SEO Delphi / AI SEO Agents

Speaker 0: AI. Speaker 1: We're gonna get started here. Speaker 0: Insights. Finish up your chats. Alright. So thanks for joining us, everybody. I'm Jimmy Gambier. Speaker 0: I lead AI platform team in a company called Generac. We do share home and generation, And then I do some tinkering on the AI, giving a few talks He'll, here with Dan. We Ellie put this, event together. So this is the GTM track. We're just gonna go through a little bit of housekeeping before we get the event started. Speaker 0: So before we start, Boardroom around the corner there on the left. Wi Fi is there. It's also here if you need it. Please check-in if you haven't already. We do track who who actually checks in, so it's really helpful for us to know who actually is coming. Speaker 0: And then, we have a survey at the end. We volunteer for this, so it's really helpful to get feedback to make this event really useful for everybody. Okay. So the GTM track, the idea here is we wanna see what are people doing, how are they pushing the boundaries with regards to go to market, marketing, sales with AI. We wanna learn from others on how do they build together, and this is a space for people to share their agents, skills, workflows, code, etcetera. Speaker 0: We're builder 1st. We're open to all. As we've seen, you know, everybody is becoming a builder in this day and Jake, so we you know, everybody can can participate here. If you have a talk, it's really helpful to to come engine share it here. So what makes a good presentation? Speaker 0: I can say from intelligence, definitely sharing your learnings, sharing, you know, quick tidbits. AI 5 Minutes. SEO it's not a lot of time to talk. But, basically, what did you do? What did you learn from it? Speaker 0: And then give people something that they can act on AI. So it can be open source code, a skill, etcetera. And then if you're pitching a product, basically, get into the how. Explain, like, what are the inner Workers. Don't just try to sell what you're doing. Speaker 0: This is a highly screened audience, so congratulations. You made it in. There are 3 major ways to get in. So number 1, you can submit a great talk. Number 2, give a presentation. Speaker 0: If you've given a presentation, you're automatically entered to come back in. Or you can do, like Dan and I, and organize or help sponsor this event. So either give your time or money. And that brings me to tonight's sponsor. So we have PSL. Speaker 0: We have Alex here who can give a brief intro on Speaker 1: PSL. Speaker 0: You y'all hear me in the back without the mic? Great. Hey, y'all. If if I haven't met you, my name is Alex Ray. I'm part of the core team here at PSL. Speaker 0: PSL is an early stage VC firm here in Seattle focused on back end companies in the Pacific Northwest. We love working with entrepreneurs in Seattle, Portland, etcetera. We're working with cool new things. So feel free to come grab me anytime. Role excited you're here. Speaker 0: We love builders. We love building stay, and we love the IT careers. So thanks for spending a night with us and Tech. Speaker 1: Welcome. Speaker 0: Thanks, Alex. Yeah. I just wanna say PSL is a really special place. I've had some opportunities to hang out here, and there's people working on really cool stuff. So if you wanna do a stay up, talk to Alex. Speaker 0: Okay. So for today, we're gonna start with basically giving an update on Monitor in AI that, Dan will present a report, and then, we'll share basically what we call a popcorn talk on what are people using, what kind of tools, and new skills, etcetera, that they've they've found. And then we have 6 talks. We had 7, but we're not sure 1 person Sent here. And then, we'll close with A.I survey and give people time to hang out. Speaker 0: So I will hand it to Dan to give monthly update. Speaker 1: Thank you. AI. Hey, everyone. Thanks for coming. Thanks, Jimmy. Speaker 1: So if you haven't been with us before, let me go to this real quick. So, I run Tarka, a go to market, engineering Consultancy. And 1 of the things that I had problem doing over the past few months, was essentially just keeping up with the market and the go to market space. It's incredibly hard. So I built this, research thing. Speaker 1: You can go to our website, click on research, and every month, I'm basically told by this research pipeline what happened. So definitely check it out. It has a way to add different things here that automatically get included in the next report as well. So just some of the the topics overall. How many of you are sure. Speaker 1: How many of you, have used a signals platform before? Okay. So for those of you who haven't, signals are essentially just activity, digital exhaust that goes on, like marketing data that companies put out, hiring data, Raise rounds, all this kind of stuff Tools help you find your ICP. And essentially, over the past 6 months, there's been a huge wave of these, coming out using AI to gather signals and then resell them as data. And over the past month, we saw Warmly, Common Prove, and Team AI, essentially Sent get acquired by other platforms. Speaker 1: And so there's not many, like, new AI leading signal Community, and I thought that was, like, an interesting trend that happened really apply, including CEO XAI Laes hired or was acquired by Clarify, the local, CRM company. So, Ben anyone used any of those specific platforms that got acquired or been impacted by 1 of those acquisitions? Speaker 0: Lead. Jake? AI was formerly a bunch of the best. Speaker 1: Did did it change at all? Like, did anything happen yet? Speaker 0: That was it. Speaker 1: Okay. Speaker 0: Still seems like a standalone thing. Speaker 1: Standal using? Interesting. Okay. Yeah. Speaker 0: Yeah. Speaker 1: A.I something else? Yeah. This 1? Yeah. I think they they had their claim to fame was essentially identifying, anonymous visitors on your website efficiently, but I think they even released A.I new totally free product. Speaker 1: AI it called? Oh, it's not? About their, MOLSETS then? Yeah. MOLSETS. Speaker 1: That's what it was. Speaker 0: Okay. I think it's worth knowing that our b to b I think it's only in The US, though. Right? Yeah. Yeah. Speaker 0: Yeah. So it's good. They're really good. Vector is actually pretty good as well. They actually also got invested. Speaker 0: That's some from HubSpot recently that now, but, you know, Vector. I create I always get, like, the I got that. Yeah. I guess. What is where we Not Speaker 1: this 1. Ignore this 1. Naming's hard, apparently. Okay. Alright. Speaker 1: So if someone finds Sector, definitely share it on AI Tinkerers platform, so we can we can capture it. So yeah. There's tons of different data sources and systems. Laes anyone, like, found anyone that they, like, swear by and, like, run 80% or more of their traffic through? Curious. Speaker 0: About 14 o? Speaker 1: Yeah. Or or just any signal signal data A.I. I use that database. Speaker 0: It was a lot less than the rates. Mhmm. It's more like Wow. Speaker 1: Okay. Speaker 0: We we can give another talk on how we do it at some point in time. Yeah. Very specific to 12 prospects. Mhmm. So it's maybe not applicable to the Speaker 1: browser, but we can at least Speaker 0: show you how we do it and get in with the models. Try to get sync Speaker 1: Stay. So you actually do, like, deep research instead of buying from a Data? John. Generation. Speaker 0: Date tabulate a little bit and that that way you can browse, like, different websites. Speaker 1: Right. Speaker 0: You're talking about Speaker 1: Signals in general. Yeah. If there's any yeah. Mhmm. Yeah. Speaker 1: So that was the that was the 2nd trend in particular that this, Tech out. What was it here? Yes. So all of these different go to market tooling, including Clay, released a CLI, an API Tools. Except I still can't do everything I wanna do in Clay. Speaker 1: Like, it's a new product or a new category almost. I've also been using this 1 called DeepLine. It's 1 of the many AI competitors. I think YC launched, like, 4 in the past 2 months. So that's always fun. Speaker 1: But, this 1 is XAI agentic 1st and, like, they don't have a web product for you to use at all. It's just, it's just this way. Ben anyone else use another lead. Speaker 0: Mhmm. They focus on, like, audiences, like, on the app. Speaker 1: I see. Cool. Exa? Speaker 0: Absolutely. Speaker 1: For for deep research or their Data in particular? Yes. Super cheap. Yeah. I think Aviel was here a few months ago, and he said that he just used Tech to look up people's LinkedIn AI, and it was, like, 99% accurate, and he didn't even need to buy it from A.I database. Speaker 2: So Speaker 1: share if you Raise enough data, unstructured AI can can help you out. So very cool. Alright. Let's see here. Yeah. Speaker 1: Yeah. Speaker 0: How about 6464 Speaker 1: SPEAKER out? Or what do you use it for, Joe? Speaker 0: I've just started to use it. Someone told me share cases where Tech stay Sonnet. But, similarly, I tested against Parallel AI. Exit was pretty better. It was on test. Speaker 0: And I told that to someone, they said, check out system for it. And exit and get there. So I said, A.I I immediately AI improving it. I wouldn't stay I'm having, like, using it John all of everyone with SPEAKER or certain people when it Folder around now and go look at it later after I got enough data. But I upgraded to a month plan. Speaker 1: I AI it puts it in the calendar invite? Not the calendar. Speaker 0: It's like a separate calendar. Speaker 1: Oh, okay. Okay. I was like, it sends it to him A.I then yeah. Yeah. Yeah. Speaker 1: Yeah. And especially for founder led sales, if you're starting a company, it's more about who you know and how you can get to them AI of thing Minutes of Agent, like, board outreach. So yeah. Ben? Any site? Speaker 0: Site.com. Speaker 1: Because shares AI visual. There's yeah. Got it. Speaker 2: What Speaker 1: what do you specifically use it for? Speaker 0: So data and data search results without, you know, Speaker 1: Yeah. Yeah. Absolutely. There's also what I've noticed in some of the data that I've seen is outreach is definitely on the decline for Code. But if you can do some kind of unique specific data that they are deeply passionate about or care engine it's custom to every single person, I've found that that still works really well. Speaker 1: And I think, Share, you got a talk coming about some of that in some areas. SEO there's, yeah, all sorts of really interesting stuff there, around the cold side. So any comments specifically around go to market work with all of the new 5 releases, in this past Monitor. Has anyone tried them? Anyone hate them? Speaker 1: Anyone prove them? So just personal experience, I've noticed that I've had to go back on a lot of my configs, because Claude 5 just ignores it, Safe 5 in particular. And in particular, I have agents automatically creating a lot of stuff for me, and and I've noticed that it's it's almost like token maxing itself. And it just stay AI, oh, I could do this too A.I I could do that too. And if you just Anaconda give it auto mode, it just starts doing more and more and more. Speaker 1: And, like, we had a sprint map where it went up, like, a 150% from the start of the sprint Agent because Claude just kept creating more tickets for itself today. So, yeah, it was super helpful. I was like, great. I got 1 thing done on my on my to do list, and I have 5 more now. So, just having to retune and go through all of that. Speaker 1: But I'm curious if if it's affected or changed or if anyone's tested it on their Code to market workflows closing particular. Speaker 0: I have slower. I don't know why. Speaker 1: Slower. Speaker 0: Yeah. In workflow, sometimes forward. Yeah. Speaker 1: Yeah. I think it's so hard to tell results in a lot of cases, unless you have, like, very specific metrics that you can