Jeff Breunsbach: All right, welcome back to another episode of the Chief Customer Officer Podcast. It is September twenty fourth. Jay, how are we doing? Jay Nathan: Doing great. Yeah, doing wonderful. How are you? Jeff Breunsbach: Good. Or well I guess you've only got one kid left in the house. So is Jay Nathan: One kid left in the house. Yeah. Jeff Breunsbach: she back in is she back in school? Jay Nathan: yeah. Yeah. They're six weeks in now. Is that right? Can that be? Five, six weeks? Jeff Breunsbach: Okay. Nice. What's like AI in high school right now? Are they using AI in high school? Or allowed to? Jay Nathan: they are Jeff Breunsbach: Or is it like a Jay Nathan: yeah, I mean they should be using it for research. Actually something I want to talk to you about today is how to use AI for that kind of stuff. But I think they're they actively check to make sure they're not writing papers with AI and that that kind of thing. which is good, but I I don't think these public schools are doing a great job of incorporating AI into the curriculum in in a way that, you know, shows shows these guys how to use it the right way. So, yeah. Yeah. 'Cause Jeff Breunsbach: Yeah. Would almost become like native to their thinking and like how they're like gonna Jay Nathan: it is already, right? I mean they they use it already natively. They I think they're Jeff Breunsbach: Yeah. Jay Nathan: one of the biggest segments of AI users out there is, you know, the under twenty five crowd, right? It's just part of it unless they've completely rebelled against it, which one of my daughters is not a fan of AI at all because I think maybe she's sort of drank the Kool-Aid of, you know, the environmental impact and all that kind of stuff. So different issue there. So Jeff Breunsbach: that's funny. I I I just saw like a recent chart of again, kind of like penetration of users of AI. I think I sent it to you. Did you look at that? It basically is and Jay Nathan: yeah, yeah, yeah. Jeff Breunsbach: what's it's interesting too is that is the way that they phrased it, and so again, I'm I'm not sure of like I should check the source of like where the research came from, but the way they phrased it was personal AI spend. Like what percentage of Jay Nathan: Yeah. Jeff Breunsbach: people in the world are spending their own dollars on AI? And I'm pretty sure it just eclipsed five percent. and so you just kind of think about like, I don't know, we live in this bubble sometimes, right? Of okay, I'm I'm looking and and thinking I'm so far behind and I'm in AI. And then I see a chart like that, and I'm like, holy cow, like 95% of the population, not that they're not using it, but a large percentage of them aren't even paying for anything themselves or in terms of a subscription to, you know, get these tools and have access and to be able to test. So I just thought that was interesting. Jay Nathan: It it's an interesting chart. I'm looking at it right now. First of all, it came from A16Z, Andreessen Horwitz, so you you think it's probably Jeff Breunsbach: Credible. Jay Nathan: legit. Yeah. But then the title is a little odd to me. Paid consumer AI adoption is small but growing. And then it says in the subtitle, share of US shoppers who purchased AI by age. US shoppers who purchased AI. Like, what does that even? Jeff Breunsbach: So I think they're looking at consumer spend, consumer credit card spend is probably where like they're using that verbiage because I think they're probably looking at consumer almost like how Ramp puts out their stuff, like okay, you know, Jay Nathan: Yeah. Jeff Breunsbach: Claude's the biggest subscription on our credit card. Like I think they probably aggregated credit card data and by age and looked at like, okay, who's buying Claude, who's buying Cursor, who's buying open AI. that's my interpretation of probably how they they did that. Jay Nathan: Yeah, okay. Well, the more I look at this chart, the less I like it. Just because I don't fully understand it. But anyway, it is interesting. Only Jeff Breunsbach: Yeah. Jay Nathan: three percent of US consumers pay for AI out of their own pocket, is what the headline here is. So the Jeff Breunsbach: Yeah. Jay Nathan: the youngest buyers are adopting at four X to rate of the oldest. That's the takeaway. Jeff Breunsbach: All right. That's interesting. Jay Nathan: So well I mean in the the one last thing I'll say about this, my my kids use Snapchat, which I've become a Snapchat user to try to just stay in touch with them better. I hate it, by the way. that's a boomer coming out. But there's AI built into Snapchat. So like they have AI like companions or assistants in in Snap. So and I think at some point you have to pay for it. It's not just free forever. So Jeff Breunsbach: Yeah, that's interesting. Jay Nathan: It's interesting. Interesting