How Having A Single Brain Will Level Up Your AEO, SEO, and Marketing 100x

And then we will go ahead and start. 3 2 1. Okay. Hello everyone. Welcome to today's session. How you guys doing? Welcome, welcome. Okay, while people are streaming in here, I'm just going to go ahead and intro a wonderful guest for today that many of you guys will know. Yes, happy to see you here, Sean. Yeah. Um, it didn't happen yesterday. So, today is the session. Um, welcome every day. Um, Alfred, welcome everyone to today's live workshop on how having a single brain will level up your AEO, SEO, and marketing 100 times. Today we have the pleasure of learning from how Eric and his team at Single Grain orchestrates a whole fleet of AI agents and how they share effectively the same United brain to get AI working for them so much more effectively. Uh, you will also know Eric from hosting two other podcasts, The Marketing School and Leveling Up, which have racked over 120 million downloads between them. And he has worked with companies like Amazon, Uber, Airbnb, and Salesforce on growth and customer acquisition. He's also one of our keynote speakers last year at HRS Evolve San Diego and spoke about getting blog content automation right with Tim on our HRS podcast. So, he has a a dedicated episode together of Tim. So, he's really here to help you take the next step up. So, but before we start, let's do some very quick housekeeping. This workshop is currently being uh streamed live to YouTube. So, everyone who registers can watch the replay afterwards. And if you have a question, Zoom has a Q&A function that you may see in your toolbar. Please enter your question there so it doesn't get missed uh somewhere through the chat uh and it would also get prioritized your questions to get answered. All questions will be answered after Eric finishes his main presentation and um so like you can ask your questions there but if you want your questions uh you want to hear the answer from Eric or me if you ask nature's question it will happen after Eric finishes uh what he has prepared mainly today. Okay so that's that's basically it. Eric you have the stage. Please take it away uh and begin. Yeah thank you for that wonderful intro constants. So yeah thank you all for joining. Um if you guys can type in the chat right now where you're coming from in the world. I always like to do that in the beginning. Um that's always interesting for me to see. So I'm going to be talking about um as Constance mentioned just having a single brain right and and somehow I was able to get that domain. I have single brain. Then I was like oh there needs to be a single brain. So that's how we came up with the concept. Um so I'll be talking about how kind of SEO AEO is evolving but also just to think about how to work about how to go about business and marketing in general. Um, I'm just going to take a quick look at where you all are coming from. So, we got Sa Paulo, we got New Zealand, we got Seattle, and we got a lot of your notetakers, I'll tell you that much. A ton of your note takers in here. We got Phoenix, Montreal, Pennsylvania, California, Sarasota. That's amazing. Okay, well, Canada, LA. David, I'm in LA as well. We've got Sky from Scottsdale. Okay, so I don't have a lot of time here. I'm going to jump into things. Mexico City. I was just in Mexico City for the first time last weekend, but I'm going to jump into the presentation. Drop questions as we go. I'm going to try to get to as many questions as possible. Um, I'm going to jump around doing screen shares. So, um, I do have a deck, but I'm going to jump around in that deck and I think I'm going to try to get this this one concept across first. Um, and then we'll kind of go from there. Okay. So, let's jump into this. I'm going to switch my screen share to over here. So, this deck over here, this is this is the concept. One brain, every team move 10 times faster. So I just want us to think think about the the the work that you are doing today. Okay. You you you have Slack, you have your HubSpot, you have your Salesforce, you have your HRES, you have your Google Analytics, you have your Google search console, you have all these tools and they don't talk to each other. Your team, they need to go pull all these. When you want a data pool, it's going to take a couple days, maybe even a couple weeks to get the data pool done. Then you got to strategize around it and and then you got to actually do the work, right? And so that that's very inefficient. And so we need to reimagine how we do about work and then this will directly affect how we do SEO. And then again I'll talk about how SEO is changing. Okay. So I'm going to jump around here. Uh let's go over to this piece. Okay. This exactly what I'm talking about. You got HubSpot. You got Salesforce here. You got Google Analytics, Google Search Console, Slack. Okay. None of this stuff compounds because they're all just kind of talking amongst themselves. This bear is like talking to someone else on the phone. This fox over here is looking at his iPad, right? Um this this uh lady cat over here is just bored. I don't know what's going on with her. and the rabbits are just working. So they don't talk to each other. Your knowledge stays siloed. Change management stalls. You know that word change management, right? And then you have pilots that never ship. So what you're looking for ultimately is you need some type of world brain. You can call it a company brain. We like to call it a single brain. You need one brain over here uh where everything can kind of talk to each other, right? And I'm going to show you how we're going about doing this um using the HFS API. the HRS API, the MCP has been incredible for us and and I highly recommend those of you that are uh HRS customers, which just most of you are, are using that, okay, taking advantage of it. Um, when you then have this single brain and everything's talking to each other, you're then able to connect it into these different agents, right? So, you have single brain up here. This is the main brain, okay? And that houses all the context. Maybe you have process in there. You have uh your maybe you have your client information in there. Maybe whatever your your HR data in there. Okay. Everything that your humans are calling on are calling from the single brain. Not only that, you then have your single brain will then enable a fleet commander that can manage all of the agents that you have down here. And I know a lot of people like to say, you know, they're using agents right now. And I'm actually curious, drop it into chat, how you're using agents right now. Are you using agents right now? Are you using co-pilots at all? Like what are you doing? Exactly. I'm trying to get a sense of what your AI fluency is at right now. Um, and then maybe I'll I'll continue to kind of try to tailor this this talk as we go because we don't have a lot of time. So, here's how you want to think about this this org chart. The the org charts being reimagined here. Okay. So, you have this one brain, same chain, every seat. So the human single brain is the umbrella and then the human is still saying hey okay you know whether it's the client or or or you guys listening right now maybe you're you you you have an agency you're the operator okay and then it then you have a mainrain and then the mainrain will then tell the agent fleet what to do and then the fleet the sub agents that live with under the other agents that live with under w with within live under the fleet commander they're doing the bidding of what your human wants to do at the very top. Okay, but none of this works unless you have your tools talking to each other. Okay, and there's a lot of people on X right now talking about how this all works. And there's more and more companies working like this. I'll tell you one thing, I'm about to show you our Slack in in in a couple minutes. But the way we work now, 2 months ago versus now, has completely transformed us as a company. And sometimes I take these these little uh these agents away from the team and they get