The AI Implementation Audit: What Section’s CEO Learned in 18 Months

25:24And then business models and structures and things like that kind of made it less advantage for people who like mental health. Like in the same way, when you have Toby's and people listening and myself and other like who kind of have a position where they can do those change, how much do you think they have a responsibility also to say, yes, I realize I'll be able to cut some headcounts, but I also kind of need to do this in a way that's kind of like, say, see, Maddie. I think they do, but I'm Canadian, which means in America, I'm accused of being a socialist. I'm from Denmark, so I hear you. Listen, there is going to be some very hard downstream impacts of all of this.

26:08Job loss is already coming. The data we're now seeing, this is what's different than from when we talked 18 months ago. We are now seeing some real job data, right? The Washington Post report about programmers, those who are not software engineers, right? The level below programmers in the United States,$97 ,000 a year average income. That class of technical worker has seen a significant contraction in open positions in the last two years. So we know what's going on there. Derek Thompson, The Atlantic reported this past week around the recent college grad sort of hiring disparity, I guess, or slowdown when compared to other ages in the knowledge workforce.

26:52Implication being entry-level jobs for college grads being impacted by AI. I think that's more of an assumption versus we actually have the data that proves it, but it seems to be indicating that. So listen, this is coming. And every time I suggest that companies have some obligation, because they're going to become more profitable, do they have some obligation to the employees that are impacted? You know, here in the U.S., I get blank stares. I get it. That's not the purpose of capitalism. It's not the purpose of enterprise. You know, purpose is to profit maximize, as you said. Purpose is to deliver shareholder returns that are outsized.

27:33And so reserving money or training programs or whatever it might be to support those or impact, it's not going to be in most companies' plans. And frankly, therefore, who else? The government. I just don't think we have currently governments that want to deal with this because they've got their own challenges in terms of budgets. And we want to pass tax cuts. So you want to pass tax cuts. You want to shrink government. How do you want to handle what's about to happen? The AI industry doesn't get it. We know that. So in terms of who's creating this technology and deploying it and pushing it, the AI industry, they're not going to sign up to handle any of the implications of this.

28:16They're signing up to make a shit ton of money and become the most powerful and influential companies in the world. That's all they care about. Sam Altman's not trying to cure cancer, right? And he certainly doesn't care about the workers that get laid off. So I'm not sure what the answer is. Obviously, you can tell. I'm worried about the implications. And as a minimum, I think our responsibility as CEOs is at least get everybody into the AI class, at least get them ready for what's about to happen. And then if they do leave your organization, either by choice or not by choice, they're at least capable to work in the age of AI.

28:51I want to say one more thing, Jeremy. I think I was wrong about the AI class, meaning I thought it would be, so those would be in it and those would be out of it and those in it had an advantage. We're all going to be in the AI class faster than I thought. So this idea, this meme, right, that went around for the last two years, you won't lose your job to AI, you lose your job to someone who uses AI. That was bullshit. We're losing our jobs to AI. Being in an AI class will not be enough. Being able to work with AI is going to be a mandatory requirement, I think, for all knowledge workers, but it won't distinguish or differentiate you or save you in any way.

29:32In terms of companies are going to deploy these technologies as fast as they can to expand operating margins and create business advantage. And frankly, the recession that we're in or about to be in, depending on what data you look at, caused by the current economic uncertainty, will be the best thing for the AI industry. Not so good for workers, but it will accelerate the deployment of AI. Every CFO I talked to, I spent last week in New York, as a minimum, backfilling has stopped in terms of hiring. New hires are on hold. You know, we're at the beginning of what is going to accelerate, which is the recession will induce CEOs to double down on AI, try to push through all these deployment challenges and drive operating efficiency.

30:21Do you see any kind of industries that goes through your classes more than others? Is there like any pattern in kind of who is grabbing on this? You say service industries, of course, have a lot to gain and to lose. Is that the people coming through or is it just whoever CEO kind of like newness? It's everybody. I would say that we are our client basis of a mix of small, those more ambitious, more aggressive organizations, less concerned about risk, less concerned around data leakage and those kinds of concerns or issues. And then big, kind of a barbell. We have very large organizations that are just very committed.