times. And then you you probably saw we talked about Instinct in Town. Meta came out with Muse this week. the Yeah. Jeff Breunsbach: Yep. We talked about that a little bit last week. I s I saw the there's like there's already been like I don't know, use you know, ten people have like used it to go save money on a ton of different things. like go shop my insurance around, go but everyone talks about how I guess autonomous it actually is, you know, that you can actually give it a task Jay Nathan: Yeah. Jeff Breunsbach: and it can go run and and figure out the things to do and do it. I just saw another example yesterday since it's connected into like Facebook, you know, directly for the most part. a guy was selling stuff on Facebook marketplace and basically allowed it to negotiate the whole thing and it basically gave it a a price, a max and a min and said, you know, how aggressive it wants it to be and the type of thing, you know, the type of way it wants it to negotiate and like let it go do it. And it sold, you know, a bunch of items and it was basically like, okay, I don't know, you know, it was above average in terms of the price that I wanted. And you know, I didn't have to handle it, it handled itself. It just messaged everybody back and forth like autonomously and sold like five or six things for this guy. Jay Nathan: Facebook Marketplace is a very weird place. The Jeff Breunsbach: Yeah. Jay Nathan: every every time that I've tried to sell something on there, I've regretted it. Like it's so weird. Most of Jeff Breunsbach: I my Jay Nathan: them do not, actually. Jeff Breunsbach: I I still have questions about who uses Facebook. Like I have not logged in. I actually can't log into my account because it's an old email address that I've like lost the login to. So I can't like so my my account's just up there forever. but Christina still uses it, especially from like a mom group perspective. Like there's like very active mom groups that like she trades stuff on, that she borrows things and sells stuff. and she's been selling, we're trying to, you know, we're finally transitioning our guest room into a baby's room for our daughter. And so she's like now on Facebook trying to sell a bunch of stuff. and I'm just like, who's coming to our house? When? Like, did you, you know, like did we verify this person? Yeah. Like that's gonna be a thing. Yeah. Jay Nathan: I know. Like the Facebook killer? Like, yeah. When do they get in here? Like do I need help yeah. Jeff Breunsbach: all right. Well, one of the things I thought was interesting was this company Harvey. So Harvey is a legal platform that has I believe started in the last four to five years. And you know, I would say sells into what seems like maybe the the Fortune 500s of the world and those types of conglomerates that have large legal organizations and essentially is are a platform that allows them to you know standardize contracts more easily, probably do reviews. I I don't know all the features, but I'd imagine, Jay Nathan: Red lines. Yeah. Jeff Breunsbach: you know, does a lot a lot of maybe of the workflow and low level we legal work that would like allow the legals to kind of operate at the top of their profession, right? How can I focus on how can I help legal professionals focus on, I don't know, the tough negotiations versus, you know, a a simple MSA review or something like that. and I I caught a clip of their CEO yesterday talking about their transition maybe from like a software as a service product and maybe like a platform into more of a deployed For deployed engineer, more kind of custom scoping. And now it seems like they're getting into like building their own custom models. and it was just kind of an interesting clip, at least that I saw about and I think you saw something maybe similar or different, but it was interesting because he was talking about, you know, almost like watching this shift in time happen of like, okay, we're kind of this software as service platform selling by seat, realizing pretty quickly that that's going to have a kind of a cap to it, and there's not you know, what's the next level of of kind of usage on top of that? And how do we get to you know, I don't know, less seat-based and more usage based and try and drive like more work and workflows through AI. And so he talked about this transition that they made to now start kind of deploying deploying their customer teams inside of these organizations to figure out what these workflows are. And I guess like one of the unlocks from doing that for themselves and then for some of these early customers was Okay, we're using kind of the standard models that are on the market. It actually looks like this should be something that we start to train ourselves and start to like bring in-house. And so now they have what seems like is their