These AI Marketing Agents Get You Customers

I bring Cody Schneider back on the show to build two marketing agents end to end, live. The first one monitors LinkedIn posts from creators in your category, scrapes everyone who engages, waterfalls those profiles into emails and phone numbers, and then runs cold email and LinkedIn DMs with an agent managing the replies. The second one turns internal conversations, sales calls, and podcast transcripts into a daily organic LinkedIn content engine across an entire team. Cody names every tool in the stack, shares the real infrastructure costs, and shows the actual terminal commands he runs in Claude Code. By the end you have two systems you can go set up today for your startup. Timestamps 00:00 – Intro 02:27 – Agent Number One: Cold Outbound Agent 04:26 – Finding Creators in Your Category on LinkedIn 09:09 – Apify Explained and the API Maestro Actors 10:59 – Extracting Engagers Live in Claude Code 12:59 – Agent Versus Automation 15:45 – Waterfall Enrichment: GitLeads, Apollo, Origami 17:13 – Compliance, Data Brokers, and What Stays Legal 21:40 – Waterfall Enrichment: Million Verifier and LeadMagic 25:38 – The Cold Outbound Infrastructure 28:33 – Software Factories and Marketing as Code 31:41 – Agent Number Two: The Organic LinkedIn Engine 39:34 – Earned Media Math at $22 CPM 42:31 – Closing Thoughts Key Points LinkedIn engagement is a hand raise, so it beats firmographics as a targeting signal. Ten to twenty source accounts give you roughly 80% surface area coverage of an industry. Waterfall enrichment moves cheapest to most expensive: GitLeads, then Apollo, then Origami or Prospeo. Roughly $200 a month covers sending software plus inboxes for about 10,000 cold emails. An agent here is plain code on a cron job with an LLM attached where judgment is needed. Organic content works best when it starts from real human source material like calls, Slack, and transcripts. The #1 tool to find startup ideas/trends - https://www.ideabrowser.com LCA helps Fortune 500s and fast-growing startups build their future - from Warner Music to Fortnite to Dropbox. We turn 'what if' into reality with AI, apps, and next-gen products https://latecheckout.agency/ The Vibe Marketer - Resources for people into vibe marketing/marketing with AI: https://www.thevibemarketer.com/ FIND ME ON SOCIAL X/Twitter: https://twitter.com/gregisenberg Instagram: https://instagram.com/gregisenberg/ LinkedIn: https://www.linkedin.com/in/gisenberg/ FIND CODY ON SOCIAL: Cody’s startup: https://www.graphed.com/ X/Twitter: https://x.com/codyschneiderxx Youtube: https://www.youtube.com/@codyschneiderx

• 0:00Speaker 00

  It's true. Marketing agents are the new coding agents. Just like coding agents were such a big deal and people were able to create software on demand. Deploying marketing agents are so important because you're able to get customers on autopilot. So how do you actually set them up? What do they look like? Well, this has got to be my most requested episode in a long time. I bring back Cody Schneider and he shares all the sauce. How you can use Codex or Claude Code to build these. What are the other 20 tools that you need for the marketing infrastructure? In order to deploy these marketing agents? And by the end of this episode, you're going to get your creative juices flowing around some of these growth tactics that are going to help you stand out, that are going to help you get customers.

• 0:45Speaker 00

  So that whatever it is you're building, you don't have to worry too much about traffic. You don't have to worry about too much about revenue and you can focus on building an incredible product. While your marketing machine is running. Enjoy the episode.

• 1:07Speaker 01

  Welcome to Greg Eisenberg's podcast called Sip Baby. I'm your co-host or guest today, not co-host. I'm never the co-host. I'm Cody Schneider. I'm going to be your guest today and today I'm going to teach you how to build an AI agent that does cold outbound, both on email and on LinkedIn. This is based off of the comments from last video. If you want to learn other go to market motions, you need to comment below right now. Do it right now. It also helps us for the algorithms. So you're supporting the show and it keeps the lights on here.

• 1:37Speaker 00

  Welcome to the show, Cody. Marketing agents are the new coding agents. We only shared one marketing agent last episode, but the people aren't satisfied with one. So you need to come back on. You came back quickly. And by the end of this episode, you're not going to share one end to end marketing agent, right? You're going to share two marketing agents, how people could set it up. So by the end of this episode, people can go stop the video and actually go set this up and actually get customers to their vibe coded startup, right?