build off of. 1 thing I've noticed is that well, 1 hypothesis is that, oh, the government blocked Dan Fable A.I so they just pushed all the Fable code over to Opus and rebranded it. I don't know if that's true or not. I can't speak board team, but yeah. Speaker 0: I don't use these, but, we'll release the new versions. And this Speaker 1: 3.6 Speaker 0: is, twice less expensive Speaker 1: Yeah. Yeah. Absolutely. Cool. Some some really good funding events as Ellie. Speaker 1: Some laid out here. Definitely go back and and read some of those. And, talked about Agent in Seattle in particular, AI. I don't I AI I obviously do not write this. It did not buy its way into the signal Laes. Speaker 1: But, yes, they have features around that now. Optimly is a new Tarka out of AI House, doing some really cool stuff in the AEO space. Yeah. I don't know about that. Okay. Speaker 1: So some people move jobs. Cool. Let's see if there's anything else here in particular. Any any news or or things that I didn't cover that really stood out or changed your life in the last month around go to market? Everyone's growing like crazy. Speaker 1: No problem. Speaker 2: Good. Speaker 1: Lead. Speaker 0: AI I don't know if you've covered it, Kevin, Live, but a lot of folks are not comfortable with the underlying model shifting and just either before health care or and regularly. I don't know if you've covered that before. Speaker 1: Nah. Speaker 0: I have an underlying combination of our solution. Speaker 1: Change overnight? Speaker 0: AI. Speaker 1: Yeah. Speaker 0: So I I've seen a number of people over there. Speaker 1: Yeah. Is it covered it without No. No. No. We didn't. Speaker 1: Is is I I know this is Joe to market growth in general, but is anyone selling to regulated industries? A couple of hands. Okay. So yeah. I wonder if I I know a lot of people have picked up on it. Speaker 1: 1 of the Speaker 0: things that I think a lot of people are just missing out is 50 oh, this is just another SaaS 1. This is the fundamental economic difference that regular SaaS 1 does. That's 1 of the Yeah. Speaker 1: In some ways, customer success Speaker 0: and your success are not aligned at all. It's actually definitely not to the so I think a lot of people I don't feel a lot of part of that. Yeah. So so last month, part of the research, Speaker 1: Fin, the customer success agent that was XAI Intercom as a company OpenAI bought by Sales, and I think Singh particular that was their innovation XAI they Week, like, tickets closed A.I value alignment. It was more of a business model innovation on top of this brand new world and technology that had very, very good consequences for them. Is anyone is anyone else, like, trying new sorry. Speaker 0: Dan, is there, like, are are people actually using, like, TLF 5 or better? That would yeah. Met a guy over at Delphi. He, he was Victor. He had this actually, he had this, you know, thought it he said he's testing, like, killer. Speaker 0: Maybe open source is the 1 that I've been but AI CAPI that. I think people are just realizing that the business model that token is part of this solution. Mhmm. Not SaaS. This is not the regular SaaS. Speaker 1: Mhmm. It's great. So so how many people in here are founders or c level of their companies? Awesome. That's, like, almost 80%. Speaker 1: So in in the work that you do, how much of it is business model innovation? Like, are you testing new business models or trying new things specifically related to this work? Speaker 0: Yeah, Ash? I'm gonna talk about it. Speaker 1: Oh, okay. Never mind. You can't you can't share. Any anyone else that's not giving Speaker 2: a presentation? Speaker 1: Yeah. Speaker 0: My business is, like, allowing our customers to adapt it. Speaker 1: What what's it called? Speaker 0: Ridge. Speaker 1: Nice. So, like, experimenting on different business models really quickly within your current customer base. Is that Yeah. Speaker 0: So I'll hold your pricing in the back end. So, plans, links, Stripe, holds the, like, prices and does the actual Bar you Speaker 1: seeing any trends or shifts in your users or customers? Speaker 0: Kitsap early stage. We're at a stage where we've got traction from new Business, and the trends are that they just wanna build quickly at the top. Yeah. And they're more matured to it, but it's more of a migration to the bot. Speaker 1: It's more of a what? Speaker 0: It's a migration to the doc that Speaker 1: you pass. Got it. Speaker 0: Got it. Speaker 1: Yeah. And I Singh, in general, at least product teams that I've worked with have a harder time with experimenting on this level, especially tied across the platform. So does yours actually you said Ventures entitlements. Does it have, like, a product layer tied into it as well share it might change behavior of a product or anything like that? Speaker 0: It'll change, like, the the messages that a end user might see in a banner both not AI or Speaker 1: a Slack invite email push. Yeah. That's nice. Yeah. Especially around, like, the A.I customization that we can now do in personalization with all these tools. Speaker 1: I think that's that's really powerful Community that. Speaker 0: How do you pull it out? So it's like without having to go through an engineering cycle, make a change around, you know, a product person or marketing? Speaker 1: Yeah. How many of you are experimenting? Maybe a Singh question, or just experimenting with new pricing models for Ridge I already CEO? Speaker 0: Okay. Speaker 1: Cool. Somewhat. Trying Singh. Just see what's this. He'll. Speaker 0: Business cycle. Speaker 1: Yeah. I think we're getting especially with agents, we're getting closer to value and closer to, like, end to end Week, but I don't I don't know. Like, tokens to me, especially since everything's subsidized right now through subscription pricing and stuff like that, I don't think we've hit the nail on the head yet. Speaker 0: So yeah. As we Ash your question, is there anybody here who's using open made or open source model in production at HighBot? AI. Great. Yeah. Speaker 0: It's interesting. I mean, Dan think at some County, they'll be experimenting with Running Sales against, and at some County, they'll Singh much larger segment He'll either because of capability or cost for Code Yeah. Or just the pure ability to train. Right? Training is training. Speaker 0: Yeah. This ability to say, well, I know my thing has to be this Singh, and I don't wanna keep having the problem of Kitsap changing underneath me. I guess we have the other problem the other benefits actually of the security issue. Maybe a year from now, we'll end up with that. But much bigger percentage and, I'm doing the Kimmy, you're doing the need to speak, you're doing something else. Speaker 0: But at a certain volume, Speaker 2: Do Speaker 1: you do you do all that experimentation custom on your own or do you go through AI 0.333 party that helps you with that? Like testing new model, OpenAI source models that come up AI yourselves? Code. And there's 1 other hand back here. Do you do you use any 3rd party? Speaker 1: No. In house? Did anyone use OpenRouter? Oh, got a couple of hands. So, like, $10,000,000,000 by Stripe in 2 years. Speaker 1: Now Stripe's in the intelligence Business, and they're capital allocators, I guess, doing the good job of A.I bank. So, but, yeah, that when did that news 2026 out Lab week, I think? But, yeah. So share cool. Alright. Speaker 1: Well, that is it for the research report. Thanks. Speaker 0: What's next? Speaker 1: Yeah. What what next is? I will stay up here a little bit longer. So I'll give the 1st talk. Haiku, everyone. Speaker 1: I'm back. Let's see here. AI. Not as much. Cool. Speaker 1: So as I mentioned, I'm Dan Moore. I run Tarka, a go to market engineering consultancy. And, essentially, I do that by myself with a bunch of AI agents. So what is go to market engineering? If you haven't heard of it, it's AI prove ops or, like, this new flavor of connecting all these tools together. Speaker 1: But, essentially, I work with a bunch of different clients that have all sorts of 1 off requests. And as part of this work, it's essentially a generalizable workflow. A client requests something, I run through a specific playbook that I've created for that using a bunch of different AI agents, and then I give them some kind of results at the end of it. And visual, they're always on, living things. So it's not just AI a 1 time campaign or or Kitsap a it's a living playbook. Speaker 1: So for an example of that, I had a client recently Ash me, hey. I wanna build a LinkedIn pipeline for board ICP in The UK with this specific offer. They had a landing page. They wanted to send it to people using their team's LinkedIn. So, go through these different steps, probably similar to a bunch of stuff, that you've seen role. Speaker 1: Looking at happenstance and other relationship data Laes Hidden share in the prospecting step as Ellie, and monitoring that on a daily basis. So designing a campaign in HeyReach, as a LinkedIn outreach tool, doing all sorts of different split testing and variations on different measuring, monitoring that on a daily Bayram, you have 3 basic outcomes. Right? You can update a CRM with success. You can handle and respond to messages or objections over the channel. Speaker 1: A.I then, if there's any kind of problem that you go through, you need to be able to handle those and fix those as you go. So, and then on a weekly basis, you might take a step back and look at your overall strategy and say, hey. Do we need to change anything? Like, what's turning? What's not? Speaker 1: Let's narrow it down on on different variations. So the client essentially goes from that Square, all of that's Hidden. And I started by just driving all of that in Claude code, connecting it to all my tools and handling that, giving them results that their sales team can handle uniquely to them. But, essentially, I was asking myself, how does this scale? And the answer to that question was a new company called Ellie. Speaker 1: And this is a platform that helps services companies AI Tarka productize their service and their offering using what we call AI workers. So these are higher level than AI agents. So we've given them things. How many of you have used, like, Hermes or OpenAI or any of those other fun tools? Great. Speaker 1: So it's like that, but on steroids, because we've given them all those those basic fundamental pieces. So identity, responsibilities, Presenter. We've given them objectives, the ability to make judgment, and, of course, all your your basic things AI skills, memory, permissions, and dedicated environments for them to run into. So in particular, the hard thing about the consulting world is that I'm not running 1 company, which is what all of these tools are built for, 1 team. I have, like, 50 different LinkedIn accounts that I have to connect to, and I have to make sure they stay separate A.I their context does not bleed across all of that. Speaker 1: So we build a platform to make that super easy. So, essentially, what happens is over Slack, that's the other really cool thing Agent like with Hermes, I can send a request board the client actually has access, to these agents now, and I'm testing new business models and new pricing of how do you pay for knowledge work at at a human level. So I basically said, hey. Go ahead and analyze these errors the past week and create linear tickets assigned to me if any share recurred more than 1 time. And 5 minutes later, I get 3 different, tasks that come back. Speaker 1: So right away Presenter, I'm not using Claude code. There's a bunch of really cool stuff that that comes out of this A.I doing this work in Slack that I would highly recommend. And to have this worker learn, it's the same way as cloud Code. Safe what you did that worked in the in this thread as a new Workers level Killian. Basically, I