own model and their own training and their own like methodology that they're Jay Nathan: Yep. Right. Jeff Breunsbach: now deploying through that. And so they've basically transitioned from a software as a service platform, I guess, to a I don't even know, an AI company. I don't even know what you call it. I guess a model. Jay Nathan: Yeah, I think it's an AI implementation company is sort of the way I interpreted what he was saying. Because Jeff Breunsbach: Yeah. Jay Nathan: it's like everything, right? You start out by building an app, then you in your, you know, essentially reselling tokens, you know, in the app. H by the way, Harvey, they also sell to I think their primary ICP is l large law firms, or even mid sized law firms. Jeff Breunsbach: Okay. Yep. Jay Nathan: and so it may be internal legal cancel, but also law firms you can operate the whole thing on it. and but I mean if you think of the progression, they probably started by building, you know, the next version of Ironclad, right? Ironclad is a old older SAS tool. I think they've you know probably made the pivot to AI first. If not, God help but So then they they created a SaaS application. Harvey was always AI first, so they're probably reselling tokens, right? And there's probably some, my only point in saying that, there's probably some seat-based plus consumption-based. I don't actually know their go-to-market model, but I have to believe it's a combination of the two. Jeff Breunsbach: Yeah. Jay Nathan: And then they built a harness, right? They built a harness around the LLM for legal. purposes, right? So it had access to, you know, probably the the old think about how much public case law there is out there that you could just consume. So then Jeff Breunsbach: Yeah. Jay Nathan: they took it instead of just having the context available to the LLM, then they started building their own models and deploying those for themselves to cut down on their own costs. So what they've done is they've gone through the entire progression of application to building AI infrastructure, which gives them a ton of knowledge on actually how that's done and how to do it. And so I think what he was saying, the way I interpreted what he was saying in that video is we've learned so much, we're now just going to do that for our our clients as a service Jeff Breunsbach: Yes. Jay Nathan: or as a deployment service. And so that's the Ford Engine deployed engineering model that we've talked so much about. Jeff Breunsbach: Yeah, and I I guess like the part of the clip that I found pretty interesting is he talked very specifically about this agent infrastructure layer and how maybe twelve or eighteen months ago, that's like you said, I think they were doing so many things internally that they've started or maybe it wasn't even that long, maybe it was six to twelve months ago. But he talked about he talked about that very specifically. Like, okay, you know what we came to the conclusion on pretty early was like, okay, if we are in this world where agents are going to operate, like How do we build this entire layer and and tech stack that is for them to engage with, interact, have context, like connect all these tools, but like that Jay Nathan: Yep. Jeff Breunsbach: just doesn't happen overnight. And I still think that that that seems to me to be I've had some conversations the last two weeks with some CS leaders, and that seems to be the I guess the point at which I think there's a lot of a lot of companies getting stuck because each department wants to go deploy agents. But it actually, in my opinion, makes much more sense for the company to centralize some of this agent infrastructure so that you're all operating on the same context, that you're all like, cause it doesn't make sense for me to go build an entire context on just our customers from our customer success lens, right? Like I should probably incorporate support. I should probably incorporate engineering and like what they're using our product for and product and maybe the features they request, right? There's all these layers that like it'd be great if we were all operating from the same. Song sheet. And it seems like I don't Jay Nathan: Yeah. Jeff Breunsbach: know if you've seen this, but it seems like to me that is the that point of like, okay, I can't just go build this for myself and my team. And this seems to be a bigger, like longer horizon initiative for the company. and it I would imagine right now, maybe like the thing that I'm thinking about is, you know, kind of AI forward and forward leaning CEOs are probably looking at that saying, we should probably do that now. We should invest in how do I get this context? How do I get this ontology as you've described it? And how do I Like go drive that. I think about like Shopify as like an example. Like Toby, you