really upset. Okay? And so it's it's I take it away just to see how how useful it is because the more pissed off they get, the more useful I know it is. Okay? And that's happening, right? They're like, "Oh man, Eric, I know I become dependent on this." We had one intern who started with us. They're like, "Oh man, um I I can't live without this." Right? And so this doesn't just tie in with SEO, AEO. This ties in with all the things that we're doing today. The the way in which we work is completely shifted. Okay? So, the way your agents are going to be set up, and I'll get to the AEO SEO stuff in a moment. I I'll talk about how things are changing, but um we think about it as we have six different agents. Okay? So, we name them. We have this otter over here named Alfred. That's our chief of staff, right? Plans the week. We have Hermes over here, which is like a second pair of eyes. We have Oracle over here, which is like an intelligence um that can, you know, handle like your your analytics intelligence. You might have another one. Um my face froze, but you guys can still hear me. Yes. Here, I'll turn my camera on and off. Okay, I'm back. So, um you might have one for analytics, you might have one for sales intelligence, you might have one for creating content. And I I'll give you one example. We have one that creates creative for us. Okay. Um one of the content ones for us on the um the AEO side, SEO side, it it uses the HS API, does the research for us, okay? And then we use our API using ClickFlow, which is our SEO software for content generation. And then we create content. Not only that, we use it for social media too. Social media, we have this this uh agent called Flash. And what Flash does for us is it creates X articles for me in my voice. And some of these articles, I'm telling you, 500,000 views, 350,000 views, 100,000 views, and it generates enterprise leads for us. So my business, Single Grain is the agency side, right? But single brain is a revenue agent side. It's generating leads for both sides. And as I continue to share this stuff more and more, I'm building skills that Flash or any of these agents can use. And these agents get stronger and stronger over time, right? And I don't have that much time. That's a whole another conversation. Um, all I'll say is this is how you need to start thinking about. But if you don't have that one brain talking to tying everything together, this stuff is not going to work cuz it's it's it's too much in isolation, right? It's not ideal. Okay. So, I'm going to move on here. And then what then what you want to do is you want to make sure that every human being on your team has its own fleet. So you might have one for paid media, you might want have one for SEO. So maybe the SEO one, you have a data analyst. Okay, that data analyst pulls from Google search console. It pulls from HF's API. All right, it pulls from uh I mentioned Google Analytics, maybe Google Search Console if I didn't mention it. Um, one handles data an analysis. Another one will then handle SEO strategy. You might have one that then helps with content creation. You might have one that then helps with outreach. You might have another one, okay, that lives within kind of the the SEO side that maybe helps with the Reddit side of things or maybe helps with the YouTube side of things. Thumbnails, headlines, the right topics, looking at trending topics, all these things. What are your competitors doing over there on the YouTube side? All of this can be done for every person. But it all starts with this little thing up here. Okay, so I'm going to read the chats real quick and I'm going to keep moving here. So, um, what brain are HR customers using? Constants, I can't answer that. What is is But it would still be the same thing. It would be whatever that you have uh in your workspace or your dashboard or if you are an agent a user, it would be whatever context or or task you've uh added to your workspace there. And yeah, it will not naturally be talking to any of your other tools like Eric mentioned unless you set it up that way, unless you've sort of had that perspective in the get-go when you use AI. Yeah. All right. Constant looks like a pro. I love it. By the way, Constance is working overtime today. working very hard. I just want to say that. So, thank you, Constance. Um, okay. So, you think about this. You got to think about where your team is at right now. So, some of you are saying you're you're you're not using agent. Some of you are saying you are, right? So, some of you are saying, "Oh, chat GPT is amazing." That's do-it-yourself. Okay. Chat GP, Claude Gemini, that's fine. You can use cloud co like you can use, you know, cloud skills, cloud co-work. Um, oh, and and cloud cloud co-work is more done with you. But you, we all have to think about moving away from do-it-yourself DIY to done with you. Okay. So this is where we have our agents in. You can use cloud co-work. There's there's other stuff that you can use, but you do want to get to a point where the agents are the brain and they're kind of running things for you and you're working with it. You're the you're kind of the human that that's collaborating with it. The true leverage is at the done for you point, which I'm going to show you in a second. The last thing I'm going to show you on these slides um maybe for right now before I do a screen share on uh Slack and then maybe I'll jump back to this is you want to think about this as the individuals on your team who are super AIDS. They're not becoming two times faster, three times faster. They're becoming 10 times faster, maybe even 100 times faster in some cases. Um, even Coinbase earlier this week, they talked about the whole, yes, they had to go through a round of layoffs, but they also talked about this new concept for them known as the oneperson product team. Okay, I'm going to slow down my talking for a moment. So, the oneperson product team, how did product teams look in the past? as early as 3 months ago, you would have a designer, you would have an engineer, and you would have a product manager. Product manager knows what to work on. They have they have, you know, good taste in terms of how to prioritize things. The designer just has good taste in general when it comes to design, hopefully. And then the engineer, the developer makes the build happen. Okay, that used to be three people at the very least. And you think about all the coordination. Oh, it's it's this person needs to talk to this person. That person has this context over here. Oh, this person has these priorities, so it's going to take two more weeks to do it. And by the way, every single week in a year is 2% of the year, right? So when we, oh, it's going to take two weeks long, that's 4% of the year. Are you kidding me? Right? Like so if we are becoming one person product teams according to think about how Coinbase is looking at it you're you're able to get a lot lot more things done faster and in order to get a lot more things done faster you need brains like this working together right and so I'll give you an example this is a like a Slack this is a fake Slack interaction over here okay you can see um I'm tagging Oracle Oracle's for paid media it's doing a data pull right now okay and it's show oh this this is performing well for uh PMAX over here Um, okay. I want to refresh the meta ad creative over here. What should we be working on? I want to see you two coordinate. Okay. I need this meta performance data over here. What are the angles that are actually converting right now? I have Google access. I need to check on meta access. You can see it's going back and forth. And not only that, you might look this is our other agent over here, Picasso. That's our creative agent. It spins up uh ad creatives very quickly. And over here in this Slack interaction from data analysis to creative, that's happening right here. You can do that with SEO as I'm going to I'm going to show you how we're we're we're thinking about it from an SEO standpoint, too. Um, so I'm going to move off these slides for a