You don't actually want an AI employee

00:00:00,320 --> 00:00:01,360 [Eric] I have a great question 00:00:02,500 --> 00:00:08,520 [Eric] to start this episode out, and that is, do-- would you consider any of the agents that you run 00:00:09,880 --> 00:00:10,720 [Eric] AI employees? 00:00:13,160 --> 00:00:16,379 [John] I don't refer to them that way. I don't, I don't think. 00:00:16,800 --> 00:00:16,880 [Eric] Okay. 00:00:16,880 --> 00:00:19,940 [John] I try not to. [both laughing] 00:00:19,940 --> 00:00:21,360 [Eric] Try not to. 00:00:21,360 --> 00:00:26,840 [John] But, but potentially worse, we, we do have names for them. Um- 00:00:26,900 --> 00:00:28,100 [Eric] Oh, you actually... Yeah. 00:00:28,100 --> 00:00:35,860 [John] Yeah. But here, we'll, we'll get right into this. We have names for them, and they are all fictional TV characters. 00:00:36,520 --> 00:00:36,740 [Eric] Oh, interesting. 00:00:36,740 --> 00:00:38,000 [John] From shows people would recognize. 00:00:38,040 --> 00:00:39,700 [Eric] Okay, give me... Can you give me a couple examples? 00:00:39,720 --> 00:00:43,660 [John] Yeah, yeah. Um, so far we've kind of worked through "Parks and Rec." 00:00:43,980 --> 00:00:44,160 [Eric] Okay. 00:00:44,160 --> 00:00:50,180 [John] "The Office," and then a little more subscure, we've got... Obscure. We have some from "Good Place." 00:00:50,280 --> 00:00:51,520 [Eric] How many agents are you running? 00:00:51,520 --> 00:00:54,820 [John] [chuckles] We haven't worked through, like, every character in any show. 00:00:54,820 --> 00:00:56,240 [Eric] Oh, oh, okay. You're just mixing it up. 00:00:56,340 --> 00:00:56,820 [John] Yeah. 00:00:56,880 --> 00:00:58,920 [Eric] I was like, "Wow, that's a, that's quite a fleet." 00:00:58,980 --> 00:00:59,880 [John] We have like 10 to 15. 00:01:00,480 --> 00:01:04,980 [Eric] Okay, yeah. What's your favorite agent name based on the name, 00:01:06,020 --> 00:01:08,760 [Eric] the character it represents, and then the job that you have it do? 00:01:08,840 --> 00:01:15,280 [John] Yeah. For sure, um... Oh, I missed a TV show. "Brooklyn Nine-Nine" also, honorable mention. 00:01:15,300 --> 00:01:15,800 [Eric] Okay, there you go. 00:01:15,880 --> 00:01:17,780 [John] We have an employee that's a big fan of that show. 00:01:18,420 --> 00:01:18,500 [Eric] Mm-hmm. 00:01:18,500 --> 00:01:25,660 [John] So she's on a client, and I was like, "You know what? What do you wanna do?" She's like, "All right, 'Brooklyn Nine-Nine.'" And we went with Captain Holt. She's the- 00:01:25,660 --> 00:01:26,540 [Eric] Okay, yeah 00:01:26,540 --> 00:01:28,320 [John] ... captain of the squad, you know? 00:01:28,320 --> 00:01:29,380 [Eric] This is, this is fun. 00:01:29,380 --> 00:01:29,520 [John] Yeah. 00:01:29,540 --> 00:01:30,320 [Eric] This is great. 00:01:30,320 --> 00:01:46,740 [John] [chuckles] So, and like, at first it was like, "We'll just, like, do this internally, like, for fun." And then it just kinda escalated, and now, like, the client is keeping the persona and we have, like, an image [chuckling] there. Um, we gotta be careful with copyright laws here, but, you know, uh- 00:01:46,780 --> 00:01:47,480 [Eric] Yeah. 00:01:47,480 --> 00:01:48,980 [John] But we just, like, kept it 'cause I- 00:01:49,040 --> 00:01:49,360 [Eric] Yeah. 00:01:49,360 --> 00:01:52,140 [John] ... 'cause I just... The original plan was like, "Let's just keep it internal." 00:01:52,340 --> 00:01:52,740 [Eric] Yeah. 00:01:52,740 --> 00:01:58,180 [John] And then we'll... And then it'll be some boring, like, blah, blah, blah data agent, you know, type thing. 00:01:58,400 --> 00:02:02,460 [Eric] Sure, sure. That's great. The, the most useful agent 00:02:03,740 --> 00:02:09,380 [Eric] that I've ever built, that actually someone this week, it, it went down because it, um, 00:02:10,600 --> 00:02:14,680 [Eric] it hit a budget cap because I adjusted it to run a check way more often. 00:02:14,780 --> 00:02:15,520 [John] That's, that's happened. That's happened to me too. 00:02:15,600 --> 00:02:21,560 [Eric] And it ended up, like, looping, like, a bunch because there was an error, and so it looped and hit a budget cap. 00:02:22,620 --> 00:02:28,100 [Eric] And so it quit working, and people actually complained. I don't even know if they knew it was an agent, but people were like, "Why isn't this happening?" 00:02:28,440 --> 00:02:28,700 [John] Mm-hmm. 00:02:28,880 --> 00:02:34,280 [Eric] It's just called Content Triage Agent. It's the most boring name, which I now have inspiration. But- 00:02:34,320 --> 00:02:34,500 [John] Yeah. 00:02:34,500 --> 00:02:34,500 [Eric] ... 00:02:36,000 --> 00:02:38,520 [Eric] I wa- So you run 10 to 15 agents. 00:02:39,800 --> 00:02:43,780 [Eric] Can you give me a range of what they do? 'Cause in this episode, we're talking all about 00:02:44,860 --> 00:02:45,940 [Eric] AI employees. 00:02:46,180 --> 00:02:46,420 [John] Right. 00:02:46,520 --> 00:02:46,860 [Eric] And 00:02:48,060 --> 00:03:01,400 [Eric] I wanna define that, but let's do it through the lens of what are you... how are you even thinking about, uh, this AI employee, right? Because there are all sorts of different employees, um, within a business. 00:03:01,780 --> 00:03:02,080 [John] Yeah. 00:03:02,080 --> 00:03:06,400 [Eric] So do you have a framework for how you sort of categorize different agents? 00:03:07,860 --> 00:03:14,560 [John] Yeah. I think, I think there's a couple of ways to categorize. But starting off, like, just defining what we even mean 00:03:15,700 --> 00:03:18,160 [John] by an AI employee or an AI agent, is that the same thing? 00:03:18,260 --> 00:03:18,300 [Eric] Yep. 00:03:18,300 --> 00:03:22,200 [John] I think that's an interesting question to, to answer. And I would arg- 00:03:23,960 --> 00:03:25,540 [John] I think people use them interchangeably. 00:03:25,600 --> 00:03:25,880 [Eric] Mm-hmm. 00:03:25,980 --> 00:03:26,220 [John] Um, 00:03:29,520 --> 00:03:30,940 [John] and I don't know. I, I don't think it's... 00:03:32,220 --> 00:03:41,900 [John] Uh, we'll get into this, but I don't think it's... The AI employee thing can lead, can kind of lead down some bad paths as far as expectations and, like, how you interact with it. We'll cover that later. 00:03:42,120 --> 00:03:42,740 [Eric] Yep. 00:03:42,740 --> 00:03:54,580 [John] But I think, uh, more from a technical side, we do two different things, which I, I think is becoming more common. We have, like, named AI agents that have, like, roles and do things. 00:03:54,760 --> 00:03:55,260 [Eric] Mm-hmm. 00:03:55,280 --> 00:04:01,700 [John] And then we also use AI with agents and sub-agents that do things like coding. 00:04:01,740 --> 00:04:02,000 [Eric] Mm-hmm. 00:04:02,020 --> 00:04:04,720 [John] And it gets weird. It's almost like named versus unnamed. 00:04:05,240 --> 00:04:05,500 [Eric] Hmm. 00:04:05,580 --> 00:04:15,940 [John] Um, 'cause that's imp- important to dis- distinguish, 'cause we may work on a project, like a coding project, and afterwards, we've used, like, hundreds of agents technically- 00:04:16,240 --> 00:04:16,380 [Eric] Right 00:04:16,380 --> 00:04:21,040 [John] ... to, to do the work and orchestrate stuff. But they're not, like, named long-living agents. 00:04:21,079 --> 00:04:21,899 [Eric] Yes. 00:04:21,899 --> 00:04:23,640 [John] Which is where we get into the weird employee thing. 00:04:23,800 --> 00:04:24,000 [Eric] Mm-hmm. 00:04:24,080 --> 00:04:31,160 [John] Because I think the employee metaphor comes in when you want something long-living with a bit of a, like, history and memory and personality- 00:04:31,500 --> 00:04:31,840 [Eric] Yep 00:04:31,840 --> 00:04:39,980 [John] ... versus a little bit more of a transient thing that can connect to various things or make a PowerPoint or, you know, update a, 00:04:41,420 --> 00:04:42,620 [John] a SOP or something for you. 00:04:43,420 --> 00:04:53,800 [Eric] Yep. I think that the AI employee metaphor emerged for two reasons. Probably way more, but these are the two that stick out in my mind. 00:04:56,480 --> 00:04:59,400 [Eric] One is that it's a conversation. 