know, like went all in on River and like that meant, you know, they were doing this kind of stuff six to nine months ago, which means they have such a Jay Nathan: Yeah. Jeff Breunsbach: a huge advantage on what they're doing now. And now and so then you probably have these CEOs that haven't leaned in on that who are, you know, kind of sitting there thinking, No, let's let's wait, let's figure out, let's, you know, let's see what this is gonna ha like to do. And you I don't know, you feel like they're gonna be passed by. Jay Nathan: Yeah, I I was I actually went back and listened to our episode from last week or two weeks ago, I can't remember which one and when we talked about like y you're never that far behind on this stuff. You can always catch up. It's moving so quickly that everybody's a little bit behind anyway. So Jeff Breunsbach: Yeah. Jay Nathan: I don't I don't necessarily think that people are gonna get passed by, but it sort of depends on what your competitors are doing, right? If if your competitors are finding ways to to reduce cost and use these tools to to drive a more effective go to market, then I think you you are gonna get behind. You know, but I wouldn't compare my company to Shopify, right? Or Jeff Breunsbach: Yeah, yeah. Jay Nathan: you know your company to Shopify. Like it's just two different two different leagues almost. Just fine, Jeff Breunsbach: Yeah. Jay Nathan: right? But but yeah I think you're right. Like getting this infrastructure right It's interesting 'cause we're we're trying to sell right now into multiple different companies to do almost localized agentic work. And it's challenging because they don't have these infrastructure layers set up yet. And you almost have to go build some of the infrastructure layer, like some of the you know, the core you know, central brain for customers, like Salesforce is still a mess. there's multiple tools in the mix, like there is a data warehouse, nobody's quite sure, you know, how to get value out of that. Like there's a lot of pieces that you know, you can't just go build agents yet, right? As a service. You really have to provide the the plumbing layer underneath it to to get the full value out of it. So it's interesting. Jeff Breunsbach: Well this this is like another this kind of leads into another question I was gonna ask you, which is we kind of started this off talking about, you know, town and muse and instinct and all these things that are popping up. what is your it it feels like there's too many things popping up for them all to survive. And it feels like, right, like it it and it I Jay Nathan: Yeah. Jeff Breunsbach: almost I I I have this like fear of like missing out, right? This FOMO of like, Instinct, let me go try that in town. And like you see all these people on Twitter and in and LinkedIn doing it. And in my mind, I'm just like, Jay Nathan: Yeah. Jeff Breunsbach: I like I it yes, I guess like being on the edge of innovation and like trying to find the next thing, I guess like matters to a degree, but it but at this stage, it feels like you have a lot of these people who are making noise in the market and you're just like, okay, that's the same tool, just a different model. It's like, you know, like And and why are you just going to go share bits of information with different tools? So now I have to go manage an instinct and a town and a and a muse. Like this is actually making it harder for me. Like AI was supposed to make everything easier, but now there's so many tools that it's actually making harder. Like it's almost like you're back to the same Jay Nathan: Yeah. Jeff Breunsbach: spot you're in today, which is like, cool. I have a conglomeration of tools. Like, you know, I've got Figma and I've got Salesforce and I have whatever, Pylon, and I've got all these tools, right? And so then, okay, like which tool do I go to? It feels like you're it it's almost like becoming the same problem with all these AI tools that are coming out. And it's like, Jay Nathan: Yeah. Jeff Breunsbach: you know, is it really advantageous for you to go test all of these things like at this stage? or is it better to like, I guess, find something that you know works for you. You can build into your own kind of like day-to-day context of like, okay, this works for me. This is my workflow. Like, okay, let me use this and then like wait for, I don't know, wait for this stuff to I don't know, wait for all this stuff to kind of pop and and come back down. It just feels I don't know. It just feels like there's so Jay Nathan: Yeah. Jeff Breunsbach: much hype on all these things that Jay Nathan: Even even just Jeff Breunsbach: I myself. Jay Nathan: a year ago, even just a year ago there there was a a product I saw, I can't I'm not gonna try to name it 'cause I think I'll use the wrong name, but there was a product