moment. I'm going to come over to the screen share and I'm going to kind of jump around here. But hopefully you guys are getting this. Drop in the chat if if you guys are kind of getting a sense for where this is going because once all your data is unsiloed and you're able to pull whatever data you want and your team can do that, they're being everyone's being unblocked at the same time. And not only that, you're able to have these agents execute for you. Um, and I'm going to answer some of your questions later. I see some of you asking good questions on what what are you using for agents and things like that. Don't worry, I'm going to get to that. But before I get to that, I'm going to switch my screen share. So, let me stop my share. I'm going to share something else. Let me share my Let's go over here. Okay. Okay. So, we are now and I think I'm sharing this right now. Constant, just let me know if I'm not, but I'm sharing my Slack channel. Okay. This is a Okay. Thank you. So, this is Oracle and we'll show you what Oracle looks like over here. This is a we just we named them after these DC characters. Okay. Okay, so Oracle is our SEO agent. What did I say here? I said, "Hey, uh, use that." So, I dictated. I said, "Use the HFS API." It interpreted as HRS API. But even it, even though I said HRS API, it still knew I was talking about HFS. Okay, I'm going to pull HS keyword trend data across the core agent theme. So, this is what I'm trying to get the Single Grain website to rank for. Okay, I think we all know by now SEO ain't dead, right? Because a lot of these LLM, Reddit, it's it's grounding on uh grounding on Google, right? So u I'm going to pull this data over here. I ran the HFS on the core themes um the HS API on core themes. Uh and I'm grading them from impact confidence east. That's a marketing framework, right? So excuse me. We should stop treating all this one blog. Okay. It's it's basically telling me the strategy. Here's what we should be attacking from an agentic marketing standpoint. Here's why I like it. Here's the the score here. Hey, I think we should be attacking AI sales agents, too. By the way, it's pulling this information, this keyword data from Hrefs, it's also looking at um the data that HRES has on my site, but it's also pulling from the other analytic sources that it's connected to on my site, right? And so all these things over here, if I wanted to ask about, hey, based on what you're seeing from my website from a keyword analysis standpoint or like what am I ranking for now that that's low hanging fruit that we should be attacking? What should I be building more links to? I don't need to. I I don't want to log into a UI and this is my I love HR. HS is a tremendous tool, right? But I guarantee you HRS is going in a direction where their UI is not going to be nearly as important as their APIs and their MCPs because agents need to pull this stuff. Okay? Nobody wants to log into a UI. I don't want to log into a UI. Consistence doesn't want to log into a UI and learn learn new UI. It's it's annoying, right? Especially when we're tired, by the way, right? We just want, hey, can you just do this for me? Right? Everyone wants that. Can you just do this for me? Okay. AI visibility, all this other stuff that we should be going after. Keyword difficulty, all these things. Okay, here I'm going to show you another example. This is one of our our clients. Okay, I'm not going to try to share sensitive information, but he can't. We installed a single brain for him. This one's called Next Diva Brain because there's companies called Next Diva. This is Joe from their SEO team. He's like, "I want all this. I'm not going to show you the chats over here, but he's going back and forth with the agent in here named Nexus." Okay? He's like, "Hey, I want you to provide a summary. Hey, I want you to set up this um I want you to set up auto reminders for this. I want you to do this. Um I want you to, you know, show me internal linking opportunities. I want you to reach out over here. All these things can be done. It can be within Slack. It can be within Microsoft Teams. It can be within a UI if you want. It doesn't matter. The what matters is that you have that brain that everything can integrate into. And then you have people going back and forth, right? There's a lot more within this channel. You can see I'm kind of scrolled up over here, but I see it going back and forth within here's the keyword data over here. Here's what competitors are doing over here. By the way, would you like us to just do this for you? Hey, we think the strategy should be this should be X, Y, and Z. Would you like us to do it for you? Okay, if you want us to do it for you, we're gonna need this. This, right? Point is, you are no longer just one person. You are now 10 people. You're now 100 people. That is where this is going. Okay. Now, a couple things I want to touch on here. Um, one thing is where this is all going. Okay. So, earlier what I just mentioned is, uh, HRS has an API. Okay. And in fact, the API is now available on all plans I believe. Um, and different use cases, right? There's there's an MCP as well. Brand radar. HRS has you taken care of there. Okay, I'm not even shilling HR. They're not even paying me to say this. I'm just saying HRS is a great product. Now, when you think about what's happening, Stripe this week has enabled agents to start paying for things. Okay, there's agents that can buy things on your behalf. Cloudflare has now enabled the ability for agents to purchase Cloudflare infrastructure. So imagine if you wanted to do tests on your domain with different landing pages or maybe you want to do different programmatic tests, something like that. Okay? And then you've created that strategy inside of Slack with your agents. Hey, go out there uh use the Stripe agents and then go and and buy some Cloudflare infrastructure here because I want to get these landing pages up. Maybe I want to get an entire subdomain up and I want to run these uh these these these tests rapidly. You can do that now and everything starts upstream from from the strategy. Get the use the HRS API. Get the strategy over there. Pull from all your other data sources. Maybe pull from your paid media sources too. Pull from everything. Pull from pull from your marketing school like like I'm talking about my podcast, right? Pull from the marketing school podcast on YouTube. Use the YouTube API for that. Use the X API. Pull from all my best articles. What am I talking about there? You see where this is all going? It's no longer just about I'm just an SEO anymore. It's about I am just good at driving good outcomes for my customers. I'm good at driving customers to my business and I'm good at driving customers uh for my customers businesses. And so you no longer need to think about yourself being pigeon holed. And in fact, those of you listening right now who if you have an SEO background, I have an SEO background. SEOs are the most adaptable people by far, the most adaptable. And I think because you're so adaptable, you're able you you are SEOs are are most equipped to deal with this change. Okay? And that's why I'm showing you all this stuff because I I think you all get this, right? You guys are you guys are my people. All right, so I'm going to share one more thing. Um, so I just talked about agents kind of going back and forth with Slack. We talked about the single growing piece. We talked about the HS API. Okay. Um, I think we should also talk about the There's one more thing that I wanted to to bring up. Um, hey Eric, I have a quick question I want to ask. Uh is are all your agents actually uh reading the history of like Slack chats uh that they have shared together when you like put out your first task or is like the history of like previous task or uh like other parts of context for your what your company or like what your client's doing for example is it living somewhere else? Yeah. So there's there's two answers to that. So that the first one for the way I work with it is the reason I like using I use telegram as