31:01And frankly, many have failed in phase one or their first deployment basically landed with a thud. They deployed this like software, CTO or the VP of IT, it was a technology initiative. They bought it. They paid for the licenses. They deployed it. And there was no training. There was no change management. There was no AI champion program. They weren't like Moderna and these other organizations. And obviously, you know, Bryce, they didn't have those kinds of people really driving adoption. And it's kind of failed. And so we often get called at that moment, which is we're looking at our dashboard and 5 % of employees are using this thing, but it's costing us$3 million a year.

31:42What do we do now? And you basically have to restart and do the hard work. It's not rocket science. The good news, if you're the leader, this is not rocket science. Making the AI is the rocket science part. We're not making the AI. This is change management. This is, you know, a lot of this we know how to do. It's just hard. What do you do there? Say you're the CEO. Say you've got a failed first deployment. You treated it like a tech initiative. There's no change management, no training, as you said. And you're at this pivotal moment. What's the playbook for the first three or four actions to restart on solid ground?

32:19Well, you need some goals. So just reset. First of all, reset, reset the plan. What level of adoption and weekly or daily active usage are you targeting in what period of time? So take a 12 and 24 month time horizon. If your organization is larger, over a thousand employees, if you're under a thousand employees, you should do all this in a year and you have no excuses. And you'll be held responsible if you don't do it this quickly. Because your organization will suffer either from a, from an operating margin perspective or from a competitive perspective. Over a thousand employees takes a little bit longer, but not, but not that long.

32:51That's the first thing. Second thing is, and you guys, you guys talk a lot about this. You need to have a point of view about AI. You need to have a manifesto, not rules. You know, AI governance, AI policies depress usage. We know that. Now that was the point of them. You know, if you, your general counsel and your AI, you know, technologists come up with the rules, you know, that, that was stupid because they came up with a bunch of rules that just scared people from using AI and they couldn't put data into the AI and like, you know, okay, so you got to reset that. Can we double click on resetting that?

33:26I mean, I realized just to, just to put a bookmark, we're in number two and there's more I know, but say more about that because I see so many organizations say we can't do anything until we have our policy in place. And so I just, I want you to double click on that point. Well, I'd say the first thing you need to know, or we all need to know, is there is very little risk if you're using enterprise AI. So if you are, or even the Teams-based AI, meaning ChatGPT for Teams or Cloud for Teams, those products in the last couple of years have been built to be enterprise or company safe. and certainly the enterprise versions that if you're buying that from open AI or Microsoft, you know, a lot of the concerns around data have been dealt with in the technology.

34:12Now, you might not trust big tech or, you know, you might not trust that they're not going to change your business models later and use your data in a way, you know, that you didn't necessarily think they would today. That's fine. That's a different issue. But this idea of your data is going to be leaked into the model or, you know, and I hear that still excuses. I think they're just excuses. All the time. All the time. I'm executive saying, oh, I'm not sure it's safe. And no, no, that's just bullshit. But so you've got a policy or the legal team has written something up, right? How do you shift from policy to manifesto as a leader?

34:43You have to say to your team, this is not cheating. You have to say that working with AI is working smarter. You have to say that we want you doing this. We have all these deeply held cultural beliefs about work. Don't cut corners. Hard work is a virtue, right? Don't cheat. And we've got to shift all that. Because right now, for a lot of people, it sounds like it's cheating using AI. It sounds like you are getting an unfair advantage. It sounds like it's not my work. And I need to be credited for my work because that's how I get promoted. And so on, right? So you just need as a leader to be really clear around you want people to be using AI.

35:26It will have some downstream impacts, which will include people's jobs will change. We're going to create some new jobs for sure. I'm not worried about that over time. There's all kinds of jobs that we don't do today that we're going to create and do with AI in the future. So there's going to be all kinds of new jobs created, but there's going to be some stuff in the transition that's a little more painful. So we just need to model that behavior and talk about it in terms of let's use AI, make the organization more efficient. Let's create some new products. Let's have fun with it. let's outsource the drudge work.