and we're like, my gosh, this thing's amazing. It's it's like an agent that reads your inbox and I read the other day that they went out of business already, right? And this is an AI native company. Like they that was what they do. So to your point, these tools are almost indistinguishable from one another. Like talent and instinct, I'm sorry, there are probably people who can tell the difference. I can't tell the difference, right? I mean they they both do the same exact damn thing. Jeff Breunsbach: Yeah. Jay Nathan: they can connect to the same stuff. Like if you go configure So I think what it's gonna end up happening is you're gonna buy into an ecosystem. What like we talked about Facebook, right? Well, the reason that Facebook Muse might be interesting to somebody, or MetaMuse, I should say, is that it is probably the only one of those that can natively integrate with Facebook marketplace because they're not gonna in in interestingly, I just read what what did I read? Amazon blocked Muse from shopping on its site, right? So there you go. The ecosystem taking place. Gets who's got their own personal agent that you can use to shop on the Amazon website. Amazon. And it's it's called Alexa, right? So I I think it's really gonna be we're buying into an ecosystem and it depends on what your needs are As to which ecosystem you're gonna buy into. And you're probably gonna buy into multiple because I'm gonna be an Amazon user. Right? I'm also I'm not gonna be a Facebook user probably, but I'm also going to use other platforms where my data is stored. Google would be a great example, right? Google Jeff Breunsbach: Yeah. Jay Nathan: might end up shipping a personal agent. And guess what? That might make a lot of sense for me to use for my personal life. Now it's gonna suck that I can't use that one for Amazon, but That's just the the trade off, right? So I think these these the it's still a platform war, not not Jeff Breunsbach: Yeah. Jay Nathan: an app war. Jeff Breunsbach: That's interesting. yeah, well then you start to think about okay, what does that mean? Like in terms of the ecosystems, right? And it means who's who's striking the partnerships that you can you know, essentially go have all these agents work together and you know, who's who's basically bringing these things together? a company like Apple comes to mind. Like that would be pretty interesting, you know, to be able to to use just like one that goes and connects to all these other agents and lets them, you know, do whatever they need to. Jay Nathan: Yeah. Well, but Apple's put a a giant stake in the ground in terms of privacy. So they have a dilemma. Is that they're Jeff Breunsbach: Yeah. Jay Nathan: and maybe not a dilemma, but they have a a stance on that, which is they're not gonna just share your data with any agent, but they have an agent too. It's called Siri. And it just got a big upgrade, and so we'll see if it's any better. But like in theory, nobody knows more than you than Apple no knows more about you than Apple or Android if you're using who whoever your phone provider is, nobody knows more about you than them. Jeff Breunsbach: Yeah. That's where your behavior is the most. well Jay Nathan: Mm-hmm. Jeff Breunsbach: cool. what I know you've been digging into or maybe digging into the wrong one. Maybe you've been getting some frustrated with some AI writing and some AI slop. And I'm curious if if you you mentioned that you've been trying to write just more on your own again and I'm curious if you know what you've been finding as you've like gone down that path. Jay Nathan: Yeah, it's funny. I sort of made this decision. I'm like, you know what, I I just can't get it to work. I can't get i and and I think there's a lot of good reasons for that. you know, I've I've tried to have AI like Claude and and Chat GPT to be honest with you, and maybe even Grok at one point, try to help write things for me. 'Cause I wanna post on LinkedIn every day. I just want to try to do it I wanna get get out of some of the work. The reality is I don't think there's getting out of the work. Any getting out of the work. I was listening to Thomas Tomas Tunguz was on the top top line podcast a couple of weeks ago and he was talking about this. Did you listen to this episode? Jeff Breunsbach: It actually. I saw this one. Yeah. Jay Nathan: You remember remember what he said? He's like, Yeah, you know, I mean it was funny 'cause I I saw one of his newsletters and I was like, This is oddly like it sounds like oddly AI for for him. Like he's got a very curated, you know, his his posts are always very tight, they're short, and they're punchy. And I read this thing, I was like, this sounds like garbage. Like what what is he doing? but anyway, been hearing him talk on that podcast, he was like, Yeah, you know, I've I've been using