well with my agents. um same agents by the way and I I like Slack because it makes it multiplayer but also has threaded conversations and so within these threaded conversations it does tend to keep the concept uh the the sorry the context pretty well. Um now when it comes to clients for example we might be pulling from their ASA board right because we use ASA or we use um you know other tools out there that that that have the APIs. Um so we would we if we do have a lot of work that we do with the client inside of their channels we would pull from that channel as well. Um, and by the way, a lot of you might be, some of you, not a lot of you are probably thinking about, okay, what are the security risks here? Okay, so when you think about building a a a single brain, you got to think about that like we call it SSR, which is you want stability, you want security, and you want reliability. Again, we want stable, secure, and reliable. And so for us, we're getting our sock 2 um our sock 2 certification for single brain because that's what we sell to our clients on that side. Oops, I hit my camera over here. Um, so you need to think about that because what we're using right now to answer one of the questions and then I'm going to show you some some other things. We are using Nemo Claw for our clients which is the enterpriseg grade than Nvidia grade version of OpenClaw. We do use OpenClaw internally like I have my own openclaw. We use Hermes agent as well which is actually been a lot more reliable than OpenClaw in the last one to two weeks or so. So these are the agents that we use and then they can start to plug in with the other tools that you have. Um, I would just say that don't let this stuff scare you. It's it's more important than anything for you to just go out there and experiment. Um because it's this conversation wasn't just an AEO SEO one even though that's that's what you know that was supposed to be like that. I'm trying to get you to think even broader. Um and then there's one more thing that I wanted to pull up here which is around um so we talked about AP. Okay. So we talked about agents in Slack. We talked about the HS API. Use it abuse it. In fact, I I'm willing to bet constants that the API you're g you guys are going to make a lot more money off your API calls in the next couple years or so. That's my bet because when you look at Atlassian um their stock just jumped like 30% or so and oh my god the SAS apocalypse, right? But no, they're like oh we're good man. Like we're making way more money off our API usage. Salesforce no different, right? Oh okay. Yeah, we don't need to worry about this by seat anymore. We what's happening is this is known as Jevans paradox, right? When electricity came out, the usage skyrocketed. When you think about um the printing press, when that came out, the the level of literature came out much more than like the previous thousand years, right? And then the the world went up in literatur uh in terms of being uh literate. We'll just leave it at that. Um so I would encourage you if you're working at a SAS company right now, if you have a SAS company, you have to think about building out APIs. You got to think about making things crawable to agents. And I I'll give you a little SEO tip here. So um this is this is Clickflow, right? This is our SEO content uh content generating product and we have an API for it. Right? Literally what we do um is if you scroll to the very bottom, we want to actually is it over here? Let make sure it's over here. I think it's docs.clickflow.com. So this, you know, we have our developer docs over here. We use a tool called Mitlfi for it. But you want to make sure that whatever company you're working at right now, if it's in SAS, you want to make sure that your developer docs are out there so the agents can crawl it and they know how to use your product. And for us with ClickFlow, we're looking at probably having like a a premium version of the API that people can use to create content with it. Um, and you can see all these things you can do, right? And so we have an API for it. And and guess what? I'll give you a little tip here. Mintfi, they're a company that just raised another $50 million at a $500 million valuation. You know what they do? All they do is developer developer content generation. Okay, that's all that they do. And then they have these that link back to their website. Okay, so couple things. Make sure if you're working on a SAS company, if you're making software at all, you got to have APIs. MC maybe you can even think about MCPs. Have the developer documentation out there. Um the single brain piece again. Um I think you you guys should you guys should all get this um in terms of like, oh, you can ask Slack what's our rorowaz this week? Oh, can you can you scale it to 200 variants? All these things you can do. Um the ROI for us has been exponential. Uh let me answer some questions now in the next four minutes. We have uh constants. I see a lot of questions. So I'm assuming that this is a this is okay webinar. So somebody asked by the way um around the I think someone's asking about the agents, but I'm also going to go a little more deeper into that. So when you're when you're thinking about your agents, the model that you use right now we're using more so we're using uh open AIS. Okay. Um and we're using claude. If I I made a video yesterday on how you can how my $12,000 a month spend on tokens, my personal spend, I dropped it to 87 cents a day. Let me tell you how to do that. Okay, pay the $200 a month for the Cloud Max. Okay, or for the OpenAI uh Max subscription, whatever they call it, Pro Switch your Cloud, make a just go to your cloud and say, "Hey, how do I make a long-term like one-year token or something like that?" They'll give you a one-year token. You use that Claude CLI command line interface token. Plug it in. okay to your the clause that you're using one one of the agents that you're using your cost will drop to almost nothing. I was using it so much I went up to like $12,000 a month. Okay. Now, if you run out of usage there, you have a fallback. Open AI. Okay. Use the OpenAI one. That's the next one. Okay. 5.5 or you can switch them vice versa. I I go OpenAI first. Then you can you you can buy local models. I have these Nvidia DGX sparks over here. Number three is Quinn, okay, from Alibaba. And then if I need to fall all the way back to APIs, I will do that. Okay, just to keep that in mind. Okay. All right. So, how do I deal with agent hallucinations, Edgars? I mean, I would just say that this is getting better and better over time. I I think you still need a human in the loop for for judgment. We still have a human in the loop when we're producing content using ClickFlow the the API. Um, we definitely we definitely will look at it. I I think hallucinations are going to be part of the game. Um, so that's that. And Moses, thank you for the takeaways from the meeting. That was good. Um, and oh, Q&A is after 30 minutes. Okay, thank you for that, Constance. I didn't know that. Um, I might jump back into the slides. Who knows? Um, I think maybe we can start. I I checked for the questions that people asked. Craig asked a question about QC checkpoints. What if a data points hallucinated early? How do you check that? Because it could compound. So, like for example, if the first AI bot makes one mistake, uh, does the second one have sort of some guard rails to help make sure that um any issues get checked or catched early? I think I kind of answered that, but it's kind of the same question or the same answer. It's you got to make sure that you're when you're building these things initially that you're checking um along the way and then only by the way when we launch something we're like we're going to all we're going to check it a first few times and then once after we trust it for the first couple of times or after we might just let it we just might just give it a longer leash but we're still going to have a human loop especially if there's a lot of risk ti like a compliance risk tied to it. Yeah. At