00:05:00,520 --> 00:05:01,100 [John] Sure. 00:05:01,100 --> 00:05:07,740 [Eric] Is, is the entire form factor, right? So, which is very different than other technology or other software, right? 00:05:07,740 --> 00:05:08,280 [John] Yep. 00:05:08,280 --> 00:05:14,180 [Eric] So I use a to-do app, right? Or a pr- even a productivity application, 00:05:15,600 --> 00:05:24,940 [Eric] you know, to, to manage my tasks or manage work for my team, ticket management system, right? These are all software tools that are sort of, you know, uh, 00:05:25,960 --> 00:05:35,220 [Eric] you know, an, an anonymous, you know, benign set of, you know, techno- technological functionality that, that helps me do something, right? 00:05:36,820 --> 00:05:41,060 [Eric] AI's very nature is that it responds to you in conversational form. 00:05:41,060 --> 00:05:41,220 [John] Right. 00:05:41,220 --> 00:05:45,320 [Eric] And so it feels like you're talking with a human, right? So that's the baseline that we're starting with. 00:05:45,320 --> 00:05:45,560 [John] Right. 00:05:46,440 --> 00:05:46,740 [Eric] And 00:05:48,440 --> 00:06:00,240 [Eric] on top of that, the type of work that people use AI for in our world, which is knowledge work, can be extremely high leverage. And so people started 00:06:01,260 --> 00:06:03,696 [Eric] saying- ... years ago now 00:06:04,876 --> 00:06:11,576 [Eric] that what I used to have an intern do, I can now outsource to AI. So let me- 00:06:11,636 --> 00:06:11,656 [John] Yeah 00:06:11,656 --> 00:06:12,576 [Eric] ... give you an example. 00:06:13,756 --> 00:06:17,576 [Eric] Let me give you an example from just this week, right? So, 00:06:19,096 --> 00:06:27,836 [Eric] uh, we are working on, um, maybe rethinking the template that we use for a certain type of content that we publish multiple times per week. 00:06:29,256 --> 00:06:34,076 [Eric] And the question came up around, do we want more standardization here? Does that make sense? 00:06:34,816 --> 00:06:34,816 [John] Right. 00:06:34,915 --> 00:06:58,776 [Eric] And what do you do as the first step in that process? Okay, let's go pull six months of this content and compare all of the different formats and see if there are patterns that emerge that are, you know, we wanna keep or that we wanna change, right? Previously, you would have had an intern or, you know, someone on your team that's like, "Hey, I have a tough job for someone. This is gonna be annoying, but we just have to do it." 00:06:59,056 --> 00:06:59,276 [John] Right. 00:06:59,276 --> 00:07:07,416 [Eric] "Can you go pull all of these, you know, sort of look at the format and whatever," right? And normally, you know, you could give that to an intern and it's all they work on for an entire week. 00:07:07,715 --> 00:07:08,236 [John] Right. 00:07:08,236 --> 00:07:12,696 [Eric] Um, if it's someone who I don't wanna use all their time that way, it's like maybe just pull a month of them or like- 00:07:12,736 --> 00:07:12,796 [John] Right 00:07:12,796 --> 00:07:13,976 [Eric] ... spot check or whatever, right? 00:07:15,096 --> 00:07:30,916 [Eric] And so instead, I can just release Claude on that, and it can crunch through all of those and produce an incredible report in a short amount of time, right? So the conversational nature of generative AI and the type of work that it can- 00:07:30,936 --> 00:07:31,016 [John] Mm-hmm 00:07:31,016 --> 00:07:33,836 [Eric] ... execute as previously done by humans 00:07:34,896 --> 00:07:41,016 [Eric] just creates a context where it almost feels inevitable that people are going to use the term AI employee. But- 00:07:41,176 --> 00:07:41,696 [John] Right 00:07:41,696 --> 00:07:45,056 [Eric] ... it's kind of a charged term, which we'll talk about in a bit. 00:07:45,536 --> 00:07:49,896 [John] Yeah. Well, in the medium, right? Like if, like for us, we're using Slack as a way- 00:07:49,916 --> 00:07:49,956 [Eric] Mm-hmm 00:07:49,956 --> 00:07:52,796 [John] ... to interact with the people and the AI. 00:07:52,856 --> 00:07:53,716 [Eric] And the agents, yep. 00:07:53,776 --> 00:07:55,376 [John] And that, I think that contributes to the- 00:07:55,436 --> 00:07:56,456 [Eric] Yeah, totally. 00:07:57,136 --> 00:08:00,816 [John] Okay, so I had three categories for these employees, agents. 00:08:01,216 --> 00:08:01,496 [Eric] Yes. 00:08:01,816 --> 00:08:02,236 [John] You know? 00:08:02,296 --> 00:08:03,616 [Eric] Oh, yes. The framework I asked you about. 00:08:03,676 --> 00:08:04,616 [John] Yeah, the framework. Um, 00:08:05,696 --> 00:08:10,396 [John] you've already touched on one of them, and I wanna start with one and then get into the other two. So this first- 00:08:10,436 --> 00:08:10,436 [Eric] Yeah 00:08:10,436 --> 00:08:12,576 [John] ... one is the one that I think 00:08:13,736 --> 00:08:20,556 [John] a vast majority of people are like, this is what they think of, if they've interacted much with AI beyond, like, very basic chat. 00:08:20,736 --> 00:08:21,196 [Eric] Yep. 00:08:21,196 --> 00:08:23,396 [John] So that's person- the personal assistant would be what- 00:08:23,416 --> 00:08:23,476 [Eric] Yep 00:08:23,476 --> 00:08:27,116 [John] ... we would call it. There's a ton of productized things out there around that. 00:08:27,316 --> 00:08:27,696 [Eric] Yep. 00:08:27,696 --> 00:08:40,256 [John] The first was something called OpenClaw, and there's a follow-up to that called Hermes. These are more like really technical, mostly like kind of hobbyist people that are already in, in kind of IT or technical jobs would be into that. 00:08:40,296 --> 00:08:40,496 [Eric] Yep. 00:08:41,236 --> 00:08:41,756 [John] Since then, 00:08:42,916 --> 00:08:47,836 [John] um, kind of crossing over to, like, some less technical people, uh, something called GrokBot- 00:08:48,156 --> 00:08:48,356 [Eric] Mm-hmm 00:08:48,356 --> 00:08:52,996 [John] ... has come out. Facebook came out with Muse this week. 00:08:53,476 --> 00:08:53,536 [Eric] Mm-hmm. 00:08:53,536 --> 00:08:55,056 [John] Also kind of that personal assistant thing. 00:08:55,096 --> 00:08:55,376 [Eric] Yep. 00:08:55,376 --> 00:08:58,516 [John] There's another one that's really neat that I've been using called Instinct. 00:08:58,876 --> 00:08:59,676 [Eric] Hmm. 00:08:59,756 --> 00:09:01,956 [John] Um, it is all via Apple iMessage. 00:09:02,576 --> 00:09:03,176 [Eric] Oh. 00:09:03,176 --> 00:09:03,536 [John] And... 00:09:04,376 --> 00:09:07,416 [Eric] There's a company called Poke that does that as well. 00:09:07,436 --> 00:09:09,216 [John] Mm. Mm-hmm. Poke- 00:09:09,296 --> 00:09:12,116 [Eric] Poke or Poke? I always thought about it as Poke. 00:09:12,136 --> 00:09:13,076 [John] I don't, I don't know. I've only seen it written. 00:09:13,076 --> 00:09:13,956 [Eric] I think it's called Poke. 00:09:13,996 --> 00:09:14,236 [John] Yeah. 00:09:14,356 --> 00:09:14,656 [Eric] Um, 00:09:15,876 --> 00:09:17,276 [Eric] super popular as well. 00:09:17,316 --> 00:09:17,876 [John] Mm-hmm. 00:09:17,936 --> 00:09:19,396 [Eric] And it's all, all through iMessage. 00:09:19,456 --> 00:09:38,316 [John] Right. And then there's all of these online posts about like, like this one that I saw for that class of personal assistant, somebody that moved around a lot and, and somehow this thing went and, like, searched through emails and, like, public records to see if he was... something to do with, like, taxes and then, like- 00:09:38,356 --> 00:09:38,416 [Eric] Hmm 00:09:38,416 --> 00:09:41,296 [John] ... auto-filed some things to get, to get money back. 00:09:42,316 --> 00:09:44,056 [John] Uh, like it was a crazy use case. 00:09:44,056 --> 00:09:44,716 [Eric] Super advanced, yeah. 00:09:44,736 --> 00:09:49,056 [John] And then less things of, like, look through my emails and look for subscriptions to cancel- 00:09:49,076 --> 00:09:49,116 [Eric] Mm-hmm 00:09:49,116 --> 00:09:49,756 [John] ... or something. 