• 2:10Speaker 01

  This is exactly what I'm promising you today. You're going to have two of these in a while. I'm going to teach you everything that you need to know. I'm also going to share all the tools that you need. There's no gatekeeping here. I despise people that do this. Don't buy a course, literally DM me, I'll teach you anything. I'll just make a public video for everybody. So let's do it, G.

• 2:28Speaker 00

  All right, let's run it.

• 2:30Speaker 01

  Awesome, man. All right, so today we're going to build a system that basically monitors LinkedIn posts of influencers within your niche, within your category. And then it's going to go and extract the engagers from those posts. And then we're going to do what's called a waterfall enrichment to find the emails. And even potentially the phone numbers of these people. So that you can then go and do an outbound motion to them, doing cold email and then also doing LinkedIn DMs. So that's what is going to happen. And then I'm going to teach you how to basically have it. So you can have an agent that's wired up to both of those inboxes. Like managing those inboxes, say for example, answering questions or trying to push them into like booking a demo with you as an example. So yeah, man, that's, that's really it.

• 3:16Speaker 01

  The, I don't know if there's any other like specifications on the high level. I think the only thing to mention with this is like the strategy around this. So right now, cold email is getting decimated. Reply rates are down. Everything is down. Actually, every marketing channel is down right now. Let's be, let's be real. The reason is just because like AI slop is flooding the zone and it's becoming just red ocean everywhere. But the way that we have found that you can stand out is you have to look for signals or triggers that basically show that people are hand raising, saying, hey, I want, I want this thing. I have an interest in this thing, right? And a great way to do this is with these LinkedIn engagements.

• 3:57Speaker 01

  They're basically, when they like content, that is a hand raise or a signal that I am interested in this, you know, specific thing. And from that, we can use that as a way to measure, okay, is this my target customer that I'm trying to sell to? And not just like their firmographics or their demographics or their psychographics, which is like what we would traditionally use for, for outbound. This is specifically like, no, they, they have a propensity or an interest in this topic. And we are going to go and now get in front of them. Okay, so how do we actually do this? And this is an exact strategy that we implement for, you know, the companies that we're working with. So I'm going to teach you that right now. So let me screen share and I'm going to walk through it. So the first thing that you're going to want to go to do is literally go to LinkedIn

• 4:41Speaker 01

  and find influencers within your category. So last episode, we talked about AI for WordPress or AI WordPress. And so I'm just going to use this again as an, the example, you know, like target demographic that we're going after. So on LinkedIn, what I would go do is I would go try and find people that are talking about WordPress development potentially. Let's see what comes up with that. Development.

• 5:11Speaker 01

  And I would try to find posts. This might actually be a terrible category, so we might have to explore something entirely different. But I would try to find posts or creators that are talking about these specific topics, like on a daily cadence, right? So like this, like again, just for this example today, this is probably going to be like a lot of like just not good signal. So a better way to look at this is like, we'll say, we'll do AI for our AI marketing, right?

• 5:42Speaker 01

  Let's see what comes up and we're going to try these, find these posts here. So.

• 5:46Speaker 00

  And what makes a good search? Like why, why was AI for WordPress not good? And why is AI marketing better?

• 5:52Speaker 01

  Yeah, so it's really just like, is the content that's being served, what your target customer would be interacting with? Like that's what you're trying to get down to here, right? So like, how I would be going through this and honestly, I use the, the for you page of all these algorithms is so good now that it's like, it's going to show you the content that's relevant, right? Like this is literally an exact, like perfect, like perfect example. First one that comes off, it's like awesome, people trying to do some type of video editing for, it's probably for marketing. Everybody that's potentially engaging with this is like a target customer, right? So I would say, okay, cool. I'm going to find these creators and then I'm going to build a spreadsheet of all of them, right?

• 6:34Speaker 01

  Like all of these people that I'm going to try, that I'm going to source these leads from. So right, I would build this spreadsheet out and we'll just do a handful of these like from my own feed. It can even be business accounts. And I think this is the thing that people don't realize. Like if there's business accounts that the people would be interacting with, that would be your target customer, that can work as well, right? So it can be literally Clay. And we're going to do the posts from Clay and we'll just keep going down on this. So MCP, it's probably too broad.