HRS, we also talk about how it's really important to have a human in a loop. You definitely can't just let like AI run uh and not pay attention to what's going on. Um, yeah. Cool. All right. So, let me see. Maybe you can pop. How do you decide? Well, Bob asked, "How do you decide what's important to put in a brain and what makes it pro proverial junk?" Proverbial junk. Proverbial. No, I got it. I got it. Um, so how can I stop clustering my single brain? Good question. I I would just that that's good. Thank you for that. So the way you want to think about it is just what are your most what what is a job to be done right now and and what are the tools that you use for that on a day-to-day basis right now? You don't need to try to connect everything under the sun. But again, if I think about myself, if I put my SEO hat on, I'm like, yeah, I'm using Google Search Console, Google Analytics. Um maybe I'm using, you know, HRS for the brand radar, for the kind of the the um you know, the the LLM tracking over time. Um and I'm using the HS API, right? Um so that's my jobs to be done if I'm trying to increase organic traffic is just that tool set. Um so you don't need to try to connect everything. In fact, the more you try to connect, the more complex that it gets. Um and you want like there's a lot of scaffolding that you need to build around this, right? That's what we've learned with with building our our own single brain. Um and every company's going to be different. And so when we when we talk to a lot of companies right now, um, a lot of them are mainly looking for two things. They're either looking to scale their ad creatives a lot, which we have an agent for, or they're looking to build out cold email infrastructure and send out cold emails, which our agent actually does that end to end. Like I'm telling you right now, if you connect one of your your your single brain to instantly, which is what we use for um sending emails. Well, what you can do with that is it will literally it will find the domains for you to buy. It will come up with the naming conventions. Okay? it will write the sequences for you and if you wanted to recursively self-improve using like Andre Karpathy's auto research which is a which which will help test it rapidly over time you can do that so you can build a recursively improving loop self-recursive right loop and it's a closed loop and it'll get better and better over time and then eventually what you want to think about is the way in which we all do work it's all about just managing loops okay so if I'm SEO manager I'm no longer just an SEO manager I'm just managing organic traffic loops and then those loops tie together and they they create an outcome ultimately. I know also like um we saw Eric's uh Slack that you see that conversations all happen in a threaded manner. So like that's another way to make sure that like if it's one task is not related to another task, you like split the context. So that seems to be useful way uh to make sure that you're not confusing two things uh to the AI that are not related. Cool. So Edgar says, let's see, I'm curious how you decide what belongs in the single brain versus what should stay outside. So this is looks like the same same question. Okay, the next one, Angelos, what is the best way to manage memory and inter agent communication? That's a great question. So there are a bunch of skills that you can install. So I would recommend looking into this one's not a skill, but there's a GitHub repo made by um the Shopify founder Toby um and it's called QMD. Okay, QMD is good for memory. There's all these little upgrades. Okay, any of you that have played video games or maybe you upgraded your cars or something like that, there's a lot of these upgrades to to think about. So, uh QMD is one of them. Um there's a lot of other memory tools that I I I've added to mine. But I will say that what's important is my agents run on local devices and you want to make sure that your local devices have a lot of um have a lot of memory there because they can get clogged up very quickly. Um, and so when you think there there's a reason why Apple is sold out of their 512 GB Max Studio. There's a reason for that, right? Um, and then I remember I was trying to buy it and then I was like, "Oh, I'm going to wait. Maybe I don't need to spend all this money. Maybe I don't need to spend $10,000 or $15,000 right now. Oh, guess what? Sold out." And then I was like, "Okay, I see the 256. Okay, maybe I'm going to wait a little longer. Wait a little long." Sold out. Now it's down to the 96 GB for the the Mac Studio. And I'm just like, "It's not even worth it. I might as well just wait for the new ones to come out." So memory is a big piece of it. There's a lot of uh memory plugins that you can be adding to it. Um I would also say let's see the best way to manage memory and inter agent communication. Ideally you have each agent focused on doing one thing well. Okay maybe it's an SEO specialist maybe it's a sales specialist. Maybe it is a paid media specialist. But the moment you have them doing too many things that's not good. And I'll give you another one. I remember I had one main agent and you learn these over time by experimenting. and I have one main agent and there's like I had five or six sub agents under it. You don't want to do that either because what happens is context bleed starts to happen and they start um you know one agent's talking to somebody else and then like what was supposed to go to the SEO specialist went to the sales specialist and it goes to the wrong channel. It starts to get really messy and the memory starts to get all messed up too because you clog it with too much stuff. So keep it specialized. Um if you're going to run it on a local device, make sure there's enough memory there. I prefer to have it a local device. That's just my preference because that's, you know, that can be the attack vector. Um, and you know, there's a lot more I can talk about there that from an infrastructure standpoint. Um, so I would say, uh, Kevin, the answer to your question was it right memory as much as possible. Okay? Because I I don't think you're going to be if intelligence is on tap now, you want to have as much infrastructure as possible to make this happen, which is why you see all the big companies out there from a capex standpoint, they're spending so much money, right? Um, so you should look at that as well. I like I never saw myself like I'm looking at my my my D my Nvidia DJX Sparks and my Mac Mini over here. I never thought I'd buy all this stuff and I think we're going to have to buy a lot more um if I'm being honest. All right, so let's see. Uh, what setup do you Oliver asks, what setup do you have for your local devices? Like could you is that something you can talk about like one of your local devices? Yeah, sure. So I'm I'm I'll stare at them and talk about them. Um, so I have a first of all, it's in a rack over here. I never thought I'd own a rack at my home, right? Um, so the Mac Mini is small. It's only like uh 30 something, 32 gigabytes or so. And now I'm like, "Oh man, I really need to upgrade the Mac Mini because one of my main agents lives on it for myself personally. Um, our and then for the the the setup that we have, um, I'll continue on this and then I'll talk about clients, but I have two Nvidia DJX Sparks. So the Mac Mini cost maybe $700 or so. I should definitely upgrade it. Um, I bought two Nvidia DGX Sparks. Each one cost $4,700. So, they're not cheap. Okay. Um, each one of those has about 128 gigabytes of RAM. Um, but they're very strong on on Nvidia uh infrastructure. And these can run local models for you or you can have, you know, like one agent fleet uh living on them, right? And then I I hook them all up to a there's a switch over here and so I I connect them all directly to the internet and so I don't do Wi-Fi for it because you want to be fast and there's a big difference. When I had it on Wi-Fi, it was so slow. you don't want it to do that. There's there's definitely packet loss happening. And then