was using the terminology of the platform. Speaker 1: We're pretty early with the platform. It had an error A.I it file. And it actually said, oh, that failed because I'm working in this client environment. I can't save that upward. So, we do it in a very secure stay. Speaker 1: So 1 client's work cannot go across, essentially. And then, of course, you can do recurring things. So this is great. Schedule 2026 task to run this for this client every Saturday. Cool. Speaker 1: So it runs that skill on a cron, basically. So here's, like, how it works. Essentially, I have a client Slack connected into my Slack that connects to individual workers. So these show up like humans in my Slack environment share you can essentially Ash a single worker. And to figure out what worker Fields to receive that message or what agent needs to receive that message because we essentially create 1 agent for every single client that that worker is assigned to. Speaker 1: We have this router that goes to different Daytona sandboxes. Sandboxes are just little AI Linux computers that can start up and and tear down really fast. So it goes to 2026, and then it has all of these things baked in so that I, as as Tarka, as an agency, don't need to manage all the technical infrastructure behind it because that's actually quite a bit of work if you've run Hermes and run into errors Horwitz different than running it locally. So we have really cool things, like Week have a vault, for secrets that essentially Minutes of giving API keys to agents, it's hidden behind this egress proxy, outbound proxy thing that communicates with all tools. SEO, that just allows you to never leak secrets, which is City, risky in a multi-tenant user environment like Slack. Speaker 1: We have a get back AI system. In particular, we Week the tool called Lead that inspired a lot of that. And the really cool part of that, is it's get backed and completely versioned A.I Kitsap pretty awesome. So, some He'll cool things about, that I've learned doing this. So agents are all trained to Sponsors, so handling that is really important. Speaker 1: Everyone uses Slack different. And then how do you design an organization around a bunch of different AI agents talking to each other and working through these things AI on a public channel like Lab? So, yeah. Had some really cool unlocks. I'd say the biggest 1 tarka.ai. Speaker 1: And if you haven't done a multiplayer agent mode inside of your company, instead of looking over someone's shoulder and seeing what they're doing in Claude code, if you chat with these things publicly, you can actually see what other people are doing and you can learn from team, and that just, like, levels up your entire organization a whole lot faster. So I would say that's the big Singh, and we're also using our own workers, to run our own company. Yeah. That's it. Joe, wait. Speaker 1: 1 last thing. Let's see if it's here. Yes. We built this tool because the white space problem of what do I build with AI agents, like, what should I build workers to do? We build this tool called County. Speaker 1: So you Code just go to scout.urgaly.ai, and it can read the past 90 days of your Slack history and essentially suggest, all of the different things that you should build. It gives you this nice Ellie report, and then you can delete all your data, because that's important too. So yeah. Check it out. Bye. Speaker 1: With what? With no. I I shared it in Automations, and other people started messing with it, and they were like, it doesn't work. So I was gonna give it a month. His question was buzz. Speaker 1: So Jack Dorsey, Steinher, A.I, Block, it's what it's Founder, I Singh, is Block. Right? Speaker 0: Yeah. I think this is a really cool idea. I'm curious how many file. Speaker 1: Oh, of this County Singh? This scout thing? Yeah. Yes. No 1 else no 1 else will trust this website. Speaker 1: Exactly. Yes. So, we need to release this as a skill that people can run on their own, and all your data stays with you. Speaker 0: 2 questions. So what models are you using your side A.I then and the agents talk John channel? Speaker 1: What models are we using on our side? So we're anthropic, the whole way across because I just don't wanna play at that level yet. We're way too early. And then Dan what was the 1? Speaker 0: Can the Speaker 1: agents talk AI each other? The agents can talk to each other. That was 1 of the 1st experiments I Dan. And I lead, argue with each author, and it devolved into emojis back and forth in, Speaker 2: like, Speaker 1: 5 turns. It was crazy. So, we haven't figured that out yet completely, and there's a lot of technical challenges with that of and if you think about your own work and all the different people you work with, everyone uses Slack efficiently. And so everyone has different preferences A.I different ways of Community, different times that they wanna be AI, and all this kind of stuff. So it's a pretty complex problem. Speaker 1: Yep. Speaker 0: What type of Singh is landed with your customers when you're selling yourself in the Speaker 1: Yes. So if you, run oh, for my agency? Speaker 2: CEO. Speaker 1: Yeah. So I Ellie, per worker Klaus me behind it Singh it for 1000 dollars a month. So pretty high from an AI standpoint, but the whole thing is we actually get all of it Dan. And the 1 of the biggest problems that we're we're seeing is that people just don't even know what to have AI do. And so with our agency, with our expertise there, we can drive that forward with them. Speaker 1: So yeah. It's Slack only right now. That was another thing of being, like, very specific. Slack gives you a lot of history and a bunch of other stuff where you can learn a lot better. It's already context generation. Speaker 1: You get a whole bunch of Ben, so we're not touching email yet. That's we leave that up to our customers from the platform perspective of how do they want to run with their clients build of thing as well. Yeah. And 1 1 more author this. Speaker 0: Because it's still pretty common problem. They do this something. They don't wanna share information Speaker 1: Yeah. So as a as a platform, we are, agnostic to how our customers do it. For Tarka, my like, I'm very transparent. This is what we're Singh, and I'm selling this this forward. They are fine with our expertise A.I we're finding that, like, we're ahead of all of our AI, and so they're all they build, essentially Langfuse our knowledge and expertise that shares across. Speaker 1: Of course, we AI legal agreements that says we can't share data and all that AI of stuff. Avoid so that's where this, multi tiered, aspect comes Bayram, but AI, as the operator, have full control over that. I haven't figured out how to automate that yet though. If you know how to, I wanna learn that. Yeah. Speaker 0: 1 more. Speaker 1: I need to figure that out. I don't do any of it today. I've seen PSL even has an exciting new Tarka that I would definitely recommend checking out called Speaker 0: Meter Graph. Speaker 1: Meeting Graph. They're doing some Yes. Amazing Raise stuff there. They're a lot smarter than me, and and have some really cool stuff. There's a few others that I've seen and talked to. Speaker 1: Everyone that I've asked that's building this level of Agent, like A.I Hermes thing that's really low bespoke workflows, value don't work in that model. So, and the amount of Tech A.I amount of signals and Data, so that is a unique problem and there are some awesome people solving Singh. Cool. Come talk to me afterwards, Week love to hear. Thanks. Speaker 0: Next Week have Ellie talking about the global data analytics, how it's hard. And I gotta get up with this. Speaker 2: I'll Speaker 1: get you in. Come on up. Oh, yeah. Both? Speaker 0: If you Sent people to Speaker 1: hear you in the back of AI system. That one's perfect. I see. I see. Okay. Speaker 3: Alright. Speaker 0: Recording in progress. Speaker 3: I'm John be doing all Speaker 1: the demo. AI. Speaker 3: That's a lot. Alright. I'm gonna be doing all demo. So what what we're working on is embedded analytics. And you may know embedded analytics, but what what problem are we trying to solve? Speaker 3: We're not just trying to put data on the web. Oh, that's a lot of what, my Community, Ridge, does. We're trying to apply the way that Singh people think about and are able to communicate their product value. The GTM problem we're trying to solve is proving value for customers. So board for your customers, for for A.I employees customers. Speaker 3: How many people know Clayton Christensen John to be done framework? I love this framework. This basically says that you get hired to do a John. A product gets hired to do a job. And when you look at the way, most companies report on doing that job to their customers, they report on very operational things. Speaker 3: And so you might report how many County of how many things were done. Back to the pricing discussion, I think there's a really interesting opportunity in today's moment where we're all doing different kinds of things and pricing it differently to tell customers what kind of, value we're actually creating for them. Let me give you an example. We're working with 1 company that is doing a collections agent to try and help their customers collect dollars faster. You don't actually really care how many invoices got sent. Speaker 3: You care if your cash collection cycle is Synter, if you get more cash at the end of the month, if you can do that with fewer touches from a human or how however you do it. But 1 of the problems in in go to market is not many people know how to do this. So this is what we're building. We're we're building a system to help people communicate value. I'll show you a little bit of a demo. Speaker 3: This is a fake website. Imagine you're a company that is trying to help, property managers figure out where to invest in EV chargers. We're we're, as I said, embedded analytics and trying to solve a couple of the big problems of embedded analytics. The 1st 1 was actually performance. So, a lot of times, customers don't end up looking at the results that a platform has generated because those are in a really bad format team executives don't wanna use. Speaker 3: And so you end up in your customer success conversations in this really bad loop where you get super tactical people into, looking at your product, looking at your analytics HINTS your QBRs. They ask tactical questions. You end up answering tactical questions. No executive would be caught dead in share, and that's that's a loop that is, destined to fail for your platform because you're not actually communicating the job you're trying to get done. So like I said, performance was 1 of those big problems, and we Sonnet that. Speaker 3: This is 1 of my 1st learnings, not with AI. I love AI. I'll show you some AI, but we solved that with browser based technology and some open source that my cofounder build. And we're basically doing computation in the browser. That's 1 of the things that, that I think is a big takeaway in building today is Ash amazing as AI is, it's not everything. Speaker 3: The other thing that we're trying to do is marry up AI with more traditional experiences. So this this is obviously a board, and, you know, you'll you'll have folks say, hey. You don't need dashboards anymore. Everything will be agentic. Well, at some point, I do think human beings need to interact with something. Speaker 3: It doesn't mean all data workflows will have a human in them, but there are data workflows that will have a human in them, and communicating value is 1 of those workflows. And so part of what we're doing is we've we're adding a data agent, to the dashboard backed by the same Data so that you can actually get to deeper detailing the data agent. You can ask questions about this EV data, for example, AI, how does average range A.I price compare by make? A.I, and you come back and