me, then I found the next use case and the next use case. So it kind of like became easier to imagine what it could do for me. But also, I found myself teaching other people who were sort of more stuck in the, why would I do this? So anyway, I loved, I really loved that point that like, actually becoming an expert is a few hours away or a few weeks away from you just playing around, you know? And the last thing I'll just say that I really took away from this conversation is, you know, I think a lot of folks, friends of mine, relatives of mine, find this whole thing overwhelming and find it exhausting and are sort of like, why? And is that really worth my time? And I heard Max say, for sure, it's worth your time. Like, I also heard him say though, I mean, sure, you could spend your Saturday on it, but actually it's worth your Monday. Do you know what I mean? Like this is a critical part of our jobs. Whether your employer is pushing it on you or not, it actually should be worth hours of your week, not your weekend. This isn't a hobby. This is a part of like our lives. And that it can be a really invaluable companion, you know, that it can unlock things for you that previously were inaccessible. And that's, that's really cool. That's exciting to imagine that we can create things that we previously couldn't have done and to imagine all that creativity out in the world that maybe we'll get to see an experience. That's exciting to me. I know it's also scary and hard, but there's exciting things. Everything has two sides, right? It has the scary and hard side, but it also has the possibilities. WorkLife is a production of TED and Pushkin Industries. This episode was produced by Isaac Carter and Leah Rose. Mixing by Hansdale Shi. TED's executive producer is Daniela Bolognese. Constanza Gallardo is the executive producer for Pushkin. Special thanks to Roxanne Highlash, Valentina Bohanini, Lainey Lott, Tonsika Sungmanivong, and Ashley Murphy. If you like the show and want more, come join the discussion on my Substack Lessons. I'm Molly Graham. Thanks for listening.