• 7:07Speaker 00

  And you're doing this manually, like you're not using agents to do this, why?

• 7:10Speaker 01

  I wouldn't even, typically the company knows who is interacting, like when we're working with a business, right? They know who they're like, their target customer is interacting with, right? So you can, all you need is typically like 10, 10 to 20 of these. And you have more than enough to be able to like source the lead volume that's necessary to actually make this like a viable channel.

• 7:35Speaker 01

  You, I'm using the feed here because like what it's going to show you is what is most like relevant to you. So it's probably going to be stuff that's, you know, in your, the niche that you're in. But you can also use the search for this as well. We used to do this where we'd like do the search and we'd find the trending posts from that period. In reality, it's like, there is a handful of outliers within any niche. And everybody is engaging with those handful of outliers. If you just monitor those outliers, you're actually going to get, you know, 80% surface area coverage for that entire industry. You don't need more than that, right? Or it's, it's, it's just like the marginal return of trying to go for all of it. It's not, it's not there for, for that system.

• 8:17Speaker 01

  This is the same idea with, we do this a lot, like we try to solve entropy, this entropy problem with, within like ads, paid ads in particular. Where it's like, if you just have the agent like go in this loop, it'll just kind of make the same ideas over and over again. How do you solve for that? Well, you find human creators, like 10 of them on Instagram. And you track the content that they're, they're publishing. You look for the outliers and then from that you typically can get signal of like, oh, here's this new hook format. Or here's this new topic and I can just pull that, I can remix that. And that's the way to do this. So, all right, I find a handful of these, these, these companies. And then from that, what I'll go and do, and just for the sake of, you know, the example today, we use Louise as an example.

• 9:03Speaker 01

  So we'll say, everybody. That interacted with this post. We're gonna use this post as an example. So, um, once I have these people, I need to use Appify. And I will find the actual one that we like.

• 9:17Speaker 00

  And what's Appify for people who don't know?

• 9:19Speaker 01

  Yeah, so Appify is a scraping API. So I can use a single API key. And then I can use it to scrape LinkedIn. Um, I can use it to scrape, uh, Twitter. I can use it to scrape all of these different channels. So it's a way for me to get data into the context for my agent so that it can have, you know, awareness, uh, and have that context for it to make decisions on or make content based off of, et cetera. So. Okay. So the one that you're gonna wanna use or the one that we like, we've worked with him like a decent amount because it's the most stable connections. There's tons of these and the challenge with Appify is finding good ones that are actually, um, uh, like being monitored and being maintained.

• 9:58Speaker 01

  And so this, uh, uh, this guy API maestro has a ton of these for LinkedIn. You can see all of these here. It's all of these different functions that you can do. So how Appify functions is you get an API from Appify. And then this enables for you to be able to have your coding agent like Claude Code or Codex. Call from, uh, Appify or call the app through the Appify API to one of these endpoints that are here. So, uh, for example, you can do this post scraper, uh, for the one that we're going to do, it's going to be engagements. So let me find that, uh, post reactions on LinkedIn.

• 10:38Speaker 01

  Believe this is it. This is exactly it. Yep. So post comments and then post reactions are the two that you're going to use. And what this enables you to do is everybody that has engaged with that post. So the post that we are just looking at here. So everybody that's interacted with this and commented on this, we're gonna be able to pull this out. I'm gonna show you how you can actually do this in, uh, uh, Claude Code right now. So I'm just gonna spin up a terminal real quick and let me reshare, uh, my screen. And so I have that, uh, I have that Appify API key, um, in, uh, already saved locally within the directory that I work out of. For, uh, all of my growth work. And if you don't know what I'm talking about here, I have a whole video on my, uh, channel.

• 11:20Speaker 01

  That's basically a crash course into how to do this. It's called go to market engineering or marketing engineering. And it will walk through the entire setup process. Takes about 10 minutes, but basically this Appify API key is shared here. And I'm already written this script. I had the agent going to read, how do I use this endpoint? To pull out all of the posts and comments information, though, all, all the people that have interacted with this. So I can give it this post URL and I can say, extract the engagers using, uh, the Appify API key.