00:09:49,836 --> 00:09:50,096 [Eric] Sure. 00:09:50,196 --> 00:09:56,596 [John] Um, which I, that's something I did. I had on my calendar, like, I need to s- cancel this annual subscription that I put on my calendar from a year ago. 00:09:56,636 --> 00:09:56,796 [Eric] Mm-hmm. 00:09:56,816 --> 00:09:59,876 [John] And I was like, "Hey," to one of those things, like, "Can, can you cancel this?" And it did. 00:10:00,156 --> 00:10:00,516 [Eric] Yep. Yep. 00:10:00,516 --> 00:10:09,276 [John] Which was cool. Um, but I think that's what people typically think of that, that, th- that are in the space. Like, oh, like it's a, you know, personal assistant. 00:10:09,276 --> 00:10:10,616 [Eric] Which makes sense because 00:10:11,716 --> 00:10:30,036 [Eric] when you start to interact with AI and it's, oh man, it's like, it's very good at Google searching and, um, finding information or looking through my email or these other points that are significant time friction for things that are oftentimes pretty low value, it just, you naturally gravitate- 00:10:30,056 --> 00:10:30,196 [John] Yep 00:10:30,196 --> 00:10:34,936 [Eric] ... towards those friction points that you feel in your day-to-day life, right? Or in your day-to-day work. 00:10:34,956 --> 00:10:34,976 [John] Right. 00:10:34,996 --> 00:10:40,436 [Eric] Which tend to be more menial, time-consuming, manual type, type tasks. 00:10:40,636 --> 00:10:42,716 [John] Yeah. Okay, second one. 00:10:42,756 --> 00:10:43,096 [Eric] Second. 00:10:43,836 --> 00:10:53,536 [John] The... And this one, there's actually, there was a release, like, yesterday, um, I think it's just called, like, Data Analyst or something. OpenAI came out- 00:10:53,576 --> 00:10:53,576 [Eric] Mm-hmm 00:10:53,576 --> 00:10:55,116 [John] ... with Data Analyst and it was- 00:10:55,116 --> 00:10:55,516 [Eric] Mm-hmm. 00:10:55,536 --> 00:10:59,336 [John] It's essentially a agent that lives inside your, you know, subscription- 00:11:00,016 --> 00:11:00,136 [Eric] Mm-hmm 00:11:00,136 --> 00:11:03,436 [John] ... to OpenAI that you can interact with data if, whatever data's connected- 00:11:03,516 --> 00:11:03,636 [Eric] Mm-hmm 00:11:03,636 --> 00:11:09,156 [John] ... to it. So I mean, that's the s- the place my team and I, like, work the most- 00:11:09,216 --> 00:11:09,276 [Eric] Mm-hmm 00:11:09,276 --> 00:11:14,076 [John] ... with agents. Um, we're, we're spending time customizing agents 00:11:15,176 --> 00:11:18,096 [John] to pull data together from a lot of different systems- 00:11:18,296 --> 00:11:18,456 [Eric] Mm-hmm 00:11:18,456 --> 00:11:22,636 [John] ... that have a lot of different interfaces, [chuckles] many of which the interfaces are really bad. 00:11:22,956 --> 00:11:23,536 [Eric] Yep. 00:11:23,536 --> 00:11:24,676 [John] And then customizing 00:11:25,756 --> 00:11:34,636 [John] the outputs, which could honestly, like, Excel is back. Like, people love using agents to pull from a bunch of things and, like, get an Excel file. 00:11:34,736 --> 00:11:35,176 [Eric] Isn't that great? 00:11:35,216 --> 00:11:35,716 [John] Yeah. 00:11:35,776 --> 00:11:39,036 [Eric] It's the best. [both laughing] It is absolutely the best- 00:11:39,116 --> 00:11:39,496 [John] Yeah 00:11:39,496 --> 00:11:43,796 [Eric] ... sort of ir- ironic result of AI. 00:11:44,876 --> 00:11:44,996 [John] Right. 00:11:44,996 --> 00:11:51,836 [Eric] But just to, just to, to reiterate this, it's incredible to use a spreadsheet with AI. 00:11:52,176 --> 00:11:52,796 [John] Yeah. Yeah. 00:11:52,796 --> 00:11:53,856 [Eric] Absolutely incredible. 00:11:53,936 --> 00:12:07,056 [John] Well, and it's such an interesting form factor, 'cause I'll just use, like, Anthropic Claude for an example. So we're, we're building agents, like data analyst agents, like customer research agents, like things like that. 00:12:07,096 --> 00:12:07,256 [Eric] Mm-hmm. 00:12:07,276 --> 00:12:13,660 [John] And, and companies, like, especially like tech companies- The software subscription sprawl is incredible. 00:12:13,800 --> 00:12:13,820 [Eric] Yes. 00:12:13,820 --> 00:12:19,620 [John] Like if, if you're like, "I want an agent across sales and marketing," you're like, guess what? It needs to connect to 30 different tools. 00:12:20,140 --> 00:12:20,620 [Eric] It's wild. 00:12:20,760 --> 00:12:22,500 [John] Yeah. Or 50 different tools. 00:12:22,540 --> 00:12:22,620 [Eric] Mm-hmm. 00:12:22,700 --> 00:12:33,360 [John] It's wild. Um, and so that's a challenge. And then, you know, obviously we, we do work with data and data warehousing, so part of it, like, well, how much data do you like put in a database versus connecting- 00:12:33,360 --> 00:12:33,380 [Eric] Mm-hmm 00:12:33,380 --> 00:12:39,900 [John] ... directly, like a whole thing. But the end result, what people actually want, often it's a spreadsheet- 00:12:40,140 --> 00:12:40,300 [Eric] Mm-hmm 00:12:40,300 --> 00:12:48,760 [John] ... and I like wanna compare things. Sometimes it's analysis of things I wouldn't be able to analyze before, which would be call transcripts, email- 00:12:48,900 --> 00:12:49,720 [Eric] Hmm 00:12:49,720 --> 00:12:52,860 [John] ... or, um, uh, customer support tickets. 00:12:53,060 --> 00:12:53,340 [Eric] Yep. 00:12:53,360 --> 00:12:55,740 [John] Those are kind of like you couldn't really analyze that easily. 00:12:55,880 --> 00:12:56,660 [Eric] Yep. 00:12:56,660 --> 00:12:58,200 [John] Um, or some combination of both. 00:12:58,620 --> 00:13:05,580 [Eric] Totally. I think this is probably the highest leverage, 00:13:07,580 --> 00:13:09,780 [Eric] least used use case- 00:13:09,940 --> 00:13:10,040 [John] Yeah 00:13:10,040 --> 00:13:15,700 [Eric] ... for AI, or for the purpose of this episode, having an AI employee- 00:13:15,740 --> 00:13:15,860 [John] Right 00:13:15,860 --> 00:13:17,420 [Eric] ... that does research and analysis. 00:13:17,460 --> 00:13:18,340 [John] Right. 00:13:18,340 --> 00:13:20,340 [Eric] It's, it is truly incredible. 00:13:20,400 --> 00:13:20,600 [John] Yeah. 00:13:20,700 --> 00:13:22,460 [Eric] Um, it is truly incredible. 00:13:22,520 --> 00:13:32,260 [John] It's so good. And from, from building them and just 'cause we, we've done a lot. I mean, we've done this for, honestly for years, kind of like locally. 00:13:32,320 --> 00:13:33,060 [Eric] Mm-hmm. 00:13:33,060 --> 00:13:39,580 [John] We'll get into this in a minute, but moving it into kind of like a named entity with more of a, like a role and a description- 00:13:39,600 --> 00:13:39,680 [Eric] Yeah 00:13:39,680 --> 00:13:41,580 [John] ... that multiple people work on is a really big- 00:13:41,580 --> 00:13:42,840 [Eric] A job description. [chuckles] 00:13:42,880 --> 00:13:48,520 [John] Yeah. But it's a really big change, and sometimes it's hard to articulate why it's better. 00:13:49,400 --> 00:13:50,360 [Eric] Hmm. 00:13:50,380 --> 00:13:51,560 [John] Um, and we'll get into that a little bit- 00:13:51,720 --> 00:13:51,860 [Eric] Yeah, yeah 00:13:51,860 --> 00:13:53,500 [John] ... in a second. But I've got a third one- 00:13:53,700 --> 00:13:53,900 [Eric] Okay 00:13:53,900 --> 00:13:55,720 [John] ... that touches on- 00:13:55,800 --> 00:13:56,820 [Eric] So personal assistant. 00:13:56,860 --> 00:13:58,180 [John] Yeah, personal assistant. 00:13:58,180 --> 00:13:59,220 [Eric] Analyst or researcher. 00:13:59,320 --> 00:14:01,580 [John] Yeah, analyst, researcher. The third one 00:14:02,900 --> 00:14:05,880 [John] is advisor slash coach. 00:14:06,160 --> 00:14:08,580 [Eric] Hmm. Ooh. Okay, this is... [chuckles] 00:14:08,660 --> 00:14:15,140 [John] This one is so interesting because it has the potential to be incredibly helpful and incredibly damaging. 00:14:15,400 --> 00:14:21,080 [Eric] Hmm. Yes. Immediately my, my warning flags went up. 00:14:21,420 --> 00:14:21,580 [John] Right. 