you can Anaconda see different makes here. Tesla has high Raise. Speaker 3: Porsche Laes lower range, high price, and and all these other in the middle. Another thing that we're doing here is limiting where we use AI. And in this case, it's to generate the Killian, and then we use it to some degree in the validation. But in the actual chart that's generated is, heuristics and deterministic. And that's partly because there's a lot of science and research behind how human beings can consume data in the right way, in the best way, I should say, to, to understand it. Speaker 3: And trying to get an agent to parse through all of that and and really apply it is really difficult. A.I SEO, for example, I asked a different question. You can see I AI got a different kind of answer, a different view of it A.I answer. And, again, deterministic on how you communicate the data using AI to set up the query A.I Sent, to some degree, validate and and make sure that that query is correct. And we spent we spent quite a bit of time on that. Speaker 3: You show, what is deterministic? What is, what is AI? I'll show you 1 more thing, and then I'll take some questions. I mentioned at the Running, we're trying to help people actually communicate the value of their platform, not just put bar charts and line charts in in, in the product. And so I'll show you a bit about how we do that. Speaker 3: This is what we call our build agent. This is the most agentic part of our platform. And, basically, when you add some data, Code be CEO Data, you could be connecting to a Data. Think of gold level data. So this isn't we're not trying to do data exploration where you're spelunking down and doing ETL and Date AI stuff. Speaker 3: You're you've you've got some data. You wanna get it out to your customers. You know City it shows the value. We actually, in our build agent, help you drive down to how do you communicate that value? What are you trying to communicate? Speaker 3: And the interactions with the users are all at this business level. This is actually 1 of the biggest problems AI think teams have when they're communicating value Ash they need to get a designer involved and a product manager and Ben engineering team. And at the end of the day, you end up, usually with something pretty tactical or something that was poorly designed. And so we're we're bringing to bear a lot of that research and so John, and then using the AI for what it's great at to to do some planning, bring in, different tools Anaconda and so on, and and John the fly Jake determinations about what kind of questions it wants to ask. Like, for example, here, when I said I wanna understand adoption, it came back and said, what does that mean to you? Speaker 3: And I met I said, well, that means active users to me. I never had to say Joe put this on the x and that and the y. And then it comes back A.I and it says, okay. Well, are you are you more interested in this or in that? It AI, in a different flow, ask team, do you care about time trends or point in time? Speaker 3: Do you care about, you know, seeing seeing the big elephant in the prove, like Tesla A.I that car data, or do you wanna, do you wanna exclude that? So it's it's got that nice Agent flow to City, and then it does some confirmation and so on and then builds me basically a a board, again, using best practice. And then we're we're doing some traditional things that we know from visual analytics, like meeting you, move things around, letting you make some changes. Undo and redo is, is very popular. But that's that's, my big learning in this building is how to break every single 1 of these things Dan, not only into is it deterministic or is it Agent, but when it's agentic, how hard is it? Speaker 3: So we actually use multiple models in here, some very cheap, small, fast ones for easy stuff and some bigger ones for harder stay. That helps us control cost and speed and so on. But that that breakdown is a is a classic thing you do in product and engineering. I don't think that changes. I think the tools that we Date, at at our hands to go take care of that breakdown is, is very new and different. Speaker 3: And then the last thing I'll say is it's not just is it deterministic or is it agentic A.I the breakdown that hasn't changed. I think that we still have to go back to those frameworks AI what job are we actually doing for our customers because it is so easy to throw features out there that I think it's important to, always remember what are you trying to actually do and really focus there. And that's my Tech talk. Speaker 1: In the background question? Speaker 0: Chris, is it AI, bring your own data? Speaker 1: BYOD. It's BYOD. Speaker 3: We bring the models. You literally put in a Date and start talking about the business value you're meeting. Obviously guided by our Agent, and that's that's all Speaker 1: it takes. Yeah. Speaker 3: AI not sure I understand your question, but I'll try to answer it and tell me if I'm off. We think there's a lot of signal in what people ask. So if you say, hey. I'm gonna tell my customers about all this value we Live, and you create the story, which is the board, and Ben, obviously, there's some, answers that they can get out of the data Agent. But everyone's asking the same question avoid data agent. Speaker 3: We wanna let our customer, the product Community, know that because there's something that's missing. Right? The idea of, the idea of this format is that there's a big picture and then long tail questions, like all the exploratory questions. If everyone's asking the same thing over there, that's information for your user journey to your point. I'm not sure if I asked your answered your question. Speaker 3: How do you know I have enough data? I I think that the company has to have a set of data it wants to share publicly. Like, the company has to, 1st of all, feel like they're creating value. And Sector, say, hey. We know we hopefully, you know, back to John to be Dan, what value you're trying to create. Speaker 3: Right? Like, I I have a collections Agent, and it wants to go collect cash for the cash AI. And there's there's some kind of value proposition there. The data that you have should be relative to that value proposition. Speaker 0: I guess, I mean, the company that would probably find, like, a business with the older. Speaker 3: I think I see what you mean. I think is it a good or a bad user experience? You have to go to outcomes. Like, are you trying to save somebody cash? Again, the John Tools done. Speaker 3: It's not how many of a thing, but it's AI you're trying to save a company cash. You're trying to reduce their sales cycle. You're trying to generate leads. There's some business outcome that I think I think, you have to drive to. Speaker 1: Yeah. I think Speaker 3: that's it. Alright. Thanks, y'all. Speaker 1: AI. Speaker 0: AI. Next, we have Shake, and Shake is gonna be talking about selling to the public sector with deep research settings. Speaker 4: This will look way more interesting in a couple minutes. Hi. So I'm Share. My Co-Founder in enterprise sales, and the thing I love is finding sales signals in places where people aren't looking. 1 piece, customer profile that I think people neglect a lot is state and local governments. Speaker 4: Some of the best deals Live ever done are within these, Growth, and they often have special stay up tracks there. And the thing that's great about selling to the public sector is that most of their information is legally required to be transparent. So if you know where to look, you'll find contracts, board meeting notes, Synter Gems, and the like. The tough part is this information is super hard to find. It's all behind JavaScript walls and websites that haven't been touched in 20 years, lead Media. Speaker 4: This is not a new concept to mine through this Data, share shares established companies like GovWind. They do an amazing job, dollars 2,300,000,000 a year in revenue Tools give you the privilege of accessing public data. The hard part is the insights you get from them are something like King County spent $200,000 in ChargePoint last year and that contract is gonna expire in 24 months. That's pretty team, but they're also John charge you $30,000 for individual license, which is astronomical for stay use cases. So this weekend, I did something that I think is way better than Govwind engine it cost me about $19. Speaker 4: So let me just get things set up here. Speaker 1: Okay. Speaker 4: So I live in, in Bainbridge. So I I focused on diving into what opportunities are there for a company within Kitsap County. Within Kitsap County, there are about 40 different what are called government units. Think each unit as a transit authority, a local school board, things like that. Across the across The US, there Bar, 90,000 of these government units. Speaker 4: So 90,000 potential customer information that you have on here. The cool thing is that as you mine through all of these contracts, meeting notes, etcetera, you get a lot of juicy Automations, tension between board members, internal politics, all sorts of weird stuff that AI Ben enterprise seller would five loved share selling to the private sector. The, so stay thing that I launched here is here, let me get 1 part going. The 1st part that's really hard is how do you actually scrape any of this data there, primarily because it's mostly under JavaScript. As I mentioned, there are about 40 different of these agencies across Kitsap. Speaker 4: And so in terms of trying to scrape this and build Dan initial corpus to test things out, I built A.I agent that took a course pass Date 1 AI all of the HTML on the website, finding out what what the structure is, and then trying to see how much of that information it could scrape. On my very 1st pass, there XAI only a 70 percent success rate. There were lots of things that weren't Reading, there was lots of hallucinations of different Recording, and so I kept going back and forth and found that giving A.I agent 6 attempts at finding information was the was the sweet spot. And after about 6 attempts, AI, my Reading rate jumped from about 70 percent to 87, which was pretty good. And so within Kitsap, what I identified was within the past year, there have been about 396 different types of contracts, all of which are Public A.I about 23,000 pages of meeting Minutes. Speaker 4: They have some very interesting things. When I mentioned offering types of, different types of checks against hallucinations in this data, the way that I tackle that is about 2% of all of the data that I trap was just junk. Primarily because it started hallucinating things like shares 1 contract that it scraped that said that, it Laes, started in 2021, but the contract wasn't, granted until 2026. And so to check against this, I put in a couple of internal logic checks. Is the start date after the award date? Speaker 4: Is it more than 2 pages? Having 5 or 6 rules like that greatly increased what the what the, output quality was. And as an example of something, so as all this extraction was happening, I I really wanted to focus on the dynamics of the people in the room. What are the what do board members really care about? Which ones are the dissenters? Speaker 4: What is what are the internal politics that I can lean into and try to leverage Ash a seller there? So as the extraction was happening, there Raise also an enrichment Laes where I would do these people specific things AI, who brought up a motion? Who seconded it? Who dissented against it? What are the reasons that are captured around that? Speaker 4: And that led to, some really testing things. 