been using AI to write, and you know, what I found is that I'm going back and forth with this thing, and he's like, I think maybe I'm 10% faster at doing it now. Maybe. But not as good. You know, he he basically Jeff Breunsbach: Yeah. Jay Nathan: said like it was it was terrible. So I it ki Jeff Breunsbach: It's like worse writing for 10% efficiency gain. And so you're like, okay, is that really, is that really worth the trade-off? Jay Nathan: Yeah, it well it's not, and I've got a stat I'll share with you that I just read this morning in a minute, but so I was already sort of headed down this path. I was like, just let me let me start writing again because The honest to goodness, like when you write stuff down, when you're forced to put your own words on the page or on the screen, you think through them differently. It's different than reading what an LLM spits out at you and and and trying to interpret that, which is often a big hurdle in and of itself. It's like, what is this thing trying to say to me? The reality is is it's not trying to say anything to you. It's only predicted words, right? That come next. Keep that in mind. That's how LLMs work. They predict the next word. They don't think. So anyway, long story short, I'd started writing for myself again. Then I start to see all these articles pop up. Maybe it's just confirmation bias, but it's like, like article I saw last week, how do you write with an LLM? I shared this with my team. Last week. So I've I've always had two simple rules for my team. Rule number one is you have to use AI. Like there's no going back now. Like we're all in this, we're going AI first. Rule number two is I should never really know that you've used AI. It should never be truly apparent. And I think what's happened is the pendulum has swung. And I would tell you this has happened in my company, it's happened with my own work personally. The pendulum has swung too far toward being able to see that I've used AI and it's become very Jeff Breunsbach: Yes. Jay Nathan: obvious. So I ran across Across this other article last week called How to How to Write With LLM. So rule number one of this, how to write with of this article. Rule number one, you may not use a single word an LLM suggests to you. Right? Jeff Breunsbach: Yeah. Jay Nathan: Break breaking this rule is what gets you in trouble. The reason is frontier models are supernaturally good at selecting pleasant turns of phrase. It's sort of their whole thing. The problem with what models suggest are subtle. Think of it this way: Frontier models are wedged in a mode where everything they write is a magazine headline. Headlines are good, but you'd wonder about someone who wrote an article with dozens of them. So I think that's spot on. I think maybe this article or a comment where I found this article said something about like Like infused with corn syrup, right? Like there's empty calories and everything that an LLM is is outputting, which is so true. and then rule number two, which is a little bit more subtle here, and there's only two in this article, but avoid encouragement. So, like, even if you wrote something terrible and you fed it to an LLM. It would tell you, this is a great start, right? it would it would basically encourage you. So it's it said, you know, here's how this is gonna basically screw you up. Said you're gonna double down on all your first draft impulses, but that's not normally what you do. You'd edit, rethink, and replace paragraphs. Those re I think this is meant to be ironic. Those rethinks are load bearing parts of your voice. Readers won't put their fingers on what's wrong, but they'll sense that you've become artificially flavored. I love that terminology. Jeff Breunsbach: It's a good line. Yeah. Jay Nathan: so, all right, so real practically, there's another article I read this morning called I Don't Want to Read What You Didn't Write. And it's sort of yeah, Jeff Breunsbach: It's a great headline. Jay Nathan: isn't it? I mean, just simple. So and in it, there's a couple of stats. There's this stat by this woman Cynthia Dunlop, who recently shared survey results on reading on reading AI-written articles. The report is called How Developers, this is more of a developer kind of thing, but doesn't matter. How people react to AI-sented blog posts. For the majority of readers, if they think that an article is AI assisted or AI authored, they will stop reading. 