00:14:21,580 --> 00:14:27,360 [Eric] But d- okay, can you give me an exa- like a concrete example of 00:14:28,500 --> 00:14:30,120 [Eric] the, the positive- 00:14:30,540 --> 00:14:30,820 [John] Yeah 00:14:30,820 --> 00:14:35,460 [Eric] ... angle on AI employee as advisor or coach, AI advisor? 00:14:35,520 --> 00:14:44,760 [John] Right. Um, I think it is really positive to help increase your personal competency into something that you're not like super competent at. 00:14:44,800 --> 00:14:45,740 [Eric] Hmm. 00:14:45,740 --> 00:14:50,520 [John] So like I have one for, that I started for like sales. 00:14:50,940 --> 00:14:50,980 [Eric] Hmm. 00:14:50,980 --> 00:14:54,520 [John] Kind of like sales, sales and marketing, which I don't have like a strong background in sales and marketing. 00:14:55,120 --> 00:14:55,340 [Eric] Yep. 00:14:55,340 --> 00:14:55,540 [John] Um, 00:14:56,680 --> 00:15:06,360 [John] which I think is really interesting. And some practical things like, like you might imagine with any like new skill- 00:15:06,620 --> 00:15:06,820 [Eric] Mm-hmm 00:15:06,820 --> 00:15:13,560 [John] ... is it will state things that you're like, "Yeah, duh." But did you need to hear it? Yeah, you did. 00:15:13,560 --> 00:15:14,420 [Eric] [chuckles] 00:15:14,420 --> 00:15:14,420 [John] So- 00:15:14,480 --> 00:15:14,900 [Eric] Which is- 00:15:14,900 --> 00:15:15,920 [John] ... which is a lot of what coaching- 00:15:15,920 --> 00:15:17,920 [Eric] ... which is actually just advisory. 00:15:18,000 --> 00:15:18,480 [John] Yeah. Yeah. 00:15:18,520 --> 00:15:22,660 [Eric] Just like, "I'm going to say the same thing and over and over, and then at some point- 00:15:22,760 --> 00:15:22,860 [John] Right 00:15:22,860 --> 00:15:27,740 [Eric] ... your personal experience is gonna line up with a critical situation where you're like, 'Oh, I get it.'" 00:15:27,780 --> 00:15:27,880 [John] Yeah. 00:15:27,880 --> 00:15:28,840 [Eric] [chuckles] 00:15:28,900 --> 00:15:34,840 [John] Yeah. So not to undermine at all like advisory or coaching, but that is a lot of it. 00:15:34,840 --> 00:15:35,560 [Eric] For sure. 00:15:35,560 --> 00:15:42,560 [John] And I think that's the value point of saying kind of the same things over and over, some kind of built-in accountability- 00:15:42,680 --> 00:15:42,860 [Eric] Mm-hmm 00:15:42,860 --> 00:15:47,300 [John] ... of what should I do next, blah, blah, blah. And it's like, "Well, you should reach out to five people or 10." You know what I mean? 00:15:47,320 --> 00:15:47,360 [Eric] Yeah. Sure. 00:15:47,360 --> 00:15:53,720 [John] Just like it's, like it's helpful but simple, and you probably could have built a non-AI thing to do it- 00:15:54,020 --> 00:15:54,180 [Eric] Yeah 00:15:54,180 --> 00:15:55,060 [John] ... in a lot of ways. 00:15:55,080 --> 00:15:55,160 [Eric] Yeah. 00:15:55,180 --> 00:16:15,080 [John] And, uh, in other ways, like say you wanted more of a like, had like a legal advisor or things like that, like sure, there's some risk there, but there's also more advanced use case where you're, where, where you're kind of over- overlapping research analyst and advisor in just a domain like a, like, I don't know, tax or legal or whatever- 00:16:15,160 --> 00:16:15,220 [Eric] Yep 00:16:15,220 --> 00:16:17,380 [John] ... that you just don't know and you kind of need both skills. 00:16:17,440 --> 00:16:17,640 [Eric] Mm-hmm. 00:16:17,700 --> 00:16:18,520 [John] I think that's- 00:16:18,560 --> 00:16:18,620 [Eric] Yeah 00:16:18,620 --> 00:16:27,680 [John] ... interesting and helpful and, you know, potentially dangerous depending on [chuckles] what the stakes are. So you have to figure out what are the stakes and then make the decision. 00:16:27,700 --> 00:16:28,240 [Eric] Absolutely. 00:16:28,340 --> 00:16:28,380 [John] Right. 00:16:28,380 --> 00:16:29,800 [Eric] It's an incredible... 00:16:31,120 --> 00:16:34,460 [Eric] AI is incredibly useful for 00:16:36,180 --> 00:16:37,620 [Eric] teaching yourself a new topic. 00:16:37,980 --> 00:16:38,180 [John] Yeah. 00:16:38,220 --> 00:16:42,240 [Eric] It, it's, it can truly be amazing. I mean, we do this all the time, 00:16:43,420 --> 00:16:44,680 [Eric] um, on my team 00:16:45,700 --> 00:16:50,760 [Eric] where there may be a new... Let's say we're exploring a new product area, 00:16:52,180 --> 00:17:06,160 [Eric] um, a new product area, uh, in Vercel for the company I work at, right? And so we're thinking about the content we wanna create about, a- around that, right? And maybe there's someone on the team who just doesn't have as much experience in that. They can get up to speed super quickly. 00:17:06,480 --> 00:17:06,560 [John] Yeah. 00:17:06,560 --> 00:17:25,600 [Eric] Not only from the standpoint of I want, I need to consume information about, you know, this new technology that I'm not as familiar with, but I also can implement it. I mean, we do this... This is such a great way to learn technology is I wanna build something and implement it, but I wanna do that in a way where you walk me through step by step- 00:17:25,700 --> 00:17:25,720 [John] Yeah 00:17:25,720 --> 00:17:29,360 [Eric] ... why we're doing each piece, how it works, et cetera, right? 00:17:29,380 --> 00:17:29,460 [John] Yep. 00:17:29,460 --> 00:17:32,540 [Eric] And it truly is incredible. But also, 00:17:33,720 --> 00:17:38,640 [Eric] of course, as we've said many times on the show, AI can be very agreeable. And so- 00:17:38,780 --> 00:17:39,020 [John] Right 00:17:39,020 --> 00:17:53,640 [Eric] ... if there is something that you want to hear, uh, it can, you can easily steer it to tell you things that you wanna hear. But great advisors often tell you things that are, are difficult to hear, or they tell you the same things [chuckles] over and over again. 00:17:53,640 --> 00:17:57,000 [John] And, and the, and the good news, bad news there is 00:17:58,220 --> 00:17:58,520 [John] the 00:17:59,620 --> 00:18:10,800 [John] advisor tends to head down a direct... AI in general heads down a direction, and there's a pretty strong consistency that wants to maintain consistency- 00:18:11,260 --> 00:18:11,480 [Eric] Hmm 00:18:11,480 --> 00:18:13,924 [John] ... I've noticed. So- 00:18:13,924 --> 00:18:18,224 [John] For me, I actually, I use the same chat thread for the advisor on this- 00:18:18,284 --> 00:18:18,344 [Eric] Mm-hmm. 00:18:18,344 --> 00:18:20,664 [John] ... and, like, never close it and just use the same- 00:18:21,104 --> 00:18:21,204 [Eric] Mm-hmm. 00:18:21,204 --> 00:18:26,544 [John] ... chat thread. And, and for AI, like, there's, there's ways to ah, compact, I think is what it's called- 00:18:26,564 --> 00:18:26,624 [Eric] Mm-hmm. 00:18:26,624 --> 00:18:39,224 [John] ... universally now. Um, so I always use the same chat thread. When it's down the right course, it's really useful and if you've corrected along the way of like, "No, I want you to push back," and blah, blah, blah- 00:18:39,264 --> 00:18:39,344 [Eric] Yeah. 00:18:39,344 --> 00:18:40,704 [John] ... it, it kinda stays the course. 00:18:40,744 --> 00:18:41,504 [Eric] Yep. 00:18:41,504 --> 00:18:47,224 [John] But if you've corrected, if you've gone a bad direction, it will do the same thing. 00:18:47,264 --> 00:18:47,584 [Eric] It'll do the same thing. Yes. 00:18:47,584 --> 00:18:49,044 [John] It will stay that direction. 00:18:49,084 --> 00:18:50,044 [Eric] Yep, yep. 00:18:50,044 --> 00:18:51,564 [John] And I think that's the dangerous part. 00:18:52,244 --> 00:18:52,484 [Eric] So 00:18:54,364 --> 00:19:02,324 [Eric] on the show, we like to take a topic like AI employees and step back and say, "Okay, what questions do we want to ask about AI," 00:19:03,404 --> 00:19:04,984 [Eric] right? In, in this context. 00:19:05,064 --> 00:19:05,084 [John] Right. 00:19:05,084 --> 00:19:07,604 [Eric] And so I thought of a great one for the AI employee. 00:19:09,844 --> 00:19:20,924 [Eric] Like we said earlier, it's an easy m- it's an easy metaphor to reach t- to reach for because of the conditions, the type of work it takes on, and the fact that it's conversational. 