35:58It's kind of a cliche, but it's true. Like I can't tell you how many little use cases for AI that we see at my companies where, yeah, you're not going to do a press release about it, but the team or the individual that's figured that workflow out and has outsourced that eight hours of drudge work. Hero. Hero. And they're happy. And how much of the, well, you mentioned how many people using AI in an organization. I think for many companies increasingly that's kind of like not a solved problem but like there's something that is the next level then how many agents you have deployed or like how do you see if like the next level of goal setting yeah i'm smiling about agents because uh you know don't get me started um oh please by all means start by all means i there's a step in between i want to talk about a step in between then we'll talk about agents step three then yeah so the step in between is what we call workflow redesign.

36:54So I do think that there are two ways to deploy AI successfully in organizations of any size. And I would suggest everyone should do both, either at the same time or one after the other. The first is give everybody a full-featured, good AI. So not some crippled old model, but give them, you know, Gemini 2.5 Pro. Give them Copilot Pro, right? the upgraded paid version. Don't give them bad AI, give them good AI. That's the first thing. And train them, support them with change management, with an AI manifesto that says you encourage and reward and value people who use AI. I call that, so the bottoms up, more organic use of AI.

37:39And then aim for 50 % plus adoption levels within 12 months, at least 50 in an organization under a thousand employees. Adoption defined initially as weekly active usage and then daily active usage. That's part A or that's the first, you know. The second is workflow redesigns. Take a team of people, one or two in a small org, maybe more in a larger org, and go into teams, interview people, and decompose their workflows. Write them out. Like literally describe these workflows on paper, the inputs, the outputs, the dependencies, and the data required to make the workflow, to create the outcome or output, whatever it is, human only.

38:28Describe it, then redesign it. See which ones AI can help with. Prioritize that list. You'll have a hundred or hundreds of potential workflows that can be redesigned with AI. That's the good news. Hundreds of potential use cases. Overwhelming. That's another reason why employees struggle with AI. Back to your question, Jeremy, at the outset. They struggle in part because the software, because it's not software, is so performant. It does so many things, but nothing really perfectly. Again, it's not like software, but it does so much that people get confused. Well, where do I use it? Same thing with use cases.

39:07When you kind of look at your human workflows or human use cases, you'll get hundreds in a typical knowledge workforce organization. Then prioritize five or 10 or 15, pick a number, prioritize them, redesign them with AI, codify that redesign. How do you codify it? Well, you build a prompt library, build a one hour seminar and show everybody how to use AI to do that workflow in the new way. Build a GPT or in Microsoft, that's called Microsoft Studio, you know, build an off the shelf bot or an agent. So, yeah, I think we are moving to agents. I think they're overhyped. This is what the AI industry always does.

39:47While their existing product doesn't work very well and it hallucinates like crazy, they're like, oh, don't worry, agents are coming. You know, they're always doing that. It's what playbook tech is always used. Yeah. It can't quote a document that you give it, but don't worry. It will outsource your thing. Yeah. Exactly. No, and agents are coming. We'll solve for all this. And by the way - Before agents though, before agents on, you know, codifying and redesigning workflows. When we last spoke, I think OpenAI just earlier in the week or the previous week had launched the GPT store, which say what you will about the store, forget that.

40:20But the capability to codify workflows, what impact have you seen of that ability on adoption? Because I'll just tell you my expectation. My expectation was it's going to transform adoption. I've been surprised that it hasn't. I share that sentiment. I think some organizations have been able to really leverage GPTs. We have at Section, and you and I both know about Moderna, and they talk about hundreds of GPTs deployed. I interviewed Michael Dominic from UserTesting recently, and he has a lot of GPTs deployed internally at UserTesting. So I think some organizations have figured out they are a great way to codify AI-assisted workflows.

41:03But it's been a struggle. It's been slower, I think, for people to realize their power and then to deploy them. They're also in some ways hard to deploy, like who builds them, who manages them. If you think about an organization, again, for an individual, for us, for solopreneurs and startups, they're easier to kind of build and deploy. And if they break, it's not a big deal. You go back and fix it. When you think about GPTs. Even how do you create authority in them, right? You know, like we have five brand bots internally. Yeah, exactly. Things like that. It's kind of the Wild West, right? So I could imagine a lot of heads of technology or IT are like, no, we don't want those things running around.