1 example would Ben, Speaker 1: Share is Speaker 4: an example is on 1 particular motion, the, Kitsap County Transit Authority, there was A.I proposal to purchase, 5 Shuvo, Seemingly non controversial, but very super loud, board members against stay, Week basically saying that council member Model was opposing it vigorously because it goes directly against all of, the county's policies against 0 emission vehicles. And so Ash you go through understanding what more of the things that crucial member Maklov has pushed throughout Kitsap County, you'll realize she's really big on Growth tech, Live lots of low emission standards and is really pushing for various clean energy things. So if you sell into any of those sectors or into that space, that's something really good to lean into. The other thing that you can find is, by mining through all of this information, you can build a really interesting organization chart as well. So not just who the, who the movers and shakers in the room are and who signs 2026 things, but actually understanding which council members matter for which particular things, what they care care about, which ones are the swing Workers, which people do you want Ben your buying committee. Speaker 4: If you're selling to a department within within Milwaukee's team, you need Precision as well because that person is the 1 who always Sector all of the motions. 1 other piece is Ash as Pipeline through these 26,000 pages of meeting notes is you also get a really firm understanding of what Kitsap County cares about. And so these are the biggest vendors that it has. Big investments into telecom, big investments into various Consultancy services from mechanical engineering Ash well as several things within the cybersecurity world. And so if you play into any of those spaces, this is a gold AI of folks to have that very 1st conversation with. Speaker 4: 1 other last piece that I'll share is, most folks in this room are selling software. How do you use this for this crowd? So let's see what happens. I'm gonna enter the name of a a local stories here called called Latch. Latch is a startup that's been around about a year. Speaker 4: They sell various AI tools for context capture for, for A.I Sent enterprises. Stay takes about 10 seconds. So I enter a bit of information on what Latch does. It goes against stay stories, and the negative and positive results are both really interesting. Since Latch is a hardcore AI tool, you know, there aren't any incumbents within Kitsap. Speaker 4: There aren't any champions because no 1 uses AI tools share, and there isn't too much of information about about what the Ridge is and how it maps to Kitsap's priorities. But what's really interesting is as you go into the notes about what issues council members have, shared against other vendors in the space, It mentions that, council members Claude, and and Model, transparency A.I finding more insights about their vendors and mapping out all of those processes share things they stay as priorities for this coming year. So if I were Latch, this would be, 2 people that I would really lean hard into in finding a local champion there. This was a very small snapshot of what's possible for a particular county. Kitsap is tiny. Speaker 4: If you apply it to King County or any of the bigger spots in the area, I'm sure this information would get a lot juicier as well. The hard part here is scraping 90,000 government units across the country is a hard job and probably not worth the AI. But if anyone has done these large scale scraping and maintenance product, I'd be very curious to see how you've structured those. I'll publish this on GitHub and make sure to share this with everyone in the room as well. Speaker 0: You, planning on turning into a company? Speaker 4: I'm not sure yet. Speaker 1: The, the Speaker 0: We have a studio here. Yeah. Even more so, you then start to spread out. Right? Yeah. Speaker 0: You go find the classic this, and you're just a flat salesperson. Did you know this? And that 2nd order map is quite powerful. Speaker 4: Yeah. The Bar part Strategies me faith that there's something there is companies already spend about $15,000,000,000 a year on tools like this that all AI suck. And so this seems like the next layer beyond that. Speaker 0: They've seen queries and things like that. Right? Yeah. And they have a permission, whatever. Exactly. Speaker 1: Yeah. Speaker 0: I think Live you considered lobbying? Because I think lobbyists would Live in store. Okay. Speaker 1: Great idea. Speaker 0: Yep. Speaker 1: Yeah. Speaker 0: I AI mean, it's crazy. I have spent during this past weekend, searching contracts. Like, I was doing the show version of that role just got board separate consultants. I five, like, A.I service disabled veteran and small business focusing on, like, I would pay several 100 a month. That, like, sample set 1 and maybe the number's higher. Speaker 0: Like, that's Speaker 4: I'll show you my Stripe link next. My job is, like, the entire time. That's awesome. Cool. Speaker 1: Yeah. Speaker 4: Let's definitely chat. Speaker 0: And they just we did it last year. The the the software that's out there for sam. Guys. Yeah. Speaker 1: Yep. Speaker 4: Cool. Anyone else? Excellent. Thanks again. Speaker 0: Alright. Next, we have are you Joseph? Laes. Do you have the Zoom link? Right. Speaker 0: Do you have the Zoom link? I have not the time. Okay. Okay. So we'll have Aaron join, talking on improving lead gen agent-based Speaker 1: Luke's. He'll. Speaker 2: Let's see. Speaker 1: Okay. Yeah. Hi, everyone. I'm Byron. Product, you SEO me sharing my experience with the lead generation agent A.I, lead qualification Agent. Speaker 1: So this is the next step. So I think, AI mean, I noticed that in the last 3 months, product, everyone is talking about loop. And I'm like, what the fuck is loops? So I I Tarka, diving in A.I, I tried to apply this concept of, self improvement to the lead generation agent that we have. So, essentially, what our agent does, you enter a AI, you get a list of, Event, prospects. Speaker 1: That's that's basically what what what happens. And each step in this journey, has at least 2 aspects. 1st is the output. Like, for instance, it could be a a set of candidates, and the process. So it's like, you have the results A.I you have the process. Speaker 1: And most of our focus was on evaluating the outputs because, well, outputs matter. Right? And my, I will argue today that it's as important to actually analyze the process of getting the outputs, especially with agents, especially since they are they have this nondeterministic, nature. So, essentially, what I build is Event day, I have a routine, AI routine that does the following. Just, loop through all of the searches our customers did, in the past day. Speaker 1: It highlights what are the new ICTs they were searching for. Is it, it, for instance, immigrant founders in the Bay Area board, supply leaders, in Germany? And, then it does 2 things. It basically takes the brief and, funnels 2026 different agents. Each of them uses a different version of our, lead generation Agent. Speaker 1: And, it runs through, that task, analyzes the outputs and the process, and then basically tells us what's wrong, what are the AI. And then I and my colleague, we evaluate those AI. We decide what needs to be fixed, and then we do a regression retest sometimes later as Week, AI this. So, a pretty City interesting example board 1 of our paying customers. It's a Germany a company in Germany, and they, basically Ellie, food tracking, software. Speaker 1: And, they got basically, he asked board, for a specific AI with the email addresses, of those, prospects. And what's, funny is that in terms of the output, it turned out that our agent did a really bad John, and, we wanted to know why. And what, when that nightly routine actually looked into the traces A.I, actually, each trace is basically, is basically this. This is AI useful where you can basically trace the the entire execution of your, Agent. And you can see the the latencies, the costs, the inputs, outputs, Tools calls, etcetera. Speaker 1: So we did that analysis and, essentially came up with defined that the way our orchestrator, makes a decision between what set of prospects works best for a given ICP, was wrong. So So we did a bad job of basically taking into account the some of the aspects of the contact information because the way we structure the ICP, the JSON that we use is, it does not take into account the email availability of a given prospect. So what I'm trying to say here is that since those traces are, like, megabytes of sometimes megabytes, literally megabytes avoid data, you can't you don't have really, a capacity to review, them Tools evaluate the outputs. That's why, you just need safe, some other agent that does the job for you. But, I noticed team multiple times that the kind of findings that agent produces, sometimes they're good enough, but sometimes they're very overfit to the exact output Tools the exact trace of the agent. Speaker 1: So you wanna do a human in the loop kind of thing, and, make sure that the findings of those agents are relevant, and you can basically make the fixes that Agent, suggests. And, now I shared, a skill that basically, helps to build your own agent to trace and to evaluate the traces regardless of whether you're using Langfuse or Langsmith or what what whatever. Basically, there are 2 type of information that it analyzes the correctness of doing a given pipeline by your agent. And 2nd is the some execution details. Like, for instance, it turned out that the Date, safe, was poor, with some of our model, and we AI that the way we structure the prompts, was not optimal to reuse the the Agent. Speaker 1: And to, close this Speaker 2: with Speaker 1: what what I realized overall and what where I'm, what what's the next step for me is to I have an agent that does the outreach. I have an agent that does the inbound lead qualification. Well, actually, they should feed signals to each other. So, essentially, when outbound, agent, generates meetings which end up with deals, I need to, like, back prove the signals and the information back to Tools the targeting A.I same with the inbound, requalification. We do the same kind of look. Speaker 1: And, what I'm trying to say here is that when you are done with AI the way a given agent works, you wanna move to how to optimize agents working together and providing signals and what's wrong and what's right to each other to optimize them further. That's basically, what this is about. Thanks. Any questions? Okay. Speaker 1: Uh-huh. Speaker 0: If you were if you were to do it again, what would you differently? Like, Like, what are you what are you gonna try next? Speaker 1: AI, try next is this. Basically, use the information from 1 agent to optimize the information of the other agent because the on a AI level Agent working for a given company, they are interdependent in terms of the outputs and signals that they produce. And that's that's the next step. But what I would do differently is probably AI would not rely too much on the findings of the agents, at least on the early findings because there's a lot of overfit, findings. There's a lot of, I would say, incorrect interpretation of traces and Singh. Speaker 1: So you wanna spend some time on teaching the agent how to evaluate, the those Traces. And that that's the job that you need, to invest into. Okay. Thanks. Speaker 0: AI. Next, we have Ash to talk about AI driven SEO agents. Speaker 1: AI. Engine? And, Joe, just so I know, where does this recording go? Like, will my competitors all see this? Hey, Joe. Speaker 1: Where this is where these videos go? Do these videos go to the public and stuff? Speaker 0: The video that you're recording? Yeah. It will be uploaded by Date your video and shorts. I'll pull it, and you can do whatever long. Speaker 2: Okay. Speaker 0: But we don't do anything. Speaker 1: Well, we do attach Speaker 0: it to his page on. If you want to not have a b on your It's fine. Speaker 1: It's fine. I'm trying to figure it out. Okay. That dictates how much I'm gonna talk about. Just in case, like