00:19:22,124 --> 00:19:28,264 [Eric] But my question is, do we actually want an AI employee? It- 00:19:28,304 --> 00:19:28,404 [John] Right. 00:19:28,404 --> 00:19:30,524 [Eric] Like, is that the right metaphor? 00:19:30,604 --> 00:19:30,784 [John] Right. 00:19:30,844 --> 00:19:44,424 [Eric] It's easy. It makes sense. I mean, you almost, you almost don't even think about it. It's like, oh, AI employee, right? As a personal assistant or as an analyst, and you just frame it that way. But is that actually what we want, even though it's sort of the easiest metaphor to reach for? 00:19:45,744 --> 00:19:47,364 [John] I mean, I would argue I don't, I don't think so. 00:19:48,904 --> 00:19:49,064 [John] Well, 00:19:50,524 --> 00:19:52,984 [John] probab- I think it depends. 00:19:52,984 --> 00:19:53,384 [Eric] [laughing] 00:19:53,464 --> 00:19:53,524 [John] I- 00:19:53,544 --> 00:19:55,584 [Eric] This is... That, that means it's a good question. 00:19:55,664 --> 00:20:01,864 [John] Like, yeah. [chuckling] 'Cause I'd argue no, but then if you ask it theoretically, it's yes, 00:20:02,884 --> 00:20:10,084 [John] right? Like, do you want an employee or do you want just something that can do tasks? Do you want an employee or do you want just another piece- 00:20:10,084 --> 00:20:10,324 [Eric] Mm-hmm. 00:20:10,324 --> 00:20:12,224 [John] ... of software? Like, I want an employee. 00:20:12,444 --> 00:20:12,544 [Eric] Yeah. 00:20:12,544 --> 00:20:28,744 [John] You know? So I, I do think it is in some ways like an, an ideal for, for at least a certain group of people of like, yeah, we want this to be, like, as human as possible, to give it a job description, to do work independently, and make the company more money or whatever your goal is. 00:20:28,944 --> 00:20:29,324 [Eric] Mm-hmm. Mm-hmm. 00:20:29,344 --> 00:20:30,944 [John] So I do think that's a lot of people's goals. 00:20:31,304 --> 00:20:31,704 [Eric] Yep. 00:20:31,704 --> 00:20:35,184 [John] Um, I think there's dangers there, which we can get into in a minute. 00:20:36,324 --> 00:20:45,444 [John] But me personally, I've framed it a little bit differently of what I want is I want something that is multiplayer and collaborative- 00:20:45,644 --> 00:20:45,864 [Eric] Mm-hmm. 00:20:45,864 --> 00:20:50,144 [John] ... and I want something that can do asynchronous or delegated work. 00:20:50,204 --> 00:20:50,704 [Eric] Mm-hmm. 00:20:50,724 --> 00:20:53,084 [John] Like, those are the two things that I personally want. 00:20:53,864 --> 00:21:02,764 [Eric] So before we get... The collaborative thing is very interesting, and I wanna, I wanna dig into that because I think for most people, it's a very individual experience. 00:21:02,884 --> 00:21:03,404 [John] Mm-hmm. 00:21:03,524 --> 00:21:13,924 [Eric] Um, and that's starting to change. But what, what is... You said dangerous. What is dangerous about the metaphor? 00:21:14,744 --> 00:21:15,924 [John] The employee metaphor? 00:21:16,004 --> 00:21:16,764 [Eric] Yeah. 00:21:16,844 --> 00:21:24,064 [John] Yeah. I mean, if you take it to the n- to the very end, it's treating a machine like a human and then 00:21:25,144 --> 00:21:27,144 [John] potentially getting confused between humans and machines- 00:21:27,244 --> 00:21:27,364 [Eric] Hmm. 00:21:27,364 --> 00:21:28,424 [John] ... basically. 00:21:28,444 --> 00:21:33,964 [Eric] Hmm. Yeah, when you, when you said this is the ideal, 00:21:35,764 --> 00:21:35,824 [Eric] I 00:21:37,264 --> 00:21:40,044 [Eric] immediately thought, "Okay, well, 00:21:41,304 --> 00:21:43,284 [Eric] yes, in theory, that's the ideal." 00:21:43,764 --> 00:21:43,804 [John] Right. 00:21:43,804 --> 00:21:45,744 [Eric] Right? But the... I think part of the- 00:21:45,784 --> 00:21:46,464 [John] Only in theory, though- 00:21:46,604 --> 00:21:46,764 [Eric] I think- 00:21:46,764 --> 00:21:47,444 [John] ... is my opinion. 00:21:47,704 --> 00:21:50,604 [Eric] Well, uh, part of the problem with that, and I think this is where 00:21:51,984 --> 00:21:54,544 [Eric] the subconscious, 00:21:55,624 --> 00:22:01,864 [Eric] the subconscious way of thinking that can creep in can kind of be insidious- 00:22:02,224 --> 00:22:02,444 [John] Mm-hmm. 00:22:02,444 --> 00:22:02,444 [Eric] ... 00:22:04,224 --> 00:22:14,624 [Eric] uh, which I don't think is too strong of a word. Because if you think about I have this ideal of an AI employee, like do you want technology or do you want an employee? 00:22:15,184 --> 00:22:15,404 [John] Right. 00:22:15,404 --> 00:22:20,484 [Eric] The reason you want AI to be an employee is because it's so much cheaper than an actual human. 00:22:22,484 --> 00:22:22,664 [Eric] Right? 00:22:22,684 --> 00:22:22,684 [John] Yeah. 00:22:22,704 --> 00:22:25,344 [Eric] Like, the reason, the reason... That's a big reason that it's- 00:22:25,364 --> 00:22:26,024 [John] Yeah, that's one factor. 00:22:26,644 --> 00:22:27,224 [Eric] It's a... Yeah. 00:22:27,264 --> 00:22:27,464 [John] Yeah. 00:22:27,544 --> 00:22:28,744 [Eric] I would argue it's a major factor. 00:22:28,784 --> 00:22:29,044 [John] It's a ma- 00:22:29,104 --> 00:22:29,404 [Eric] Right? It's like- 00:22:29,424 --> 00:22:31,024 [John] No, no. It's a, it's a major factor, 00:22:32,064 --> 00:22:44,344 [John] but I wouldn't say it's just financially... It's not just financially cheaper. 'Cau- 'cause, like, taking the employee, like, pe- people don't actually want the full human version. People want better than the human version. They want- 00:22:44,384 --> 00:22:44,424 [Eric] Hmm. 00:22:44,424 --> 00:22:48,784 [John] I want the, I want the cheaper, I want the no drama, never quits- 00:22:49,064 --> 00:22:49,824 [Eric] Yes, yes. This is- 00:22:49,824 --> 00:22:53,044 [John] ... has perfect retention, can work 10 times faster- 00:22:53,244 --> 00:22:53,564 [Eric] Yes. 00:22:53,564 --> 00:23:00,664 [John] ... version. So even that breaks down as far as like, I, I, I just want an optimum human employee. Like, no wait. Like, no. 00:23:00,664 --> 00:23:01,004 [Eric] [chuckling] 00:23:01,004 --> 00:23:09,244 [John] We want this, which is like the best of a human employee with, like, a bunch of extra things where it's, like, cheaper, smarter, works harder, blah, blah, blah. 00:23:09,244 --> 00:23:10,104 [Eric] Yes, yes. 00:23:10,104 --> 00:23:10,284 [John] And- 00:23:10,324 --> 00:23:10,724 [Eric] Well said. 00:23:11,364 --> 00:23:16,784 [John] Yeah. And honestly, that's what... I think that's what all the labs want. I don't, I don't know that I personally want that, but- 00:23:16,844 --> 00:23:17,024 [Eric] Yeah. 00:23:17,024 --> 00:23:19,364 [John] ... but I think that's where all this is, like, heading. 00:23:19,704 --> 00:23:28,424 [Eric] Right. Right. I... Confusing humans for machines and machines for humans I think is where it really does go wrong. 00:23:28,684 --> 00:23:28,884 [John] Right. 00:23:28,884 --> 00:23:29,304 [Eric] Um, 00:23:30,704 --> 00:23:38,044 [Eric] because it can be easy to forget AI is a machine. It's generative, it's conversational- 00:23:38,324 --> 00:23:38,484 [John] Right. 00:23:38,484 --> 00:23:39,804 [Eric] ... but it's not a human, 00:23:41,104 --> 00:23:44,444 [Eric] and it's important to maintain that distinction. 00:23:44,484 --> 00:23:44,944 [John] Mm-hmm. 00:23:45,044 --> 00:23:47,164 [Eric] Um, and likewise, 00:23:48,284 --> 00:23:48,904 [Eric] if you, 00:23:50,344 --> 00:23:50,924 [Eric] if you 00:23:52,024 --> 00:23:53,404 [Eric] frame AI 00:23:54,444 --> 00:23:57,524 [Eric] in human terms in your mind or in your thinking- 00:23:58,764 --> 00:23:58,764 [John] Mm-hmm. 00:23:58,764 --> 00:24:10,644 [Eric] ... then it can start to shape the way that you think about people and even the way that you evaluate them, right? And so you can start to subconsciously think, "Well, this person's, like, hard to manage." You know? It's like, well- 00:24:10,684 --> 00:24:10,724 [John] Yeah. 00:24:10,724 --> 00:24:12,964 [Eric] ... they're, they're a person. People are hard to manage. 