41:37They're like these little bots and yeah, there's different versions. And then the person who made it left the company and now there's bots. People are using it and who owns it? So I think it was, I love the idea. I was going to build a thousand GPTs and stick them in the GPT store and own that real estate. I bought that hook, line, and sinker. I love that idea. Hey, this is like the Apple Store 2.0 and they're early winter. store are going to win. So I was going to build a thousand GPTs overnight. And I said to one of our board members, I'm thinking about doing this. He's like, are you out of your mind?

42:10Like, this is like a concept, this idea of a GPT app store, like, dude, like slow down, like wait a bit, like, like figure out, is this really going to stick? And then when you get more signals, you might want to jump in and, you know, put a thousand GPTs in the app store and see if you can turn that into either lead generation or revenue stream. Anyway, we never did. Obviously, We didn't build one GPT or build a few and stuck them in the app store. The rest we use internally and we love them, you know? So listen, agents are coming. The big reason agents are coming is the AI industry, open AI in particular, but all of them have realized a couple of things.

42:47One is these applications really are what's driving the value. The models aren't driving the value, right? Then the models have all really been commoditized. So what's really different from 18 months ago is the frontier models have all converged, basically. with the same set of capabilities. So it is, in fact, the application layer where the money is going to get made and where adoption really happens and the business value is generated. And if you think about ChatGPT, what it is is the first AI application to get to a billion users. That's what ChatGPT will be. And that has really opened everyone's eyes to, no, the way we win is you build applications and you drive adoption of applications and applications essentially, to some extent, over time, take the human out of the loop.

43:31So agents are the next generation of AI applications. The AI industry wants them because humans are getting in the way of deployment. This might save us, ironically, right? Our challenges with adoption, our challenges with using AI. That's fascinating. Not actually be something that we need. It's slowing AI down inside of our companies. These pesky humans are keeping us from deploying AI. Right. They're getting in the way. So, and that's not good. You're open AI because you've raised so much capital. You're burning whatever it is a minute, you know, in terms of compute and energy. So agents are a way to take the humans out of the loop and our way to deliver to enterprise, larger enterprise where adoption is really stalled out in terms of chatbot adoption or LLM adoption.

44:19Let's bypass the humans, take them out of the loop, generate the business value, extract that business value. If you're open AI in whatever pricing model you can come up with, it sticks. And, you know, we're all going to be happy, at least the CEOs will and the builders and providers of the agents. Well, you actually, you kind of, I feel like you foreshadowed agents a little bit, even on our call 18 months ago. I don't know if you remember this, but you talked about at Section, you're an educational organization. but you, and we can have shared some cynicism about people's desire to learn, right?

44:50And specifically you mentioned somebody doesn't want to learn how to create a strategy. They want a strategy. Correct. And you talked about shifting from a class that teaches folks strategy to, I don't know what words you use, copilot, whatever that actually just delivers the strategy. Talk about how your offering has changed based on infusing AI, not as the subject matter, but as the, not the what, but the how? Yeah. That's changed in an important way because we've launched Prof. AI two weeks ago. So Prof. AI is an AI powered coach. And the only thing it's teaching and may ever only teach is AI.

45:28At this point, Prof. AI doesn't teach other things or coach other things. It's coaching AI. So Prof. AI is powered by currently the anthropic models. We built it initially on GPT, on the GPT APIs. Now we're using Anthropics models instead. We like the output better. And it teaches people how to get good at using AI. And talk about disrupting ourselves. There's lots to talk about with this. The first thing I want to talk about is it's an amazing learning experience. It is because it's personalized almost from the first moment. So when you play with an AI-powered learning experience or an AI-powered coach or an AI-powered tutor, I think the first time you do it, it is one of those magical sort of technology experiences, like playing with ChatGPT was, you know, two years ago when we all first started doing that.