for clients, um we have we have isolated um we have isolated sequences for them, right? So um it's they're all on isolated servers and um there's so there's no context bleed um and we run it on Nemo call. So it's a little different than our our kind of you know setup that I have for myself internally. So Bob asks, "Do you give your agent team a share drive uh to expand without having to keep buying beefy machines?" Yeah. So yes. So my my personal agent uh has read only access to my drive. Um and you can give it its own like there's there's tools you can use like agent mail where they can create their own emails and then they al can also create kind of their their own drives and things like that. So you can definitely set that up. Um, the way I have it set up right now is just it most of the stuff it has access to is from like a critical standpoint like email it can't it can't write. Okay, so imagine by the way you you maybe you can let it read your emails. You don't want to let it write because if you let it write whatever emails and it starts to just go crazy or someone prompt injects it and uh it starts emailing everyone for for like I don't know credit cards or sensitive information um that can be a huge risk, right? So you have to constantly be thinking about security. Um so anyway, that's how I look at it. Uh so a few people are asking about uh memories. I think a better question would be what is like the smallest minimum memory you think that where things will actually be workable and not struggling too hard. I would just say I mean when I had my Mac Mini when I first installed Open Claw in January I got the Mac Mini for it. 32 gigs is fine to start with and then you once you realize that you've kind of discovered Fire you will realize that you need a lot more and then you might be like oh my god Eric how are you spending you know 7500 or $12,000 a month on AI tokens. You are stupid. Well let me tell you something. So, my my uh one of my agents, I had a CFO agent. It said, "Hey, Eric," it looked at all the numbers like, "Hey, uh, you shouldn't be you're overspending in these areas over here. And if you cut these areas, you'll save $500,000. You know what I did in the next three days? I cut 500 grand." Okay. Well, the ROI on that on versus uh 7500, that's 66x. Okay? So, a lot of these things you can do, the reach I'm getting on X, the leads that it's driving. Okay? the the the new service offerings that we have. I'm telling you, everyone wants this type of setup with a single brain. Um, you need to do it for your company and then if you're thinking about doing it for other people, like you might as well figure it out first, right? So, that that's how we're thinking about this. Uh, David asks, "For a lot of technical industries that require factual accuracy, how do you replicate that in creative agents?" Yeah. So, I think I would need more context to your question, but I will say one thing, David. Um, we have clients that work in semiconductors or even do clinical trials, for example. And when our agents use our ClickFlow API to create content, they actually praise us for the content. And these are very these are very picky customers. That's what I would say. Um, and like I would also say that when you talk about factual accuracy, this is especially when it's very very technical, you definitely need humans in the loop. Okay? So when we think about SEO now, I I think the job of editors is more important than ever. Okay? Okay, you're going to need to hire more editors. You think about, let's go to radiologists for a moment. Oh, like, oh, oh my god, radiologists are going to get replaced by AI. But what factually happened was that radiologists didn't have to do scans anymore. So maybe content writing itself, okay, isn't as in in as much demand because AI can can do a lot of that. But what's going to happen is editors are going to need are are in more demand now in terms of reviewing the work. You look at software engineers now, they're not writing code by hand anymore. Um, you think about Enthropic not doing that, right? their engineers are reviewing the work. So, I think we're going to have a lot of uh reviewers like marketers with taste that can review. That's how it's going to go. In fact, constants, I'm curious, h do you know how much of the code that you guys are writing right now is uh handwritten versus um AI? Uh I actually also have a follow-up answer to David on this, but I answer your question first. Most of our teams are already using AI and we're and um you may not know but HF's we uh quite proudly are using Okamel uh as our main programming language as well as things like Rust and other uh like kind of more niche programming languages to to build our databases. I mean build our entire software stack. So it's not uh it actually took us uh not like day one to be able to use AI to do a lot of the programming work but once we got things up to speed and AI got a lot better um our entire team now like I think um for me personally when I was working like maybe uh now five years at HFS from the beginning of like five years ago to like today the speed of which we push out new uh updates is insane like the number of product updates that we cont uh constantly push out uh like improving our API endpoints what we provide uh directly via API and like um new features of ways you can filter or add new graphs or new charts. These were things that like took many weeks or like even months to to come out and like now we have them like multiple times in a single week. So it's really crazy what you can do and that like a lot of this is um because uh a lot of our team members started a lot of the coding by hand. So you have a lot of good examples of like best practices of how you should like implement a particular component or a feature and they use that to train the AI to then create like much better work later on as well as many other best practices that I I as not a programmer cannot talk about. Um, but like factual accuracy is actually a problem not just in SEO, but it's also a problem like internally. You probably have very old documentation and random notes out there that is talking old stuff like outdated like maybe outdated prices or positioning of your company that is no longer true. Things that used to offer in the past that are no longer true. Um, that could be feeding into your single brain if you're trying to architect it yourself. You have to be careful. Um if so if you're creating like um technical content on stuff and you go oh yeah let me just um start with um our documentation online that we push out about for example if you're in the semiconductor industry are they outdated because if they're outdated all your content is going to be outdated so it matters not only and AI is going to be reading that and um like pushing out answering things in the wrong way as well. So uh yeah it there's a lot of work needed to make sure that all of these old documentations clean up for many reasons. I I think it's work. I don't know if it's a lot of work constants because the reality is this you have to set a rule that says okay old content should probably be decaying more and also if you're you have a bunch of skill if you have a skill.md repository the skills that rise to the top you think about Apple's uh Apple's app store for example they don't think about deleting apps that don't get a lot of downloads you just kind of let them sit there but what happens is the apps that rise to the top you know it's it's through signals right and so um but there there definitely work for sure it's like hey what what type of algorithm do we set up is there some type of decay function um with the content that you're you're producing, right? I'm going to show you kind of just to get people to reimagine a little more. I I'll share kind of uh maybe this answers a lot of questions here because people are like, "Oh, well, how are you using it? Is it hallucinating?" Whatever. So, I'm going to give you a couple examples here, right? So, this is me talking to my my Hermes over here. So, earlier this morning, because