Speaker 0: I can add a button where you can, like, delete it. Speaker 1: Okay. Sweet. Alright. Alright. My talk is SEO Delphi. Speaker 1: I don't know if anybody's watched the Odyssey CEO. Apparently, you have to, like, drive to Langley, SEO to watch, like, in full IMAX or something like that. But, okay. And the reason my talk is called SEO Delphi is because my company is called Poseidon. A little while ago, we started we were building something completely different. Speaker 1: It was just to try to understand customer reviews and stuff like this, but we built a sales agent pretty Ellie, actually. We were kind of 1 of the 1st to build these sales agents. We thought we could team Apollo, so we named our company Poseidon crucial. And it was just a Delphi. And and still who's that? Speaker 1: Oh, cool. That's not me. Cool. Yeah. So our company is called Poseidon. Speaker 1: And, essentially, what we did was we narrowed down our kind of ICP to to focus on wealth advisers. How many of you work with a financial adviser Tools? Just by a show of hands. Okay. Alright. Speaker 1: So SQLite, about 15 to 20% of you. So all of us should probably, at some point in time, be working with a financial AI. And the reason we we started working with them was because somebody came A.I they saw what was on my screen Date 1 point in time. It was, like, literally Running, and there was all these agents that were running in the background, like, kind of automating a whole bunch of stuff on LinkedIn. And AI like, holy shit. Speaker 1: This is literally what I do all day Langfuse I do outbound. I send minimal. I send emails. I send all these messages Agent basically waiting for people to respond back to me. I was like, cool. Speaker 1: You should use it. And then they became the fastest growing RIA on the launched. And that's kind of where we got started. So Date, we do a lot Moore, and that's because outbound has changed. And so today, Poseidon, my Code my wonderful cofounder, Andrew, is here. Speaker 1: We have what I kind of joke about is we have a platform that we're constantly trying to pass the, AI tour the Turning test. Right? Where it's like, hey. I'm actually human A.I I'm actually sending this. So we AI of bounce back and forth between, totally automated steps and then human steps. Speaker 1: Right? I think this is really, really important for all of you guys who are Founder, who are out, like, out there doing, you know, go to marketing. And it's like, hey. I'm I'm actually YAML. I'm actually authentic. Speaker 1: I'm actually a person, and I'm sending you this message. So an example of that might be something like sending a voice note share He'll be like, hey, dude. Robot, I'm just doing this as a demo. And then you can hit stop A.I then you can hit send. And so it'll actually send this in LinkedIn and he will get it in his LinkedIn inbox. Speaker 1: This is an example of some of the things that we do at Poseidon. But that's actually not what I'm gonna talk about today. So outbound has changed a lot. Right? So cold email, is basically rented attention. Speaker 1: It's a lot harder to get that attention in people's very, very, very saturated inboxes today because of AI slop and AI driven email. So what do we do? So what how we kind of thought about it is that every single campaign that we started to send needed somewhere really, really trustworthy to to land. And the way that, wealth advisory works is essentially the lifetime value of your customer is lifetime. It's 1 of the share few things where you work with somebody. Speaker 1: You will literally work with them probably until they die, or or you die or you pass on your business or sell your book of business to somebody else board, they basically give their assets to their heirs, their kids, A.I then they become clients of you or your kids and you just kind of run this business forever. So if the lifetime value of customers is so freaking high, then obviously Moore willing to spend a loop. But guess what? Trust is really, really essential. Right? Speaker 1: Because if you don't have trust, then you're really not gonna, you know, get this person as a as a client. So, so what I'm gonna talk about today is how we build of architected this trust in a platform that we're building next. Does anybody know this kind of directory strategy? Laes everybody heard of architecture strategy? Tried it Traces it? Speaker 1: Okay. I knew it to YAML it. Okay. So, hat tip. If you haven't watched, this dude on YouTube, his name's Raise 2026. Speaker 1: I will link it later. Really, really helpful to help you think about how to create directories. Okay? So it's a really, really underrated inbound machine, and this is the part where I was like, hey. I don't necessarily know if I want all my competitors to see kind of what we're doing. Speaker 1: So because of our supreme leader, there's a there's there's there's this AI. It's called FINRA. I'll show it to you. And there's a if you have an AI if you have an adviser's AI, this is this happens to be the person I'm working with. It seems scary. Speaker 1: And, you can go to prove Tech. Founder can see his details, and you can also see what's called the IAPD, the investor AI public disclosure. So every single AI out there by the Tech requires, is required to have this essential thing. And if you have this detailed report, it's essentially a PDF that has all of their disclosures Tools say, hey. You know, this person's good board they have, like, a felony. Speaker 1: You should probably not work with this person, so on and so forth. And it shows you how how long their years of experience are, what firms they Workers with, and so on and so forth. So what we did was essentially Week ran a very, very comprehensive kind of ingestion scraping job share we take we took 5 to 6 Minutes, different proxy servers, and we ingested all of these AI. Then we took these profiles A.I we essentially put them up, online. So we took all this fragmented supply. Speaker 1: And then what we did was we AI structured it around intent and content. Right? So if you think about, like, the very, very bottom of the funnel in search, it's like, hey. I'm looking for a financial AI in Seattle. I'm looking for a financial adviser in, you know, in in in Bellevue, in Austin. Speaker 1: Right? And SEO, basically, we allow you to be there at this sort of Agent of intent. So this is AI of what it ends up looking like. So if you look at the very, very fat tail, it's just like I'm looking for a financial AI. And then thinner and thinner and thinner, you're like, stay. Speaker 1: I'm looking for advisers that work with SPEAKER employees, or I'm working closing for advisers in 1 specific location. Right? And then the long Moore, really, for us is to create very, very niche micro content sites share depending on if it's a company that you work for board, like, Google or SpaceX or if you're like, hey. I specialize in equity comp and RSUs and concentration risk, then this is the AI that I absolutely need to work with. So what does our set of agents so a lot of this is run by a set of agents, which is really cool. Speaker 1: And so here's all of what these, agents do. Event day, we have a thing that goes out and checks Google Leader Sponsors. So there's XAI MCP that goes and connects to that. Then we look at competitor ranks. SEO, obviously, Kitsap a very, very competitive this is the most 2nd most competitive category product for search on Synter after ambulance chasers, after lawyers. Speaker 1: Right? Then we look at AI gap analysis. So, like, where are we basically where's our gap compared to any, like, top tier, like, competitor that have had a lot of money invested HINTS? And SEO let me show you this practical person. So there's kind of 5 different steps here. Speaker 1: We call it sync, sense, decide, act, and report. And so what we try to do is we try to break it down by, hey. Here's what the agents can do, and here's what human beings need to do. So, like, Event day, it's like, okay. You know, humans just need to act on the human stay, and then agents can just go and run Manya eval and say, this is what we can just sort of automatically do without, you even intervening. Speaker 1: So the examples Bar, like, things like syncing AI, syncing firms. As new firms get added, we are constantly syncing with them automatically. And then going down through the, the funnel here, this is an SEO right skill. So let me just show you an example of what the SEO right skill will do. It will create something like this. Speaker 1: So it's a very, very common search Square. It is, hey. What's the best state to retire in? Or what's the best place to retire in in Florida? And then you would click on Florida, and then you would SEO, like, okay. Speaker 1: Here's the strongest Sector, the state, your safety, lifestyle, climate, and so on and so forth. SEO, yeah, this is this is essentially all sort of automated. There's also a video agent that automatically creates videos that are based on this content and research. There's research agents team go out and do all this stuff, and then automatically come back. And I guess the key thing is it reports it back into into Slack. Speaker 1: SEO, essentially, all we need to do Event day is just check Slack-Native then humans just kind of go and reuse stuff AI, okay. Here's an IDENTIFICATION, requested. I'm not gonna share this because, obviously, it's very sensitive information that people have shared. But, yeah, it's, that's essentially how it works. Questions? Speaker 1: Confidence? Speaker 0: For someone that hasn't done a directory strategy, what are Singh that they should be on? Speaker 1: The main things are to try to figure out how to use Google Search Console really, really Ellie, and then be really timely in what your, stuff is. So whenever the SpaceX IPO happened, we knew that there was a trending search alert. So we had a Google Ben MCP. Google yeah. Google Trends MCP that basically, like, Joe, wait. Speaker 1: There's a lot of trends API people writing for, SpaceX employees. A.I then we created this, and then it, like, just shot up. There's a lot more profiles that are being, like, basically Claude because so our process is that you go in and you claim profiles. So if you can do something where you just put a profile up and then you have people claim it, it's a lot easier than having to go and, you Speaker 0: know, Speaker 1: do a whole bunch of stuff. Right? You just A.I Laes process is much easier. Speaker 2: Questions? Speaker 0: Yes. Speaker 1: Yeah. Any, any large m and a or liquidation so what we do normally on, like, on Poseidon, on outbound, any if if there's any m and a Event or things like that, we will look in that particular A.I, and we'll, like, reach out to the people. And now we've kind of started doing it on the inbound side as well. So if you're a client at a particular company, AI, if you're Anthropic and you have secondaries or something like that, then, then you should you should probably go here. And Ben, eventually, we'll it'll be do some somewhat like A.I Vabicari auction share, you you know, if you wanna be an AI on the financial planning for Anthropic employees page, that will probably be premium real estate. Speaker 0: If you step forward, we align the Speaker 1: Yeah. So let me give you an example example. 