46:18Because right away, the coach or the tutor knows who you are. So the exercises and content inside of the experience are immediately customized for you because you tell it where you work, what job you're in, what country you're in, what level you're in, you know, in the organization. and right away Prof. AI starts to work with you and coach you knowing who you are. It really is amazing. There'll be no value in video-based learning in a year or two. I mean, Chegg is dead, the rest are going down. You know, Udemy, Coursera, Skillshare, Skillsoft, you name them, anybody who's made a living producing video-based learning content and then asking people to watch it inside of learning management systems, all those companies, I mean, you know, there won't be around.

47:10These experiences are magical. The AI powered ones. The contents generate on the fly. The content is personalized. The content quality is high. And it's not just content. It's actually giving you exercises. It's producing. In our case, Prof. AI produces prompts with you, works with you to produce the prompt. So you get the output. You don't get the, you know, you get the learning, but you get the output. but you want to learn how to prompt and you actually want to cut and paste that prompt and stick it into the AI, right? Well, that's kind of what I was wrestling with even as you're talking, because as at least, and maybe this is feedback or whatever, but when you start talking about it, it's still a teaching tool.

47:45And maybe that's what, maybe there's this want versus need, right? I think people want to be a lifelong learner, but people don't need education. They need the output. So is the package teaching because that's what people think they want, but really the content, so to speak, or the sugar inside of the wrapper is actually the output because that's what people really need? I think it's both. I think people don't want really to learn, frankly. They don't. There's very few lifelong learners out there. No, it's a mix of this is what we could deliver today. And it's presented more as a coach, I'd say.

48:21It's sold to our enterprise customers as a learning experience because that's how enterprise buys upskilling. upskilling. They buy it as a learning experience primarily. But you're right. What people really want is get me faster to prompting and working with my AI to get good outcomes from my AI. So ultimately, Prof. AI or other learning experiences should be living inside of the AI, almost, if that makes sense, like living inside of the chatbot. We can't do that. Prof. AI is a, you know, it's a separate experience, but you can certainly save your prompts from Prof. AI. You can cut and paste them over to GPT or Cloud, whatever LLM you use.

49:00And eventually we'd like to make that integration as tight as possible so that, yeah, you're working and learning at the same time, essentially. At the last conversation, you also mentioned, the first that I heard talk about taking your board decks and then basically synthetic roleplay it out before you had board meeting. Still doing it. Learning something new. What's the latest board management trick? Yeah, we did it last week. I'd say this is what's new, this is what's different. from 18 months ago. So Claude used to be top of the leaderboard for the last 18 months. You've done it now, I guess six or seven board meetings.

49:32So GPT-04 jumped to the top of the leaderboard. Not 03, just 04. Yeah, for some reason. Yeah, we used, yeah, and maybe 03 could have as well. I don't know. We used 04 this month. And yeah, Claude was close behind. Yeah, we still do. It's great. And on the, on the, you think maybe that's a little bit of a jump, but you had like, you're a monitor with the three different kind of like foundational models. utago 304 what's your go-to kind of mental model for should i use deep seek uh deep think should i use research should i use ideation when do i go to grok you know what do i what do i image generate with mid-journey where do i how do you what's the go-to navigation structure i get confused like everybody else i mean i mean if you had to do a case study on the worst go-to-market at strategies in the history of technology, it would be the AI companies, right?

50:28OpenAI and Microsoft would both win the worst award. Well, Google's right there. Google's right there. I mean, it's terrible. Yeah. In terms of product marketing, in terms of naming, in terms of conventions and so on. Yeah. I get confused. You know, I, when I think about bigger projects, I think about deep research. When I think about everything else, I think about GPT or, or call them perplexity for research. But frankly, I'm, I'm confused. I'm very frustrated. Sometimes you don't even know, right? You're like, okay, remind me, is 03 better than 40? Is that? I mean, they know it. It's absurd.

50:59You know, Sam Alden's talked about it. They need to bring it all under one kind of product. But I want to go back to Prof. AI for a sec. I want to talk about business model disruption. Please. Yeah, no, I love it. Love it. Here's what's crazy about Prof. AI. It takes about an hour of conversations with Prof. AI to become very proficient in AI. Because we're measuring you, right? You're doing exercises. We're grading you, essentially. And so at the time you get completed, you're sort of certified as AI capable. And then we auto-release a LinkedIn badge if you want one to your LinkedIn profile. It takes about an hour.