00:24:12,984 --> 00:24:13,024 [John] Yeah. 00:24:13,024 --> 00:24:15,664 [Eric] Anyone who's managed people knows that that's just difficult, right? 00:24:15,684 --> 00:24:16,604 [John] Yep. 00:24:16,604 --> 00:24:18,364 [Eric] So I agree. I think that's a great way to frame it. 00:24:19,496 --> 00:24:23,916 [John] Okay, so there's two specific things that happened this week that I'm excited to bring up, 00:24:25,456 --> 00:24:27,836 [John] kind of on the downsides of the, the 00:24:29,176 --> 00:24:34,876 [John] people using AI, that the machine, the downsides of the machine versus the human piece. 00:24:35,056 --> 00:24:35,196 [Eric] Mm-hmm. 00:24:35,376 --> 00:24:40,996 [John] One of them, which everybody's heard this at this point, is the AI slop problem. 00:24:41,036 --> 00:24:41,796 [Eric] Yes. Yep. 00:24:41,816 --> 00:24:53,876 [John] And the other one, um, I'll get to in a minute. But that one, um, this was so funny. First, first I've, um, heard of this. There is a, a company I work with, um, and they have a new emoji in their Slack. 00:24:54,076 --> 00:24:54,336 [Eric] Mm-hmm. 00:24:54,996 --> 00:24:59,996 [John] To be-- I mean, I think most people work with Slack or Teams. Like, you can react to messages with a thumbs up or whatever- 00:25:00,096 --> 00:25:00,176 [Eric] Yep 00:25:00,176 --> 00:25:10,116 [John] ... or, you know, or a custom emoji. They have a new one, and it is called... So I think, I think there's one out there, I've never used it personally, called, like, TLDR or something like, like too long, didn't read. 00:25:10,156 --> 00:25:10,636 [Eric] Mm-hmm. 00:25:10,656 --> 00:25:14,816 [John] They have one called, I think it's AIDR, like AI, like, didn't read. 00:25:14,816 --> 00:25:15,616 [Eric] [chuckles] 00:25:15,616 --> 00:25:18,476 [John] So I, I have never seen it used. I just know it exists. 00:25:18,476 --> 00:25:18,856 [Eric] [chuckles] 00:25:18,856 --> 00:25:22,776 [John] But I think it, what it is, if somebody pastes or sends some long 00:25:24,216 --> 00:25:26,036 [John] AI-looking generated thing- 00:25:26,656 --> 00:25:26,656 [Eric] Uh-huh 00:25:26,656 --> 00:25:29,756 [John] ... you can react to the thing with, like, "AI didn't read." 00:25:29,756 --> 00:25:35,416 [Eric] [chuckles] That's so good. [both chuckle] Okay, what was the other one? 00:25:36,096 --> 00:25:38,056 [John] Uh, the other one, um... 00:25:39,596 --> 00:25:40,956 [John] Gosh, what was the other one? 00:25:42,936 --> 00:25:43,536 [John] I can't remember. 00:25:43,956 --> 00:25:44,456 [Eric] You said there were two. 00:25:44,456 --> 00:25:45,736 [John] I know. Yeah. I think I said there were two. 00:25:45,756 --> 00:25:46,036 [Eric] It'll come back to you. 00:25:46,056 --> 00:25:46,516 [John] It'll come back to me. Yeah. 00:25:46,516 --> 00:25:50,056 [Eric] It'll come back to you. Yeah, yeah. It'll come back to you. Okay, talk to me about collaborative- 00:25:50,396 --> 00:25:51,316 [John] Yeah. 00:25:51,316 --> 00:25:51,916 [Eric] Talk to me about- 00:25:52,416 --> 00:25:53,536 [John] Yeah, on the, on the positive side. 00:25:53,676 --> 00:25:59,856 [Eric] Yes. Okay, so AI, do we actually want AI as an employee? Is that even the right metaphor? 00:26:00,036 --> 00:26:00,056 [John] Right. 00:26:00,076 --> 00:26:05,636 [Eric] Right? In some ways, it's not the right metaphor. It can be very dangerous 'cause it can subtly change the way you think- 00:26:05,656 --> 00:26:05,736 [John] Right 00:26:05,736 --> 00:26:12,116 [Eric] ... about AI or about humans. But you mentioned collaborative. You want a collaborative experience. Now- 00:26:12,176 --> 00:26:12,296 [John] Right 00:26:12,296 --> 00:26:19,676 [Eric] ... I mean, you can kind of collaborate with a bunch of AI agents, but do you mean literally humans collaborating with AI agents? 00:26:21,416 --> 00:26:21,716 [John] Um- 00:26:21,796 --> 00:26:24,936 [Eric] Like, humans collaborating with each other and AI agents. What did you mean by that? 00:26:25,216 --> 00:26:27,936 [John] Yeah. So the term I've heard a lot recently is multiplayer. 00:26:28,256 --> 00:26:29,176 [Eric] Yep. 00:26:29,196 --> 00:26:37,656 [John] I really like that turn, term. It's a little bit, um... And it's, I, I think it's maybe even intentio- intentionally ambiguous. 00:26:38,076 --> 00:26:38,276 [Eric] Mm-hmm. 00:26:38,596 --> 00:26:43,116 [John] As in multiplayer, like, who are the players? Humans and machines, I think, are the players. 00:26:43,176 --> 00:26:43,916 [Eric] Hmm. 00:26:44,016 --> 00:27:10,256 [John] Um, and I think the positives... And I think as a real negative, and I, and sa- like, I heard a story this morning. It was really sad. So there's a, um, [lip smack] startup that I know of. I've never actually met the founder, um, but, but kind of in the local sphere, like where we live, who essentially, like, really sadly, like, has really gotten into AI, and I think had a small team. And it, and it seems like, 00:27:11,836 --> 00:27:12,116 [John] um, 00:27:13,176 --> 00:27:22,476 [John] like is, is pretty reliant on AI, like, to do all the work, and, like, doesn't seem to be being able to leverage the humans that were already part of the startup- 00:27:22,696 --> 00:27:22,816 [Eric] Hmm 00:27:22,816 --> 00:27:32,476 [John] ... is, like, the best way I can say that. Where, like, hadn't cracked, how do I make this multiplayer so we can build machines that build other machines- 00:27:33,256 --> 00:27:33,516 [Eric] Interesting 00:27:33,516 --> 00:27:35,176 [John] ... together, um, 00:27:36,476 --> 00:27:41,676 [John] [lip smack] with... And by together, I mean, like, the humans building the agents- 00:27:41,796 --> 00:27:41,916 [Eric] Mm-hmm 00:27:41,916 --> 00:27:44,276 [John] ... or employees, if you wanna call it that, but agents. 00:27:44,476 --> 00:27:44,956 [Eric] Yep. 00:27:44,976 --> 00:27:54,656 [John] Um, and I think that's, I think that's the best place because it's defense against, which we've talked about before, it's defense against AI psychosis 'cause you're like- 00:27:54,756 --> 00:27:55,116 [Eric] Hmm 00:27:55,116 --> 00:27:58,576 [John] ... may- maybe not AI psychosis group think, which could be a whole thing. 00:27:58,576 --> 00:27:58,856 [Eric] [chuckles] 00:27:58,856 --> 00:28:00,616 [John] But, like, it's a little bit of defense against, like- 00:28:00,636 --> 00:28:00,656 [Eric] Mm-hmm 00:28:00,656 --> 00:28:02,716 [John] ... psychosis 'cause you worked on something together. 00:28:03,036 --> 00:28:03,096 [Eric] Yep. 00:28:03,136 --> 00:28:06,236 [John] You have another human's opinion as part of the- 00:28:06,476 --> 00:28:06,676 [Eric] Mm-hmm 00:28:06,676 --> 00:28:11,856 [John] ... like, "No, like, I think this is bad," or, you know, "No, we should change it this way." Like, you actually have- 00:28:11,876 --> 00:28:11,936 [Eric] Yep 00:28:11,936 --> 00:28:13,556 [John] ... like, multiple humans' opinions involved. 00:28:13,796 --> 00:28:13,996 [Eric] Yep. 00:28:14,016 --> 00:28:23,836 [John] Which I think is good. And you have a collaborative nature. We're, we're, you know... I mean, technical people can already have a little bit of, like, reclusiveness to them. 00:28:23,876 --> 00:28:24,596 [Eric] Yes. Yep. 00:28:24,596 --> 00:28:33,296 [John] And I think AI can really amplify that in a negative way when it's, there's, like, a forcing function of like, "Okay, we're gonna, like, build this thing together." 00:28:33,336 --> 00:28:33,556 [Eric] Mm-hmm. 00:28:33,876 --> 00:28:37,256 [John] So I think both of those are, are positive things of- 00:28:37,395 --> 00:28:37,436 [Eric] Yep 00:28:37,436 --> 00:28:44,096 [John] ... thinking about this in more of a multiplayer way. Um, and then, and then the last positive I'll bring up is, 00:28:45,116 --> 00:28:48,116 [John] um, which is a, which is a plus of the metaphor. 