78% of people will stop reading. I'm in that category. I'm sure you are too. And furthermore, They will v seventy one percent of them will avoid the author entirely in the future. So you're Jeff Breunsbach: Interesting. That's e because that's like the worst part, right? You're trying to get more you're trying to output more to stay top of mind, to be in front of somebody, to like, you know, try and get your thoughts out. But then you're actually harming yourself in the long term because like people are more likely to turn tune you out, essentially. Jay Nathan: Hundred percent. One hundred percent. And so then you become noise, right? And you you are literally getting the opposite effect of what of what you intended by riding with AI. Jeff Breunsbach: so there is an Instagram a Instagram handle that like is a I guess like a ploy on Chat GPT and the guy it's like it is like so funny to me and my friends because like to your point, like the guy basically just like first of all is always drinking water at the beginning for no reason, but like he's guzzling water and then like it'll be like somebody will be like, Hey man, like what's going on? And he's supposed to act as if he's Chat GPT and he'll be like, What a great question. You know, or like he'll have all these things that you're like, my god, that is like what the LLM does to you. You know, it's like, you're so magical about what you just did. Like, yes, that is correct. Like, so it just reminded me of like what you're you're saying. I Jay Nathan: Yeah. So don't take the bait on the encouragement. Jeff Breunsbach: I have I have found this, I've gone down the same path too. Like I automated a bunch of stuff. I tried to get writing, even I mean, we've done it for our newsletter, I've done it for LinkedIn, and every time I've I've walked away disappointed. So I've just stopped doing it, which means my output has gone down to like zero. but I think where I'm landing, and maybe this is where you were getting to as well, is like, I have curiosities I want to write. and I guess like the thing I was trying to optimize on the writing part, but that's like the part I enjoy. Like I should actually optimize on the other parts of the process, right? So okay, can I go? Hey, I have a Jay Nathan: There you go. Jeff Breunsbach: curiosity, I have a question, I have a hypothesis. Can you go help me bring the research back? And again, I don't want it to summarize the research. I want it to bring back, sure, it can it can give me some useful hints. Okay, hey, hey, I thought this was interesting, but like bring me back the actual articles you're finding. Give like basically give me all this input to help me figure out what I want to write. And that like should allow me to do it better. and I think that's where I'm getting to is like, okay, it's it's much better for me to use it from like a discovery and like a sourcing standpoint. Okay, give me some of the things that you're finding about this hypothesis I have, give me. articles on both sides, you know, let me kind of like read for myself and figure out what, you what what my opinion is or go find, Jay Nathan: Yep. Jeff Breunsbach: you know, this interesting person and what they've done in their careers. and then, you know, I should be taking that to go write, you know, what I think is interesting based on the the kind of the content or hypothesis that I have. Jay Nathan: Yeah, use it as an input, use it as an editor. That that is the prevailing recommendations. This isn't just complaining these articles that I'm reading. It's like, Jeff Breunsbach: Yeah. Jay Nathan: hey, here's what I'm doing instead. And exactly what you described, right? It's like, okay, I'm gonna write a paragraph, then I'm gonna feed it to the LLM. Ask it for feedback, ask it for opposing views, asking it to go support my statement with data that it can find. Like use it for what it's for what it's good at. But but actually, you know actually what one really interesting idea here too was as you're writing, put put instructions for the AI reviewer in your writing. Right? Like, hey, here's a here's a reference to something I can't quite remember. Go look this up when you edit this paragraph for me. Right? Jeff Breunsbach: Yeah. Jay Nathan: I mean, do it, do that, right? And then remove it later. Jeff Breunsbach: Yeah. Jay Nathan: spelling, grammar, simplifying sentences, but Make it simplify stuff that you've already written, you know. This is really Jeff Breunsbach: Yeah, that's interesting. Jay Nathan: particularly so think I mean, we talk a lot about writing from just a content perspective, like public content, almost like a marketing standpoint, but I'm seeing it in presentations that we're preparing for our clients now too. And so yesterday we had a review. We have a big client, you know, executive readout today, and and I was going through it with my team yesterday and you know the titles of these slides, I was like, guys, these these titles don't make any sense. I could tell that you just left Claude, whatever Claude produced in there, and it's just not good enough, right? It's it's it's repeating everything that's down below and just like a few meaningless, pithy words. So what you make, you need to edit the