51:35So those conversations, the tokens, the AI cost, the inference cost of that, for me, is a dollar. One dollar. That's crazy. So let's put that in context. We charge section$750 a year for unlimited access to the AI Academy, to consumers. And we absolutely discount that. So our probably net selling price over the course of a year is probably closer to$550, something like that, right? $750 is the list price for unlimited access to the AI Academy, which includes all live sessions and then anything that's recorded and some other benefits, right? We have a Slack channel and we have other community and benefits and assets.

52:19So it's not just the courses, but the core value proposition is the AI Academy, the courses. So$750, discounted to$500. Prof. AI can do most of that. Not all of it, not yet, but it can do most. It can't do the community, but it can certainly do the upskilling, the learning, the skill acquisition in hours or less for the average student. and it cost me a dollar. How do you price it right now? Oh, I don't price it. So it's a combination. I'm not sure what to do. And frankly, I want enterprise lead. So here's what I'm doing right now. It's going to be free to consumers. So our members who are paying hundreds of dollars a year will get Prof.

53:03AI, obviously included in their membership. And I'm betting what they're really paying for is access to the live events, the live lectures and workshops and so on. If they just want the upskilling, if they just want the skill acquisition, then Prof. AI is going to be free for everybody. All consumers are going to be able to use Prof. AI at no cost. That's my launch strategy I'm launching in a couple of weeks. I launched the enterprise version a couple of weeks ago. The consumer version is going to go out as a freemium version. So I'm doing that because I can bear for now the inference cost of a dollar.

53:40If we had millions of users, I'd be like, oh, shit, I need some capital. You know, I got to fund the inference cost. But for now, it feels like the smart thing for us to do in terms of getting it out there. Frankly, helping others get ready. It's a mission, that section. As you said, Jeremy, to bring a million people into the AI class. I don't think that'll be enough to kind of save you, so to speak, in terms of your job if your job changes. But you need to be in the AI class as fast as possible. So we want PropAI in as many people's hands as possible. It's going to help them, we hope, upskill themselves.

54:12But yeah, listen, if I was an AI native startup, I'm not, right? I'm an incumbent with an old, and I'm a digital incumbent now with an AI product that costs a dollar to deliver a tremendous amount of value to the student or to the employer that might be deploying Prof. AI. So this is my point around disruption from a business model perspective. These are going to be very significant business model disruptions. And so you got to do it to yourself or at least be ready when someone does it to you. Well, how did you decide? How did you how did you as the CEO of an incumbent digital player, how did you decide?

54:49Because what I'm hearing you say, just to read it back to you, is we decided to disrupt ourselves or we decided to at least get in the self disruption game. Walk us through your thought process. Yeah, I think it was, I don't know, three inputs probably. One is, what's our mission? Why are we even doing this? I don't need to be doing this for a living. This is hard. Running a startup that's got ambitions and not enough capital and investors that are expecting a high return and all that stuff. It's not exactly easy running a startup. So like, why are we doing this? I don't need to be doing this.

55:23I'm doing this because I want to help as many people as possible get into AI class. That's our mission. How do I get to a million people? How do I help two million people? It wasn't going to be through what we were doing. It wasn't scaling fast enough. We had to create something with much less friction, much more accessibility and very high value. And that's an AI learning experience. That's the first thing. Second thing is I could not look myself in the mirror in two years and say, someone did it to me. Like I had the insight. It wasn't like I was clueless. if I was, but genuinely if I was blindsided, if I hadn't seen the future, you know, if I was in my own tunnel or whatever, you know, and not seeing the future, then okay, I got blindsided, that's on me.

56:01But I had the insight, someone was going to do this. So I could not look myself, my investors and my employees in the eye, you know, in a couple of years or a year and say, listen, you know, we didn't act and someone else did it to us. If someone else had built Prof. AI and we didn't, I mean, to your point, we were early in AI training, like 18, now two years ago. So give up all that advantage and have some, you know, two dudes and a dog and an AI in the garage to beat us with their own AI, AI tutor. Man, I, I, shame on me. I got just, I, that wasn't going to happen. Third thing is I have an amazing board.