00:28:49,256 --> 00:28:53,236 [John] A lot of what people wanna do is work asynchronously with AI- 00:28:53,276 --> 00:28:53,276 [Eric] Hmm 00:28:53,276 --> 00:28:55,096 [John] ... or delegate to AI. 00:28:55,696 --> 00:28:55,796 [Eric] Yep. 00:28:55,796 --> 00:29:02,716 [John] Um, I do think that's positive. I think there's a lot of downsides in being in, like, multiple constant chats all day- 00:29:03,416 --> 00:29:03,416 [Eric] Yeah 00:29:03,416 --> 00:29:05,356 [John] ... with AI where we're going back and forth and back and forth and back and forth- 00:29:05,556 --> 00:29:05,596 [Eric] Sure 00:29:05,596 --> 00:29:15,136 [John] ... which is a lot of people's workflows. Um, that is one of the positives that goes along well with multiplayer, um, that I'm really excited about. 00:29:15,396 --> 00:29:16,016 [Eric] Yeah. I, 00:29:18,216 --> 00:29:26,856 [Eric] I, I agree. I think it kind of brings back the, the danger of the metaphor side, not to, [chuckles] not to just- 00:29:26,896 --> 00:29:26,936 [John] Yeah 00:29:26,936 --> 00:29:27,776 [Eric] ... return to the negative. 00:29:27,796 --> 00:29:28,876 [John] Yeah. 00:29:28,896 --> 00:29:30,756 [Eric] But if we think about, 00:29:32,276 --> 00:29:43,656 [Eric] if we think about multiplayer in the context of humans and agents working together in a collaborative environment, right? So let's say Slack. This is becoming more and more common, right? Where 00:29:44,676 --> 00:29:48,436 [Eric] Slack agents are in Slack along with humans and- 00:29:48,516 --> 00:29:48,816 [John] Right 00:29:48,816 --> 00:29:52,316 [Eric] ... you have a thread where agents and humans are- 00:29:52,376 --> 00:29:52,416 [John] Yeah 00:29:52,416 --> 00:29:53,396 [Eric] ... you know, collaborating. 00:29:53,436 --> 00:29:53,616 [John] Mm-hmm. 00:29:54,776 --> 00:30:04,176 [Eric] And... But if you think about the startup that you mentioned, it kind of goes back to do you, i- in the way that you frame 00:30:05,556 --> 00:30:10,696 [Eric] AI employees, do you have m- a displacement mindset- 00:30:10,796 --> 00:30:11,016 [John] Yeah 00:30:11,016 --> 00:30:15,176 [Eric] ... where you think about AI employees displacing 00:30:16,296 --> 00:30:19,676 [Eric] other, you know, displacing em- uh, human employees? 00:30:20,096 --> 00:30:20,316 [John] Right. 00:30:21,116 --> 00:30:22,564 [Eric] Or- 00:30:22,564 --> 00:30:31,504 [Eric] Do you have more of a leverage mindset, where you think about a human employee using AI employees to dramatically increase output? Now- 00:30:31,744 --> 00:30:31,844 [John] Yeah 00:30:31,844 --> 00:30:31,844 [Eric] ... 00:30:33,124 --> 00:30:41,264 [Eric] 100%, there are jobs that are highly manual human jobs that are already being automated by AI. 00:30:41,964 --> 00:30:42,004 [John] Yeah. 00:30:42,004 --> 00:30:48,144 [Eric] And I think to some extent, that's unavoidable, and that's a, that's a tale as old as time in terms of technology, right? 00:30:48,204 --> 00:30:48,404 [John] Yeah. 00:30:48,404 --> 00:30:50,104 [Eric] Like, there are some jobs that are displaced. 00:30:50,524 --> 00:30:51,464 [John] Yep. 00:30:51,464 --> 00:31:12,184 [Eric] I think people can tend to over-rotate on that, um, view of AI and not have a leverage mindset, which says, "Okay, all things being equal, if we take the same set of employees and empower them with AI, or AI employees, if you wanna call it that, how much more output can we produce?" And- 00:31:12,224 --> 00:31:12,364 [John] Right 00:31:12,364 --> 00:31:18,724 [Eric] ... that's really the exciting thing where new jobs are created, new ways of working are created, uh, which is, which is super exciting. 00:31:18,904 --> 00:31:19,064 [John] Yeah. 00:31:20,264 --> 00:31:20,564 [Eric] Okay. 00:31:22,204 --> 00:31:22,364 [Eric] What, 00:31:23,724 --> 00:31:26,284 [Eric] what do we need to ask about ourselves as humans 00:31:27,484 --> 00:31:34,324 [Eric] u- like, interacting with, let's just say, AI employees, or even the way that we, that we think about them? 00:31:34,764 --> 00:31:34,784 [John] Yeah. 00:31:34,784 --> 00:31:36,344 [Eric] What are some of the top things that come to mind? 00:31:37,144 --> 00:31:40,084 [John] So I didn't do this intentionally, but, um, 00:31:41,684 --> 00:31:53,904 [John] yeah. So I'm not gonna claim I did it intentionally. I really love the, like, fictitious game that we kind of started with AI employees of having them be fictitious characters. 00:31:54,204 --> 00:31:54,684 [Eric] Hmm. 00:31:54,684 --> 00:32:03,324 [John] Because at least it's a... 'Cause you can... Everybody can watch TV and understand, "Oh, that's not real. That's people acting and pretending." 00:32:03,784 --> 00:32:04,404 [Eric] Yes. 00:32:04,404 --> 00:32:05,564 [John] From the, uh, very young age- 00:32:05,684 --> 00:32:05,864 [Eric] Mm-hmm 00:32:05,864 --> 00:32:07,004 [John] ... we get that. 00:32:07,064 --> 00:32:07,564 [Eric] Yep. 00:32:07,564 --> 00:32:09,904 [John] Um, so I think that's actually helpful- 00:32:10,564 --> 00:32:10,944 [Eric] Hmm 00:32:10,944 --> 00:32:18,424 [John] ... to anchor it to a known fictitious thing, so when you're interacting with it, it's like, well, you know. 00:32:18,484 --> 00:32:18,664 [Eric] Yeah. 00:32:18,704 --> 00:32:22,124 [John] At, at least you have a starting point of like, this is a fic- fictitious- 00:32:22,184 --> 00:32:22,284 [Eric] Yeah 00:32:22,284 --> 00:32:22,884 [John] ... character. 00:32:22,884 --> 00:32:23,704 [Eric] Like, we name- 00:32:23,824 --> 00:32:23,824 [John] Um- 00:32:23,824 --> 00:32:25,284 [Eric] ... our cars in our family. 00:32:25,704 --> 00:32:25,884 [John] Yeah. 00:32:25,884 --> 00:32:26,404 [Eric] Like, they each- 00:32:26,404 --> 00:32:26,504 [John] Yeah 00:32:26,504 --> 00:32:27,424 [Eric] ... have a normal human name. 00:32:27,484 --> 00:32:28,524 [John] Right, right. 00:32:28,564 --> 00:32:31,364 [Eric] Yeah, which is great. It's an inanimate object. 00:32:31,484 --> 00:32:31,744 [John] Right. 00:32:31,744 --> 00:32:33,804 [Eric] Right? AI is generative and can converse with you- 00:32:33,884 --> 00:32:33,884 [John] Right 00:32:33,884 --> 00:32:35,604 [Eric] ... but it, it's the same concept. 00:32:35,684 --> 00:32:35,864 [John] Right. 00:32:36,884 --> 00:32:46,784 [John] So whatever you can do to build out, maybe, maybe that's a small thing, but to continue to build it, like, just a reminder, we're all playing pretend here, guys. [both laughing] You know? 00:32:46,944 --> 00:32:47,124 [Eric] Yeah. 00:32:48,164 --> 00:32:48,964 [Eric] Yeah, totally. 00:32:49,324 --> 00:32:52,664 [John] And, uh, I, I think that's silly, but actually, like, helpful. 00:32:52,944 --> 00:32:53,404 [Eric] Yeah. 00:32:53,504 --> 00:33:10,724 [John] And I also think there's a practical in having multiple people interacting, um, [lip smack] and building. So there's actually two levels here. Like, one is building, like how, how are we kinda contributing to make this thing better? And one is interacting, like how are we both interacting with the thing? 00:33:11,024 --> 00:33:11,384 [Eric] Yep. 00:33:11,384 --> 00:33:12,804 [John] I think that is helpful. I think 00:33:13,924 --> 00:33:19,544 [John] all of the tooling as it has come out over the last couple years has been extremely individual-focused. 00:33:19,944 --> 00:33:20,544 [Eric] Yep. 00:33:20,544 --> 00:33:21,544 [John] Um, and 00:33:22,664 --> 00:33:24,784 [John] just think in general, that's damaging for people. 00:33:25,204 --> 00:33:25,904 [Eric] Yeah. 00:33:25,904 --> 00:33:42,704 [John] Um, I under- I understand, like, there's some technical reasons it's individually focused, but I, I think it's, I think it's a problem. I, I, [chuckles] I think the... I mean, I literally was talking to somebody this morning of like, I just, "I've had all these ideas for, like, 15 years, and, like, now I feel like I have superpowers, and I can-" 00:33:43,204 --> 00:33:43,284 [Eric] Mm-hmm. 00:33:43,284 --> 00:33:48,484 [John] "... like, do all the things." Um, I mean, some of that's fun and good, but, like, it can have... Like, there's downsides. 