today my team's doing an AI hackathon, I wanted to explain to them the difference between open and closed loops over here. Okay, I didn't have a good way to explain it. I was like, can you just draw me something? And so it drew me something over here and it explains kind of, hey, what is an endto-end workflow when it comes to AI? What is a closed loop over here? What are open loops? I'm like, guys, we can't live in this old agency world over here. We got to move to end toend workflows for this hackathon and then we're going to move to closed loops where it it recursively self-improves, right? That's where we want to be living. But if if I wanted this done, I got this done two minutes before the the the you know, we appeared for kind of the check-in today. The other thing is on the right side over here, I'm asking for mix panel data pools on um some of our product usage, right? And so it's working over here. It's looking at how uh clickflow usage is happening over the last four weeks or so uh in terms of one of the features that we have over here. And you can see how that's trending over time and based on how it's trending, I can say, okay, where else should I be attacking? Where's the opportunity right now? Uh what are we not doing right? Okay, not only that, um I I there's all this stuff I can do like sometimes I'll read stuff on X using the X API. I'll say, "Hey, uh you know, should we be doing this? I saw this get a lot of reach." Okay, novelty is low. Uh we shouldn't attack this, right? So all that to say, there's so much that you can do here from just having this creative strategies over here. And then you can see some threads are 31 replies, 16, some are 26 over here. Some are infrastructure things. It's incredible the amount. Look, this one has 173 in this one. I didn't even notice this until right now. So my point is if you start doing this now, you are going to gain a lot more leverage. you're going to be way ahead of other I I don't even think you're top 1%. I think you're top.1%. Okay, you can use any of the tools that I mentioned, but the whole concept of having that single brain unlock you. That is the whole point of this webinar. Hey, uh anyone else has uh oh, there's one other question uh from Edgar. When agents pull from multiple tools, how do you handle conflicts between sources? still similar uh question honestly con I'll answer that because there's similar questions popping up over and over honestly once you all start to play yeah once you all start to play with this it it's going to like you will see it's it's not it's not a big deal I haven't seen many issues with the work that I'm doing from a hallucination standpoint and we're constantly checking it like I remember um for one of our other clients using single brain right now they want to use it for LinkedIn so we connected it to the LinkedIn API and we kept pressing it with different questions we we we kept testing it And um the paid media manager who was working on it was like, "Yeah, this is all factually correct and this makes my life a lot easier." And think about it this way, the way you see me using it within Slack right now, MetaMCP just came out, okay? Uh Higsfield MCP just came out for, you know, creating uh creatives at scale. Um Motion MCP came out, so you can see kind of uh what ads are performing well. If you combine all these things together with your ad accounts and you can push these ads out at any given time or double down, triple down on the the variations that are working for you that you that would have been impossible for you yourself as a person long time ago, right? So, I'm just saying look if you by the way you might be thinking, oh my god, like uh how do I stay on top of this stuff? I feel like I'm behind. Well, that's okay because I feel like I'm behind. I feel like I'm behind every single day and whenever I look at X, tons of information coming in, right? So, if you want to be at the forefront of this, you're hanging out on X, but more importantly, you're going out there and you are doing. Okay. So, um Moses asked, "For an SEO agency, what would be the first few agents you'd recommend focus on?" Uh, Payton, X is the worst place, but best place to learn. If you filter it well, Payton, you'll make it good. Okay. Um, so Moses asked, "Fro for an SEO agency, what would be the first few agents you recommend focusing on?" Okay. Let's keep it simple. really for any marketer, what do you spend a lot of your time on? If you're an agency, you said agency, right? Oh my god, you spend so much time on reporting, right? The data pulling, the reporting, the different sources, the uh the strategizing around it, the thinking around the information, and then to consolidate it all for your clients. Reporting was taking our team a lot of time. I would I was I I was hurt with a reporting agent. I'll tell you that much. I would I would even just say look at all the time you're spending on things and look at all the money you're spending on things and just think about okay what parts do we actually do really well right now that can we actually start to hand off right and um you know love or hate Elon he has a nice algorithm you can use here Qessa QD QD SSA okay I'm I'm going to repeat it so you want to question the requirements okay you want to delete that's very important you want to delete things then you want to simplify things. Speed up and then accelerate or automate rather. Okay, again I'll say it again. Question, delete, simplify, speed up, automate. Most people think about automation too early. So you got to get nail down the process first as as as humans. Then you want might want to delete some things, then you simplify, then you speed it up, then you automate it. That is going to make your life a lot easier, Moses. So data analyst piece, um I would say the reporting piece, which kind of goes into data analysis. Um I think using the HRS API to help with all that. Um, and I think keyword like you can put keyword research into data analysis as well if you're doing that, right? There's a lot of things to to think about. Um, from a AEO standpoint, pull in the HS API because you're going to need that, right? It makes things a lot easier. So, audit agent. Yes, absolutely, Philip. Thanks, Constance. Go ahead. Uh, I uh if you don't know how to begin creating your own agent, uh maybe Eric can add to this, but uh I actually asked AI that I want to create my own agent. Um, I have this one task, but I'm very specific. I tell AI that I have this one task that I really want to automate so you can save me time and these are like the steps. Like it it doesn't have to be complete either. Just tell them like as much as best as you can like what you want this agent to do and then you ask AI for recommendations of how to go about building it and AI will tell you like whatever AI you use it like this they can give you a starting point of how to start building your your own thing. And that's kind of to me the best part of AI is that you can ask literally any question of like um how you want to go about doing something. Uh and you'll get like a pretty good like starting base. Cool. 100%. Okay. I'm going to answer this Payton question here. So in terms of file structure, how much do you let your agent create versus you make the path and it follows it? So I would say in terms of file structure for OpenClaw or Hermes, we just kind of we we let it do its thing. um for clients we maybe we'll have like a we'll have a folder for the client. Um I would say when I'm building things using cloud code or codeex I do have I build locally and I do have a file structure for for that. Um and and plus Payton the other thing is when I'm using open cloud Hermes it does um back things up into GitHub and it does organize it accordingly and whenever you feel like the file structure is is uh it's not sanitary or it's not clean just tell it to to to Constance's point just say hey can you clean up our file structure or can you make it more uh coherent because when I say hey go into my Google Drive and do it it does it and it does it in a very