1 of our clients, really focuses on automotive Engineer, and he Laes, like, 6 Tesla, you know, leads that have come from this process. And if you kind of look at what he does or what he focuses on, and and you go to, like, the Tesla page, you'll say, hey. Who's worked with Tesla employees? Speaker 1: And then he Killian tends to pop up. But Ellie the key author key thing, Dan, for for your just to give you another idea of what works well is that, there's there's a lot you can do with it. Ellie Build John share you feed all of this data back into, like, chat GPT and stuff like that. And so this this particular page, like, the best financial AI in San Francisco, stuff something like this would be very, very important to have on every like, for every facet. So the best AI who do estate plans, the best, you know, advisors in the divorce category or whatever it might be. Speaker 1: So we have literally pages that are built for every single 1 of those verticals. Stay takes a long time to do, but it's totally worth it. Yeah. So this is this is there's yeah. Ash there's a whole matching thing too. Speaker 1: I'm not gonna I didn't wanna bug you guys with this, but so we have a matching thing as well. And and I guess this kinda uses AI as well, but it's an entire, voice voice Agent. So you could just have a conversation with it. Basically, what you're doing here is board context loading this Singh. And then I think she talks to you. Speaker 1: Sheikh she doesn't. AI oh, so I guess she is. I guess Week can't hear City. But, but, yeah, basically, you can talk to it and then context load it, which is way, way richer than than any of the incumbents or competitors that AI I mean, they just don't have a voice agent that you you do, like, a very, very in-depth prequalification process. So that's this is for the retail side. Speaker 0: You so you've essentially, like, inserted yourself into your customers growth channel. Speaker 1: Yeah. Speaker 0: And then you're hijacking that so that you're a vital organizer source for SEO and SEO. Speaker 1: Yeah. And SEO, I mean, I guess I didn't show this Ellie, but this is our our largest competitor is they've they've raised a 160,000,000, and they have 14 years of SEO. And, like, we're like, okay. We've actually gone higher than many of our other competitors, and we're already on the 1st page. So, like, you know, obviously, this works. Speaker 1: Yeah. Speaker 0: Yeah. Speaker 1: I got this idea 2 ways. 1, we did a Sent hub for a large RIA, and they paid us a lot to do it. And we ranked higher than UBS, Organizer, like Folder for a lot of very, very high Sent stuff. And we AI we recognized how valuable that was to them and their business. And so we're AI, wait a minute. Speaker 1: If we did it on a different vertical, or if or, like, on A.I independent website, that independence is actually what, like, helps you with that Traces. And we believe that trust is, like, probably the most like, Ash, like, SaaS gets commoditized Tarka. But, like, you know, the the most important thing is, like, creating trust based channels, and that's really what we're after really long term. Yep. Speaker 0: Yeah. Speaker 1: Yep. Yeah. Actually, so it's a great question. So, so if you look at, like, how we sell to AI, so if if you so if I go to the 4 AI. Right? Speaker 1: So this is what we say. We hey. You have, like, the SCC data only. That's the vanilla profile. This is your AI guide, like, verified profile. Speaker 1: And then the the 3 steps or 4 steps, I guess, are find your profile, enhance Date verify, collect reviews. Now what's really, really interesting about this is the SCC changed how they do reviews. So you specifically have an SCC s attestation process, which we do. But what's different about how we do it than anybody else is sorry. I'm scrolling to it. Speaker 1: So it's verified reviews, but it's private. So what you do if you like, let's say you have an AI and you go in, you can verify yourself with Hidden. So you can, like, verify that you're you're who you say you are A.I Ben and steps who you say you Bar, but it can be private, which is really, really important for higher net worth, you know, privacy. Awesome. Cool. Speaker 1: Thank you. Joe, I'm sorry. Speaker 0: Alright. Next, we have Joel who's gonna be giving an update on his AI driven ad platform. Thank you so much. AI. Sorry. Speaker 0: I'm so sorry. Speaker 1: Okay. Speaker 5: AI just got off the phone with the an Agent. I think I scared the hell out of it. Alright. I'm Joel Horowitz. I'm the founder of Sent Media. Speaker 5: Let me switch to the Zoom. Right? Okay. 1 2nd. Bear with me. Speaker 5: I'm so sorry. Right He'll. Right there. Oh, no. Work. Speaker 5: Oh, Dan, do you know why this link is weird? Sales UNC let me see. Speaker 1: Yeah. Speaker 5: Is that right? Kitsap just AI that 1 in the oh, there it goes. That was weird. I'm not AI. Okay. Speaker 5: Sorry about that. Which 1 should I use? This 1 or this Speaker 1: 1 board both? The Recording. Speaker 5: The recording. Stay. Speaker 1: Okay. Good here. AI, guys. Speaker 5: Thank you so much for waiting. Okay. Let me open this up. Okay. Speaker 0: Yep. Quick shout out. So seeing a couple of demos out just whose 1st time is this at a AI Converge call? Great. So much for joining us. Speaker 0: So you see these different presentations, all this kind of stuff, above all kinds of generation together. So, if you have anything that you're working on, anything that AI, definitely, consider Recording board progress. Speaker 1: Sorry. Sorry. AI bad. Speaker 5: Sorry about that. Okay. I did the total faux loop of, not turning my mic off. Okay. Here we go. Speaker 5: So I'm Joel. I'm the founder of Sent Media. I was here, what, Dan, 3 months ago or so when I kinda launched. So I've been back in the marketing about few months now. What you're looking at is the 1st basically Ash, AI ad platform that can run campaigns on any, any AI platform in the world. Speaker 5: We're, yeah. So I'm gonna just do a test. So, someone Raise their hand. Like, what's what's the name of, like, like, a website, Live, your company, something? Like, just a website. Speaker 5: Any website He'll do. Speaker 1: Yes. AI. Speaker 5: AI. Alright. Hold on. Alright. Go ahead and build, an AI, campaign, research their company, build an audience, and launch a Google AdWords campaign. Speaker 1: See if that worked. That Week? Speaker 5: Oh, no. My oh, because I'm oh, I'm on Zoom. I can't do it. Shoot. Alright. Speaker 5: I'm gonna dang. My whisper Laes dead. Usually, that's my demo. Okay. Hold on. Speaker 5: So go ahead and analyze. No. I know it's Live typing. Oh my god. Who does this? Speaker 1: Anyways, go go Speaker 5: ahead and AI. What's the what how do you spell the website? Speaker 0: Advisor team. Speaker 5: Uh-huh. Speaker 0: Yep. .Guide. Speaker 5: .Guide Live this. Just like that? Speaker 0: Yeah. Speaker 5: Build an audience based on Sent signals. What are you optimizing for? Speaker 0: I have some questions. So What should I Speaker 1: Lessons? No. What do you do people sign up? Speaker 0: Just tell me Week I see the Speaker 1: Okay. Fine. Let's stay, Speaker 5: tell us tell us tell us based on their good good 1, man. I like that. I see what, what we should what what we should optimize. XAI can't spell so well when I'm writing AI board. Jeez. Speaker 5: That was, like, really hard. I'm so sorry to watch me type. AI. So what it's gonna do now, it's gonna open up an agent. So it's basically, going to oh, 1 thing I wanna show you 1st is that Copilot Singh. Speaker 5: 1, as it's running, you SEO this how it's, like, orange right He'll. That means it's in draft mode. So, like, 1 of the things I'm learning, if you build any agents at all, you need details, you need to have safety Date. You need things to make sure your agent, like, doesn't go off the rails. Yeah? Speaker 5: And so the 1st thing is basically, AI, I have this, like, switch apply live, don't be live. Switch July, don't be live. So, like, that's the safety gate. That's number 1. Number 2 is that, I have I can give it access Tools, an account. Speaker 5: This is just a sandbox that's running in right now. But if I Manya actually connect to 1 of my live County, so I can you can see I have, like, every account. AI I say, actually, like, go ahead and connect it to, like, Google like this, and I'll it'll actually now have access to my Google AdWords account, which is pretty safe. When I was doing this in Claude at Sourcegraph last year Running YAML code myself, I would use developer keys AI everyone does. Right? Speaker 5: Say, okay. I'm just gonna do it. All of a sudden, the Claude stories hallucinating A.I it started updating my site map A.I it stories updating my site links to a different brand. I was like, that's not so Growth. And then it started doing bad things. Speaker 5: And so, then we actually started working with other clients because I was an agency And Claude forgot that I was working in 1 client, not the other Client, started doing other bad things. So that's why I started the company because I felt this pain pretty apply. Lost 1 client over City, in fact. And so that's why I built this product. What's pretty cool about it, as you can see, it's just 1 shotting minimal going through this end to end. Speaker 5: As Dan mentioned earlier, we also have signals in Sent and you'll see them in a Agent. And we have a couple of proprietary, partners. 1 is, providing us 400,000,000 business Outputs. AI, basically share emails that are hashed. No Workers, privacy. Speaker 5: And then number 2 is we have a, person Tools basically, providing us, Minutes Singh. AI, a lot. Haiku, Board, g 2, AI, way better than, 6¢ or anything else you're gonna find. So anyways, like, it looks like your AI guys A.I 2026 sided marketplace. Yes? Speaker 0: Yes. Speaker 5: Yay. And then it's AI key findings. You're ready for ads, my friend. And so, basically, Sales you should spend about 6 k a month on maybe Google and Reddit. You can do it right now. Speaker 5: Stay go sign up, swipe your credit board, and you'll be running ads. It's saying that basically, AI autonomous Code intelligence don't know if that is, but loop pretty great. That's a group they should probably target. These are some of the channels team recommends. We have all of team, so you can just do it. Speaker 5: Founder like you're already Reading. Are you running on Reddit already? Speaker 0: Sort of. Speaker 5: Sort of. Okay. Cool. So you got that kinda right. So check this out. Speaker 5: So then you can, like, dial in, like, stay. Who do I really Manya target? Financial AI, recruiters, maybe shares, like, a board, RA, whatever. I don't know. But what's interesting is I used to be a data science in another AI, so I like k means analysis a lot. Speaker 5: Who stories k means? Speaker 1: Oh, Speaker 5: some data Speaker 1: AI in the house. Right on. There's a k means cluster. Right? Basically. Speaker 1: So now AI can say, actually, when I'm running ads, I don't Manya just spam the universe with, like, random ads. I can actually choose groups of people. And most people think, oh, I'm gonna do job Seattle. I'm gonna do profession. I'm gonna do geo. Speaker 1: None of that matters. Right? You Manya just find the people that's gonna buy your product. Right? That's what this does. Speaker 1: You can see there's different groups here. You can actually choose different traits based within those Growth, or you can just select that up here. And as you select them, they'll get added to your list, whatever. And then you can actually, like, build the audience. Right? Speaker 1: So then you can say, great. It'll It'll give you a Singh. You know, all the things stay, Growth. You show, 1 2nd. I can't hold this in AI. Speaker 1: Sometimes Team sorry. Usually, Visual, like, use the thing. So then you can basically say, okay. Build you know, upload the audience, you know, to, Google search, and go lead, and build out the campaign. So I'm showing you this, like, in, Sent, AI, a center UI, but it also works in Claude desktop. Speaker 1: So Week have an MCP Agent connecting. You can, like, use cloud and never look at me ever again. Just go and buy AI, you get all the connections, and you just go. You can also use this in ChatChippy Team. I was at my son's soccer game. Speaker 1: I should've been paying attention. I was just checking my ads. Like, okay. What's going on? Speaker 5: It also runs autonomously. Right? So after it launches, I can all say, great. Keep an eye on this campaign. If