56:37I have amazing investors and we're aligned. We're aligned that this creates risk and significant opportunity. And so, you know, it makes a lot of sense. Let's do it. The question of testosterone for my brain. One thing that I'm particularly interested right now in Greg is the tendency of older folks. We actually talked with Evan Ratliff. He's a podcaster, amazing guy. He's got this great new podcast. And he referenced the fact that it's a consistent theme throughout the history of humanity and technological process to resist change. I think AI is particularly well-suited to experienced individuals because they have more context.

57:19It's like in an MBA class, right? The experienced person gets a lot more out of the class than the person straight out of undergrad, right? The same is true in collaboration with AI. More experienced individuals actually have far more context to glean better outputs from it. And yet the tendency is to opt out, check out, say, or resist, right? In your comments about AI being testosterone for your brain, specifically in regards to call it cognitive decline, that's been on my mind. Have you learned anything or seen or observed or felt differently about that as a kind of a concept or paradigm in the last 18 months?

57:52Here's where I've ended up on how we think about using AI. I think there'll be two types of AI users. This is what's of most concern to me right now, which is that we'll all be cognitively offloading because we always have. We've always looked for these edges. You know, we used to be able to recite the Bible or the Koran from memory. Humans used to be able to recite Homer's Iliad from memory, 15 ,000 lines of text from memory. And then books came along, and so we cognitively offloaded the books, right, and used that cognitive capability for something else. GPS is the obvious example today, right?

58:30We used to know how to navigate. That part of our brain has actually been atrophying kind of that spatial sort of mapping capability because we rely on GPS. We've cognitively offloaded that part of our brain to GPS on our smartphone. So this is the mother of all cognitive offloading tool. This is like, we've never had this before. So we are going to use it and we should. We should all cognitively offload as much as possible, borrow those 30 points of IQ at any age and hopefully get the gain. But I think what's going to happen is there'll be two types of people who cognitively offload, freeloaders and managers.

59:08And I think most of us, sadly, are going to be AI freeloaders. We are going to be lazy. We're going to cut and paste from AI. We're going to take AI's work and call it our own. We're not going to add what's unique and different and special about us and our experience and our judgment to the AI's work. We're just going to use it to cut the corner, get home, put the kids to bed, walk the dog and watch Netflix. So the poster won't be remember to use AI, it'll be remember to use your brain? Absolutely. I think we're going to have to develop new habits that turn off the AI or work with AI and then improve that work and acknowledge and show the difference to yourself and to your boss.

59:54I mean, I'm not sure how we're going to do this, but we are going to have to not lose our minds. I want to be really clear about this. Big tech wants us to lose our minds. They want us to be 100 % reliant on these technologies and stop thinking. It will make them the most influential and powerful and valuable companies and individuals in the world. So if you had one human skill that you think that we ought to train ourselves in, keep training, like we stopped having to run after animals in the forest. And so now we go to a gym. What's the kind of like the innate human skill outside, like the broad of thinking that you would go to the mental gym for?

1:00:45What's right or wrong? I'll take that. We must never lose that judgment. We must not offload that question, especially when the AIs are run by companies that don't care. And how do you think we, I don't know, but you run a company that teaches people. How do you go through the process of understanding how to codify your own right and wrongness so that you can train it? Like, how do you do that? You need to write it down. I think we need to begin to describe ourselves to ourselves better. We need to stop, reflect on ourselves. So when we drift, we know that we drifted. What are our values? What are our operating principles?

1:01:46How do we make decisions? Like codify those, meaning write them down. understand them. Make sure you align your AI to those. This idea that 200 million young men are going to be using XAI, that's not comforting to me because XAI's positioning in the market is it's got no guardrails. Yeah. Right? I mean, they turned it off, but they used to have a feature called unhinged. Like this idea that, you know, that someone thinks it's a good idea to release an AI with a feature called unhinged. I like that as a, I like that as a frame is don't be unhinged. It's our, don't lose your mind. I mean, there's something there that's actually really, really profound.