title every time. And so what we've what we actually landed on yesterday was okay, like we're gonna get through this presentation. But the next one is gonna have like there's gonna be, here's the way we're gonna go present. It's gonna have three these three sections. You're gonna write them, right? Of course, use AI to get the ideas. Like, remind Jeff Breunsbach: Yeah. Jay Nathan: me what we did because our our agent has access to every call we've done with that client over the past six months, every decision we've made, every status report, every document we've reproduced, like it's got all of it, right? You're gonna forget some of that as a human. Go leverage the tool to go scan through all that, get the right ideas out, and then. I want you to write what goes especially if we're gonna call it an executive readout, right? Jeff Breunsbach: it's interesting you mentioned that too though. So, you know, Claude now is like native in Slack. I forget what they call it, the Claude Jay Nathan: Claud tag. Jeff Breunsbach: tag. and you've started I've I've started to see some people on our team use it like in a way of like in terms of like writing, you know, or like bringing something from maybe Claude Cowork and like posting it in there directly, because it'll say from at Claude at the very bottom. And you can start to see some people say have like started to notice, and people have said, like. Interesting, seeing that's like directly from Claude. So like to almost to your point, like starting to see this in our own Slack channels where it's kind of like, I know you didn't write that. And therefore, like am Jay Nathan: Mm-hmm. Jeff Breunsbach: I less likely to act on it or more likely to act on it? Or like, you know, like can I trust what's in there? Like have you verified it? Or have you just like you said, have you just put it in a a post, you know, to Claude Cowork and it's just shot over something? So it's interesting. I've just seen our I've I've just seen the curiosity and the question start to pop up right now. And to your point, like I've seen it like get into you know, slides and other things that we've been we've been doing. The one thing that I've used it for recently that I think is actually very good from a writing standpoint, is I have taken our knowledge or mate mainly our API documentation, and I have plugged that into cloud design. I've used our brand guidelines and I've used an example template and I've built two pagers for our customers on certain features, on certain outputs. And like you said, to me, that's not like it's not like going to write anything that's like brand new. It's using a source and it's trying to codify it into a template that has our branding and looks nice. And I've just started to show that around. And I think, I think people are starting to realize, okay, like we should be doing that much more because like, you know, I don't know, we don't we we should, I guess, like templatize that stuff in a way that like doesn't matter who you are in the company, just go use the brand guidelines, use the template, plug in the topic that you need for your customer. And now you've got a resource that you know, again, uses the verifiable source and comes out with like a like I showed it to like our product marketers and our like our product people and they're just kind of like, yeah, like you can go use that. That's fine. And it's like cool. Okay. Like, but like Jay Nathan: It's cool. Yeah. Jeff Breunsbach: the but to your point, like I'm not asking it to write anything new. I'm actually asking it maybe to pare it down and put it into a format. But like now I've got a beautifully branded format that like, you know, I can put in front of a customer and and is something that can be a customer facing asset. And that took me literally seven minutes versus like having to wait, you know Weeks for something to come out. And and so I think we're trying to find our footing there too. How do we create these customer-facing assets that we can all leverage? Again, question still the questions remain, right? Okay, where's the central repository so that nobody duplicates the same thing? And how do we optimize there? But cool. Well, yeah, that's interesting. I'm I'm kind of going down the same path. So all right, I think that was a fun episode. you know, we talked through a little bit of Harvey, I think late hearing. more and more about I think some of the stuff that you guys are working on from agent infrastructure too is is is being super interesting. I think it's like, you know, the the wave of where people are going right now and then how to get slop out of AI ready. So I like it. Jay Nathan: Just learning as we go. Every day. New challenge. Alright, ma'am. Jeff Breunsbach: Yep. All right. Well sounds good. We'll see you next time. Jay Nathan: See you next time.