CPAs go above $50, then, you know, dial it down. Speaker 5: Right? If it stays at team or below $50, ramp it up. You can do all of this. Look at negative keywords. Look at positive keywords. Speaker 5: Rotate my effective, like change the Klaus. So any of that. It's all autonomous all day long, which is pretty pretty sick. Anyways, what else? Tech. Speaker 5: Question. Oh, I got 1. That's it. So, like, basically, we have a Event we have a solopreneur conversions. It's $99 a month. Speaker 5: I'll send you guys all the coupon for a $100 to go AI play for for free. We have a meter Ash you can see running up here. It's pretty much transparent pricing. I'm just doing cost plus, right, of the models model much. Oh, 1 more cool thing that you'll never find out unless I told you. Speaker 5: Hold on 1 2nd. SEO as as Dan was talking, he's creating he's creating Agent. Right? As am I. SEO, basically, we have different agent-based different things. Speaker 5: So, like, if you wanna do analytics A.I you wanna do creative A.I you wanna do optimization, you wanna do prospecting, prospecting Singh fun by the stay, so use that caution. We use better models. They have skills, they have tools, they have better models. Right? So for example, the creative, agent uses, Gemini Prove, like Anaconda 2, AI, has all that build into it. Speaker 5: So anyways, we're just getting started. Ads are just the 4 Ridge just the wedge, but we're gonna have a a publicist agent. We're gonna have a which basically is Ellie m observability. We're gonna have a, outbound email Agent. And so we're building the framework for how you guys build agents and bring them into business context. Speaker 5: So that was my 5 minutes. Good? Solid. SQLite? Alright. Speaker 0: Good. Speaker 5: Code. Very good. Speaker 1: Yeah. Lessons? Speaker 0: AI, man. Speaker 1: That's such a great question. Speaker 5: The answer is well, I'm the answer is everybody. But the people that don't get don't get freaked out are AI natives. I just was on a call with an Australian based agency and they're, like, freaked out. I don't think I I I don't know. They're, like, I don't even know why we have an agency anymore. Speaker 5: Literally is what they safe on the call. It's file, I'm so sorry. Like, I don't even know what to say to that. But it's true. Like, that's what's happening. Speaker 5: So, like, you know, we have, Hyrax, which is e yuru.com. So Tristan's, using the product and they're running their campaigns there to launch their new products. We have, span.app, which is an infrastructure AI infrastructure customer. Week have an insurance company. I don't know if I can name them because it's Consultancy. Speaker 5: John Midwest that's using us. We have 1 of the largest global fast food chains that are using Sent. Yeah. It's just in every avenue. Speaker 1: Mhmm. Speaker 5: The creatives I'm still working John, I'll just you know, fair turning. Yeah. Because it's like AI. It's it's loop, but I am doing HINTS, I think, now that allow you to, like, tune the creative. Think of it XAI, like, Figma inside of this. Speaker 5: You'll see it when you AI. It's pretty cool. Yeah. 1 minute. I'm done. Speaker 5: No. No. No. No. Yeah. Speaker 5: Go ahead. Speaker 1: So in draft mode, is there, like, a base pricing or something to just get all this advice and research? Speaker 5: I mean, you don't ever have to launch a campaign. And so Ellie, you can just not launched campaign, just use it how like the regular AI. Like, this is chat 2026 p t, but better because you have all the models, not just 1. Truthfully. Right? Speaker 5: So there's a there's another angle here where it's not advertising at all. Some people come in here. In fact, when you build this, sorry. I know I know I'm, like, over going over. You know, like, download a Board doc? Speaker 5: So I do this anyways. This would cost people $5,000, and I used to run my own agency. This This is the 1st thing I do. I say, oh, I'm gonna create you an I'll create you a, a brief. Right? Speaker 5: And say, here's what what I found about your site. So there you go. And I would just, like, send this. So after I, like, get off a call with a a person, I send them this. Speaker 1: Yeah. AI send it to you. Yeah. Speaker 5: For file? Pretty wild. Right? I mean, this would take me months to build, literally. I'd have to go into their ad accounts. Speaker 5: I have to go into their analytics. Yes, Alex. Speaker 0: AI know. Speaker 5: I have to hire a designer, Alex. Okay. If you know any good designers, come talk to me. I have I'm hiring for AI. Yeah. Speaker 5: Oh, it's just me. Speaker 1: Yeah. Just me. Speaker 5: Yeah. So that's why I talk fast A.I I because I I don't AI, so I just talk. And so now I talk so fast because I'm using Whisperflow all day loop. Then I interact with normal YAML. Like, why are you talking so fast? Speaker 5: I'm like, I know. I have to turn it off. It's hard. But that's that's the truth. AI don't think I actually have 2 Code. Speaker 5: Believe it or not. Yeah. Thank you very much. Speaker 1: Yeah. Yeah. Question. Speaker 0: I'm like, I had yeah. Very, like, formal strategy. Speaker 5: But it's, like, all of a trademark. That's I honestly thought that's what this report would five would've been. But when I started, it was it was more AI all it was when I started was just an audit of the website. Like, do you have the right pixels in place? Because, like, there's a lot of competitors in my stay, Live, a lot. Speaker 5: You can throw a rock and hit a competitor. But most of them are just focused on the ad layer. Like, oh, like, here's creatives. Everyone thinks creatives. Right? Speaker 5: That's not the hard part. The hard Partner conversion tracking. Get that shit right 1st. Yeah. If you don't, then all of that money is just gonna be wasted because you can't track anything. Speaker 5: Google can't optimize for anything or Reading or anything. And so I Lucas on that stay, and then people are like, oh, man. SEO great. I don't have to go into, like, Google Tag Manya. I have to figure out service AI Reading. Speaker 5: Kitsap a pain. It's a pain. And so I focused on the biggest pain 0.1, which was that. So so to answer your question, then I met I made some friends with some people who AI some signals and some audiences, and that's how it gets really, really tight to answer your lessons. So now it's not derivative. Speaker 0: Yeah. Speaker 5: Lessons? Comments? Yes. Speaker 0: What's the next Speaker 5: step on the creative side? SEO 1 thing that's really cool for all of you A.I John and I will get to the Raise. So I my creative's alright, but, like, I think I'm HubSpot for, sorry. I'm just looking for my settings. I think I'm HubSpot for, it's down here. Speaker 5: Sorry. Again, I need a designer. I'm HubSpot basically for agent-based the way I would think about it. Right? And so I might not have, like, the best creative, but, like, then you can come down here and, you know, basically Ash Higgs Fields if you want it. Speaker 5: Like, I have an MCP server. So extend it all you Sent. AI your own MCPs Langfuse it. Right? And it'll bring it bring it all in. Speaker 5: So to answer your question about the creatives, our kids are pretty good. I'm working on, like have XAI have anyone Haiku lead, like, engine dot dev? Yes. 1, 2, 3. So I'm building pencil dot dev inside of center. Speaker 5: Right? So if you've used it, it'll look like Figma. And steps of humans designing things, it'll be AI designing things with you. And then you can give it feedback Agent like Lovable, five, grabble City, and all the other things. Like, I can show you. Speaker 5: You wanna see it? Speaker 0: Dan. Speaker 5: I know I'm, like, way over time, so I'm sorry. But, let me see if I have an example here. Bear with me. Role on. Bear with me. Speaker 5: That's not this 1. Sorry. I AI, like the other cool thing about this is you can, like, share. There you go. So this is what it looks like. Speaker 5: I haven't released this yet, so it's coming. So, like, you can just it's, like, really loop. But, anyways, you can just basically, you know, interact and, like, click on this and, like, you know, then see how it, like, updates over here Sales, oh, you're editing variant blah blah blah? So, like, now you can just basically sorry. I lead to do this 1. Speaker 5: Let's see if I yeah. There it goes. So you can, like A.I stay John Shopify today, by the way, and StackAdapt. So if you have Shopify Pro, you can go in and request. So I have a very big customer that's coming John AI. Speaker 5: I can't talk about yet, but I'm fucking excited about them because they're gonna be so cool. Anyways, so they're coming in. So if you go to Shopify, you can just request Shopify, and AI create a custom apply, and they can onboard you. And then we also are an agency for Kitsap now. That's programmatic. Speaker 5: And so, like, we can now go digital out of home, billboards, AI, everything you Singh of. SEO, anyways, I built this. So the way you can do this thing 2026 say, actually, I wanna change, like, this a little City, and it'll it'll edit. I'm not done there. I'm not done yet, but it's it's a work in progress. Speaker 5: A.I AI I'm talking too much. Any other lessons, feedback? Speaker 2: Yeah. Speaker 0: How do you, like, call Speaker 5: Is that Yeah. Yeah. Yeah. I built Dan attribution. So we use, like, mixed model marketing. Speaker 5: Believe it or not, it's an open source model that Google itself uses in their marketing, whatever. It's just a Python share. And so we use mixed model marketing for 1 thing. So basically the other question I get a lot by the way is AI, well, won't Google just do this? And Google Laes, like, Dan banana built in engine they have, like, Max AI and they have these Singh. Speaker 5: Truthfully, Google will keep taking your money. They actually are arbitraging Washington they're making money by you not being smart enough to know that they're just taking your model. Right? Purposefully. And so they're not gonna tell you, like, actually, you've role saturation on Google. Speaker 5: You should go engine the meta A.I or go to Microsoft Bing or go to Reddit. SEO our mixed model marketing, it's not perfect. I AI have to build it. Is the is gonna say, like, you're basically saturated on this channel. You should go and move to board spend it somewhere else. Speaker 5: But, yeah, I'm building, like, 1 of the things I'm most excited about, sorry, is, Sheikh. I just forgot the name of City. But it's basically when 2 models, like, fight each other. I forget what it's called. Adversarial. Speaker 5: Thank you. So, like, you can I'm starting to build adversarial model. And so just the other Dan, I had a customers, he goes he goes, Joel, I told it to, like, how I should set up my Google Labs. Should it be on, like, in Google Stay Manager? Should I build it into the, like, into the website itself? Speaker 5: I said, well, ask both A.I, like, 2 models lead do 2 different things. And I said, well, okay. But let bring in author and get 0.333 Killian. So we brought in the 3rd model A.I it can do that. So we're bringing that HINTS you actually get really good advice, but it's a really interesting thing to think about. Speaker 5: 1 last comment I'll make is, the the same person lead, you know and then it asked me, like, manually 2026 do it myself. It told me I should go into Google Tag Manager. I'm like, well, tell it OpenAI tell it you do it. He goes, really? You can do that? Speaker 5: And I go, yeah. He goes he goes, Joel says that you should do City, and so he did that, and then then it worked. So it's a really bizarre situation, but it's it's what it is. Anyways, I encourage you to give it a shot and try it out. It's free to get started. Speaker 5: Yeah. That's what I built. Speaker 1: Very entertaining. Speaker 0: Thank you. Yeah. Speaker 5: Alright. Agent nervous of that. Speaker 0: The AI time we had. Speaker 2: Thanks. Thanks AI.