1:02:32Yeah. We cannot lose our mind. And it requires that we actually connect to our own humanity as a protection against our, our worst impulses. Absolutely. Yeah. This is awesome. Greg, you're awesome. Thank you so much for coming back on. I hope we can invite you back on maybe before 18 months so that we can hear what's going on. We got to hear a post-mortem on the Prof. AI launch and hear how it went. Absolutely. I know. We are looking at the data as we speak. It's fascinating. So, Jeremy, it's so good to have him back on, huh? I mean, so much has changed, and yet so much remains the same. I thought that was a really, really fascinating insight.

1:03:14Do you want to go first? Kind of what we took away from this conversation? I mean, I always challenge myself to take notes. Sometimes folks, just so you know, Henrik tells me, I can hear your pen in the background. But it's because I know that this like passive vessel can only hold so much and I'm trying to capture it. It's the sound of wisdom. It's the sound of wisdom being transcribed into stone tablets. Anyway, I took a ton of notes. there's almost so much my thing that stood out to me. There's so many times that I always know if it's a good interview because I pull out the highlight function, not just the pen function on my remarkable tablet.

1:03:50And I'm highlighting a ton of stuff here. I mean, it's so much that it's pages and pages. That's actually one piece of feedback I would have to remarkable. It's really hard to scroll through your notes. But I think one thing I would say, if I had to highlight one thing as I'm scrolling frantically in the interest of time, the encouragement to leaders to shift from policy to manifesto. And what he said, three things specifically, it's not cheating. It's working smarter. We want you to do it. I think that is a message that is so sorely lacking. There's always a but, but be safe, but don't. And it's the but that is biting people's Achilles heel and keeping them from exploring and using.

1:04:36And I know a lot of leaders who say, yeah, we tell people to use it. Just do it in a safe way. And it's the ambiguous fear of an amorphous risk that's actually holding people back. So that's probably the biggest thing that resonated with you. What about you? You know what? I have thought a lot about a thing that he obviously framed much better. And so he put a kind of very distinct kind of like word on it, which is output outcome pricing. um what is it that you're gonna as an entrepreneur be making in the future and i don't think that you can kind of like deliver content you can't just deliver text and you can't just deliver hours you have to deliver here's this thing you ask for the strategy or here's this outcome you ask for 20 better conversion on your funnel whatever and so thinking in that way from day one also even as a human what is my output and how do i become better of doing that's interesting and that kind of goes to the second one which is you know what i've meet a lot of people and you show them replet or lovable and you show them what agents can do and stuff like that and they have a very difficult time kind of converting the awareness that those things are there to something that they can then change in their work life and so i've kind of always put it into people are just not used to thinking products you know people are used to thinking managing their bars or you know projects but not products yeah and so i think what he was talking about was really how you learn how to think of use cases how you really just take the time to write down your workflow and then you atomize them and then you figure out should these be redesigned or any of these components are they able to be be done with ai i thought was an interesting way of doing but the whole thing about like how do we become better of of productizing instead of solving problems uh i thought was fascinating um i do think this kind of he's so straightforward and honest right you know and i think this idea that you won't get out competed by somebody you know a person with ai we will all have to learn ai and there will be people who lose their job and they might find a new job but like There is just, I don't think there's really any excuse anymore in most industry to not learn it if you want to have a career in the long term.

1:07:05And so I think he talked - Not even an edge, you're saying. Not even an edge, just a career. I just, that's what I think he was pointing out. And, you know, in many ways, I hope he's wrong, but I think he's probably right. That wild stat of the day, they just surveyed 5 ,000 knowledge workers. Employee proficiency has not improved in the last one year. That's staggering, given that I would say the last year has seen the most attention towards proficiency of human history. But it also sounds like it might be the good news for everyone, because that's the thing that's going to slow everything down.

1:07:42That and what Mark Zuckerberg said in that YouTube video you sent me, which basically that it also just take a long time to build hardware, like to build like data centers. And so the two things I was slowing everything down is not the models. It's basically humans ability to change and then actually building stuff in physical space. Yeah, that's great. All right, folks, if you enjoyed this episode, if you enjoyed this conversation, please use the safe word. The safe word is right or wrong. Right or wrong. Right or wrong? Question mark or no question mark? No question mark. Ah, I see what you did there.

1:08:24Until next time, thanks for listening.