• 11:53Speaker 01

  And it's going to go and run that process for me. So this is how I would go and build this automation or build this agent. As I would basically take this code and I would deploy it into the cloud. And I would say, okay, on a daily cadence, I want you to check for net new posts. So that is where I would look at the profile posts. So this is the profile post scraper. So I would extract the post URLs from this person. Right? So every net new post daily is getting extracted. And then from that, I'm then extracting the engagers.

• 12:26Speaker 01

  Using that API endpoint as well. Right? So right now, as you can see, they're duped by public profiles. There's 63 raw. And it's about to pull all of those contacts out. So once I have those contacts, this is, this is done, man. Like game over. As long as you have the LinkedIn profiles, you can go and find the email addresses of them. You can find the phone numbers of them. You can find everything that you need on the cold outbound. And I'm gonna show you that right now. What are the tools to actually go and use to do this? So let me just show you though, again, just to, uh, the final completion of this.

• 13:00Speaker 00

  And what makes this an marketing agent versus a marketing automation?

• 13:05Speaker 01

  Yeah. So the agent component of this is that it is running on a cron job daily. And then you're gonna have an agent that's, uh, later on, we'll have it responding to the inbox. And this is this blurry line, right? Like what is an agent? People ask, this is gonna be this every sales call. And the answer to all of this is like, it's how I think about it personally. Is it something that's doing a job to be done? Right? So the job to be done here is finding leads. And outbounding to those leads and then responding to those leads as they're like asking questions or again, like driving them deeper into the pipeline. In reality though, gee, like what is a market, like what is a marketing agent?

• 13:47Speaker 01

  It's, it's code. It's maybe some thinking loop and it's a live data stream, right? That that is really how like this functions. And the thing that you can make, you know, extend this further with is like what you're, who you're outbounding to. You want it to basically do an ICP fits or it's or a, a target customer segment fit. So before it even does this enrichment that we're about to do, you would be like, okay, agent, research this person and the company that they're at. How many employees do they have? All of these things. And then based off of what we find, if it fits this customer profile, like it's, you're gonna have the agent basically think through that, right? Using an LLM. If it fits this customer profile, then it goes into this enrichment. Then we're actually gonna cold email them.

• 14:31Speaker 01

  So that's where that thinking loop could potentially be here as well. But really the, the blurriness between all this, I think about it as software anymore. Like to be transparent, like everybody, the thing, a different way to say this is like, everybody tried to put God in a box and give it access to a Facebook ads account. And we realized that is not the right way to do this whatsoever. The right way to do this is like, what was the human doing? They were running this very specific process with like media buying. They were researching ad creative angles. They were making new ad creative. They were testing the new ad creative. And then they were like pruning the losers, promoting the winners, right? Like that is what the, a top media buyer does. Okay. How do we go and make a piece of software that does that exact same thing?

• 15:12Speaker 01

  So when you hear agents like really just think software with potentially a thinking loop, like you shouldn't be paying a different like way to think about this. And this is something I'm obsessed with right now. You should not be paying Anthropic. You should not be paying chat GPT to do an API call. You should be paying them to make the software that uses CPU to do the API call. Why are you paying this tax on tokens every time that you're trying to do this marketing activity? That's ridiculous. Build the software that does the solution for you, not tokens burning every time that you're trying to do the action. So anyway. Okay. So we've got these LinkedIn URLs and what do we do with them now? So we're gonna do what's called a waterfall enrichment. And so we're basically gonna use these LinkedIn profiles to go and find the email addresses and then the phone numbers of these individuals. So.

• 15:57Speaker 01

  How do we do this? The first thing that we're gonna use in a tool stack is called gitleads.io. So this is a database of, uh, it's basically they aggregate all of these B to B contacts and you can access it via their API.