logical way. Um, and my brain doesn't like to think about all that, right? It does organize it pretty well. And then the other thing you mentioned here is you're using Hermes plus Kimmy 2.6 with O Lama. So, I think that's great. I think Kimmy 2.6 is amazing. That's one of the the models the the models out there. Um, and then using with Olama. That's I think it's good. My whole thing is using Kimmy or using uh Quinn or using any of these other ones. Um, I like using the Frontier models, whether it's Claude Opus or it's um, we're talking about Chat GBD 5.5. The reason for that is because whenever I've tried switching back to when I tried switching over to Quinn or when I tried these other models, they're solid, but you can tell there's enough of a difference there. So, my whole thing is this. If I'm going to hire someone to do a job, if they have to redo the job, even though they're cheaper, if they have to redo the job two times or three times, I rather just go with the Frontier models. And that's why the what I told you earlier, if you're paying a 200 bucks a month for Claude, they're very generous with the the the the token allocation, right? I think they're saying that for every $200 you spend on Claude, they actually reimburse like you're actually getting like five or $6,000 worth of usage. In my case, I I think $12,000 worth of usage, maybe even more. Um, so I would recommend still having one or two of the Frontier models. uh and for like heavy work and then maybe for the the that's for like Ferrari work and then for maybe Honda Honda Honda Accord or Toyota Camry which are great cars very reliable right you can have it on kind of these more uh open models and no Bob they did not ban OOTH usage uh they they actually brought it back so you can have a so what I mentioned earlier cloud CLI you can you can connect it um and so yeah So there there Bob's like, "Oh, damn." Right. So yeah, go ahead and check it out. I'm P. I'm glad that you use it for sure. That's great. Okay. Are there any other questions that I think we answered, we covered pretty much everything uh that people have asked. Angel. Yep. Do you use agents to design and develop enterprise websites? Any suggestions here? Oh, dude, this is a great place to end it. Here, I'll I'll show you right now. Okay. So here you all see this. So I deployed this site two days ago. All right. But literally, okay, this this logo and all these things, one trusted BL brain per client. All these the marketing partner companies cannot outgrow. Here's a simple chain of command. All these things. Okay. You know what I did here? By the way, like SSR, what I mentioned, security, stability, reliability. This website was deployed um two days ago. And my point of calling this one out is I leverage claw design for the design. And I also said, "Hey, I like the designs of the aesthetics of the Finn.ai website. I like this one. I like this one. Go ahead and combine it. And I want you to make me a new single brain website." Okay. Not only that, it designed it, but then I went to I went to my my agents. I went to OpenClaw. I think in this case, I went to OpenClaw. I said, "Hey, can you actually deploy this? We already have a single brain website that's up right now." Okay, so the single we already had a single brain website that was that was kind of vibe coded. It it clearly looked vivecoded and um but it was converting and it was driving enterprise leads already. So that it didn't matter that it looked vcoded. It was driving leads. Okay, we're talking about multi-billion dollar companies here. Um and so I was like I I want to make it look a little better. And so not we got the design done. Okay, that's fine. But how about actually deploying this website? Cuz I'm not I'm not an engineer, but I don't I also don't want to wait for my engineers too. But and I feel like this work is kind of beneath my engineers now. And so what I said was, "Hey, um, help me deploy." And I said, "Okay, I'm going to help you deploy. I'm going to help you test it. I'm going to make sure that the Google Analytics pixels are firing. I'm going to make sure that all the conversion stuff that's firing over to your HubSpot, it's firing over to your Slack. Everything's tied together. So if you want to go end to end design, okay, and then deploy and then you have the product sense in the very beginning, you can literally do that. And that goes that encompasses this entire talk. You can do that end to end now. And that's the beauty of this stuff. Um, so hopefully that helps you a little bit. Um, patron, that's not on go high level. Um, that's just me using kind of the the tools that we have. Um, and then Moses, you helped me build my first agents of make.com during San Diego HFS times leveled up. I had no idea. I hope maybe Constance helped with that. I don't think I helped with that. Uh, I wasn't at the conference actually. No, but like I'm sure like your sharing probably got him inspired to create his first a Yeah, Versel. This is on patron. This is on Versel. And so here's the thing. I I don't I didn't have to log into anything for this. I think I just had to give it a few keys and the fact that it was able to carry over all the conversion stuff over and was I was just worried about it. And then I was giving it feedback a little bit. I made like a loom and I gave it feedback over and over. Uh and yes, I'm using Versell. Uh so anyway, that's all. And then finally, do you plan to do more online events about AI agents in the future? That's a constance question. The the demand for AI related content is off the charts. is basically what anyone wants to watch right now. So yeah, I think uh for example uh on 19th and 22nd or 21st uh this month I'll be talking about mistakes that marketers typically make with AI and how to fix them. And I'll be showcasing a bunch of like I've basically pulled in all the mistakes that our marketing team so these are people that are not necessarily programmers or engineers though some of them kind of are. And uh and when we use AI, we're telling you all like the mistakes that we've done in the past and ways that you can basically avoid um doing our same mistakes in the future. So I'll be doing a session on that uh both one for the US time zone as well as one for Apac Europe. Okay. So um yes, there will be more. I mean um you'll there's no end a shortage of ways that AI can like you know basically change your life and your company's life. All right. Well, thank you Payton for that. Uh yeah. A AI has took me to another I think AI has taken us all to another level. So I I appreciate that. And Constance, I'll tell you one final thing before we go here. I'll let you wrap it up for us. But uh uh I remember two years ago I did a live interview. So um Tim Solo, CMO of HRS was great grace uh graceful enough to um speak at my uh this is last year actually. This was last year during this time. So he spoke at my global AI summit and um the funny thing is we did a podcast after so I was on the HRS podcast and then uh we were debating AI. He's we were arguing with each other about the usefulness of AI. He took the other side where it's like maybe it's a little overhyped. But now we all know now it's not overhyped. So Tim, I win. Thank you. So I'll be texting Tim after this. We're we're we're we're friends. Um but Constance, thanks so much for having me. Um everyone else, uh thank you for joining. Thank you for your attention. Go ahead. Thank you everyone for joining. Um like you can definitely follow um Eric on the socials of course. Um as well as um follow us for the next few um sessions that are gonna happen. Again, we're going to keep talking about AI, guys. It's going to be even more ways that you can like level up your work as well as your output. So, thank you so much for joining us. And for the people who are watching this recording, we hope to see you guys uh live next time um with us on these uh live sessions. Have a good rest of the week and uh we see you guys later. Take care. Bye bye. Yes, we'll uh we'll send a recording in the link. It's uh uh it's on YouTube. So,