00:33:48,844 --> 00:33:49,764 [Eric] Yeah, yeah. 00:33:49,784 --> 00:33:57,304 [John] And then a final kind of negative that also has occurred to me more and more, especially with people in technical roles- 00:33:57,604 --> 00:33:57,844 [Eric] Mm-hmm 00:33:57,844 --> 00:33:59,124 [John] ... like technical leadership roles, 00:34:00,304 --> 00:34:01,784 [John] there is such a pull 00:34:02,824 --> 00:34:06,244 [John] to get back into things that you probably shouldn't be doing. 00:34:06,244 --> 00:34:08,364 [Eric] [laughing] 00:34:08,384 --> 00:34:09,564 [John] And I know you see this. 00:34:10,784 --> 00:34:25,424 [John] I know you see this, 'cause there's a lot of people in technical roles, 'cause if you've ever been in a technical role and then you're a manager, director, CTO, whatever, there's, like, definitely a, like, "Why do they get to do all the fun work, and I have to do all the, like, like, meetings and convincing people of things-" 00:34:25,544 --> 00:34:25,624 [Eric] Mm-hmm. 00:34:25,624 --> 00:34:25,804 [John] "... and, like, 00:34:27,104 --> 00:34:30,524 [John] compliance and, like, like, this job is the worst." Um- 00:34:30,524 --> 00:34:31,304 [Eric] Budgeting. 00:34:31,304 --> 00:34:33,244 [John] Budgeting, [chuckles] exactly. So 00:34:34,264 --> 00:34:39,664 [John] there is something, like, really tricky here, um, and this is more AI in general, like, 00:34:40,763 --> 00:34:43,084 [John] like, uh, AI employees, like, it's adjacent. 00:34:43,143 --> 00:34:43,344 [Eric] Mm-hmm. 00:34:43,424 --> 00:34:52,624 [John] But it's interesting, because there's that pull into, like, getting more involved in the work, which can be good, and absolutely to manage well, you have to know about the work. 00:34:52,644 --> 00:34:52,664 [Eric] Mm-hmm. 00:34:52,664 --> 00:34:55,564 [John] And if the work drastically changed and you don't know how the work works anymore, like- 00:34:55,724 --> 00:34:55,764 [Eric] Yeah 00:34:55,764 --> 00:34:56,744 [John] ... sure, there's a point there. 00:34:57,064 --> 00:34:57,384 [Eric] Yep. 00:34:57,384 --> 00:35:02,264 [John] But I think there's a lot of people that, like, should be managing that aren't- 00:35:02,604 --> 00:35:02,844 [Eric] Hmm 00:35:02,844 --> 00:35:06,724 [John] ... because they can do more than they used to be able to. 00:35:06,784 --> 00:35:07,204 [Eric] Yep. 00:35:07,204 --> 00:35:11,304 [John] And they need... And I'm saying this to myself too, and they need to get back to managing. 00:35:11,344 --> 00:35:12,444 [Eric] Hmm. Yeah. 00:35:12,464 --> 00:35:15,904 [John] And figuring out how to best leverage AI in that job. 00:35:15,964 --> 00:35:20,364 [Eric] Yep. Yeah, for sure. Uh, it also crosses roles too. [chuckles] 00:35:20,364 --> 00:35:20,824 [John] Yeah, yeah. 00:35:20,824 --> 00:35:22,224 [Eric] Which I think is another good thing. 00:35:22,264 --> 00:35:22,504 [John] Right. 00:35:22,504 --> 00:35:27,984 [Eric] Like, "Oh, well, I have an AI employee who can help me do X, Y and Z, which means that I can sort of, you know-" 00:35:28,304 --> 00:35:28,344 [John] Hmm. 00:35:28,344 --> 00:35:30,044 [Eric] "... try, try my hand at design." [laughs] 00:35:30,044 --> 00:35:32,004 [John] I can, I can be pretend... I can be a pretend designer- 00:35:32,124 --> 00:35:32,204 [Eric] Yeah 00:35:32,204 --> 00:35:33,024 [John] ... or marketer. Yeah. 00:35:33,084 --> 00:35:33,524 [Eric] Exactly. 00:35:33,544 --> 00:35:33,644 [John] Sure. 00:35:33,644 --> 00:35:34,264 [Eric] Exactly, yeah. 00:35:34,284 --> 00:35:34,364 [John] Right. 00:35:34,364 --> 00:35:44,584 [Eric] It crosses roles as well, which can be... Again, I think the... I think it goes back to are you gut checking yourself on what you actually want? 00:35:44,644 --> 00:35:45,204 [John] Right. 00:35:45,204 --> 00:35:49,244 [Eric] Right? What, what do you want out of an AI employee? How do you define that, right? 00:35:49,284 --> 00:35:49,504 [John] Yeah. 00:35:49,604 --> 00:35:50,004 [Eric] Um, 00:35:51,044 --> 00:35:54,264 [Eric] and being intentional and drawing boundaries, I think, is really- 00:35:54,364 --> 00:35:54,484 [John] Right 00:35:54,484 --> 00:35:58,424 [Eric] ... really helpful. I think one of the, my big takeaways from this conversation is 00:35:59,924 --> 00:36:05,844 [Eric] just watching the metaphors that I use in my own internal mo- monologue- 00:36:05,964 --> 00:36:06,084 [John] Mm-hmm 00:36:06,084 --> 00:36:09,524 [Eric] ... around AI employees or agents that I build. 00:36:10,144 --> 00:36:10,284 [John] Mm-hmm. 00:36:10,284 --> 00:36:16,744 [Eric] Because I think that, uh, has the potential to shape the way that I think about AI and the way I think about- 00:36:16,804 --> 00:36:16,824 [John] Yeah 00:36:16,824 --> 00:36:17,444 [Eric] ... other people. 00:36:17,484 --> 00:36:17,824 [John] Mm-hmm. 00:36:17,924 --> 00:36:21,844 [Eric] Um, you know, and, and how do I value the people that I work with? 00:36:22,064 --> 00:36:22,564 [John] Yeah. 00:36:22,564 --> 00:36:26,544 [Eric] Am I sort of subconsciously thinking, "Well, anyone can be replaced by AI"? 00:36:26,644 --> 00:36:26,744 [John] Right. 00:36:26,744 --> 00:36:27,104 [Eric] Well, I don't- 00:36:27,304 --> 00:36:27,384 [John] Right 00:36:27,384 --> 00:36:29,784 [Eric] ... think anyone wants to live in a world where that's the case. 00:36:29,944 --> 00:36:30,164 [John] Yeah. Right. 00:36:30,684 --> 00:36:32,744 [Eric] But it's not too hard to get there- 00:36:32,824 --> 00:36:33,144 [John] Yeah 00:36:33,144 --> 00:36:34,244 [Eric] ... uh, in your thinking. 00:36:34,784 --> 00:36:42,904 [John] Yeah. Or evaluating... I mean, something that came up for me, like evaluating, like, evaluating somebody else's work and being like, "I could have just done that with AI and it would have been better than that output." 00:36:42,964 --> 00:36:43,484 [Eric] Right, right. 00:36:44,104 --> 00:36:44,724 [John] Um, yeah. 00:36:44,764 --> 00:36:47,724 [Eric] Which, is that true? I don't know. I think it's easy for us to think that, but- 00:36:47,724 --> 00:36:50,124 [John] I, I think it is, it's probably objectively true occasionally. 00:36:50,284 --> 00:36:50,924 [Eric] Yeah. 00:36:50,924 --> 00:36:53,484 [John] Um, but it's not, like, super helpful- 00:36:54,044 --> 00:36:54,044 [Eric] Yes 00:36:54,044 --> 00:36:55,204 [John] ... to think in that direction. 00:36:55,204 --> 00:37:04,764 [Eric] Yeah, yeah. Yeah, yeah. Uh, I think the other... The last thought that I had is, I think it, I think the per- going back to the personal assistant- 00:37:05,444 --> 00:37:05,684 [John] Mm-hmm 00:37:05,684 --> 00:37:11,324 [Eric] ... because that's so pervasive and it's so easy for us to use AI 00:37:12,464 --> 00:37:26,444 [Eric] in a hundred little different ways that make our work and our life easier on a day-to-day basis, I think it can, it can be easy to underestimate how powerful it can be- 00:37:26,524 --> 00:37:26,944 [John] Yeah 00:37:26,944 --> 00:37:29,064 [Eric] ... uh, with more complex work. 00:37:29,084 --> 00:37:29,964 [John] Right. 00:37:29,964 --> 00:37:35,384 [Eric] And, and especially when you get multiple people involved, it can be super powerful. Um- 00:37:35,424 --> 00:37:36,724 [John] Yeah. 00:37:36,724 --> 00:37:56,124 [Eric] And I also think that's a danger, right? Like, it's a danger to, to sort of let a, you know, AI employee metaphor influence the way we think. Um, but also what is good in that is sort of expanding what we believe AI to be capable of beyond just a personal assistant. 00:37:56,384 --> 00:37:57,924 [John] Right. 00:37:57,924 --> 00:38:10,764 [Eric] Um, you know, which is incredible. I mean, you can, you can do so much with it. So, um, as always, we're landing in a place of, of balance and intentionality, [both laughing] which will probably be a recurring theme on the show. 00:38:12,344 --> 00:38:13,464 [Eric] All right. Well, thanks for joining us- 00:38:13,464 --> 00:38:13,504 [John] Yeah 00:38:13,504 --> 00:38:14,824 [Eric] ... and we'll catch you on the next one. 00:38:14,864 --> 00:38:17,704 [John] Yep. Stay safe out there. [laughs]