• 16:13Speaker 01

  The emails that we don't find within gitleads, we're then gonna use some, or then get a waterfall down to something like Apollo. And then you could take this even further down into something like Origami. It's another tool that we, uh, have been using and experimenting with. Also their team is just doing awesome work, like Finn and his whole team is incredible. So anyways, for gitleads, let's go back to our, uh, uh, uh, a terminal right now. So again, this is me hands on keyboard, doing the process to teach it to you. But everything that I'm doing right now, this is all just going to be code under the hood. And once it's code, I can deploy that into a cloud system as long as it has the necessary data that it needs. And the necessary access that it needs, it can go and run this operation autonomously.

• 16:56Speaker 01

  And then you're just there basically jockeying the agent or modifying the system, right? So we're building a system here. So from here, um, what I would then go do is say, use the gitleads API.

• 17:09Speaker 01

  To, uh, find the emails and phone numbers.

• 17:13Speaker 00

  And like dumb question.

• 17:15Speaker 01

  Yeah.

• 17:15Speaker 00

  That's legit. Like, you know, like. It's not gray to get these people's emails. It's like fully legit.

• 17:24Speaker 01

  It is fully legit to get these emails. Um, what you do with those, that's where things, uh, like from a compliance standpoint change. You can cold email technically in the United States. You can also add people to a email newsletter. To be and be CAN SPAM compliant. There's like tons of, uh, like thing, you basically have a checklist of things that you have to do. With this said though, um, like this is one of these, like on the cold email side and the contact lookup. You're basically just buying data from a data broker, which is, uh, is legal, right? That, that is accessible. So these companies, how they do this is they basically are buying all these lists and then aggregating them from all these different data brokers.

• 18:05Speaker 01

  That whole piece is, it's all through shady network, but this, uh, like what we're talking about here, uh, you know, on the spectrum of like black hat to white hat is pretty far on that white hat side. So.

• 18:16Speaker 00

  Cool. Yeah. I mean, I don't think anyone would, you know, mistake you for a lawyer also.

• 18:21Speaker 01

  No, totally. Take this with a grain of salt, you know, and, and like, there's also different compliance rules within the United States.

• 18:27Speaker 00

  Do your own research.

• 18:29Speaker 01

  Exactly, exactly. The, uh, within, you know, the United States versus, uh, like the EU has totally different compliance pieces.

• 18:36Speaker 00

  Exactly.

• 18:37Speaker 01

  But with that said, like the, uh, you know, the finding of people's information and then like reaching out to them, uh, there, you, you can do this basically. It's kind of the high level. But again, this, I, I, we don't have time today to go into all of the, the specifics about like. All this, the, the finite details here. So once I've found this, um, each of these individuals and then the emails. From there, what I'm then going to do is validate these emails. So I would send it to a software called Million Verifier. So Million Verifier, um, enables me to, uh, basically check if the email is good, risky or bad. You know, uh, more technical terms would be, uh, like good, catch all, um, you know, risky, et cetera.

• 19:21Speaker 01

  The, the reasoning for this, or the reason you have, you wanna do this is the emails that come out of these providers. So out of gitleads, out of Apollo.

• 19:32Speaker 01

  Out of Origami, I think they do some checks like a little bit deeper though. So I don't know as much about this, but I know for sure with GitLeads and Apollo, it's like do the second verification. You're basically only wanting to send cold email to valid emails. Because if you send to invalid emails, you're going to basically just run into deliverability problems. And probably right now, you're asking yourself like, okay, cool. How do you send these cold emails? I'm going to show you that in a second. Bear with me. So we've done that waterfall enrichment. We found the emails. We found the phone numbers. And when I say a waterfall enrichment, what is happening here is we're taking that list of 50. And just to use this spreadsheet as an example, so say we have, you know, 50 that we have,

• 20:16Speaker 01

  50 LinkedIn URLs that we found. And on GitLeads, maybe we only find, you know, 32 emails of those people, right? So that next cohort, so those other 18 that are left, I'm then going to send those 18 to Apollo. So of those 18 that I send, maybe I only find 10. And then those eight, that's when I would send that to something else like Prosperio or Origami or these other enrichment tools. And the reasoning behind this is you're, you're starting with what is the cheapest, most accurate, and then moving your way down into the more expensive validation tools. But from this, you can pull out basically from a list, like, you know, this is the

• 21:01Speaker 01

  way that you get to, you know, an 80% fine rate, et cetera. And you can chain as many of these together as you want. It just, you know, depends on your budgets that are available, et cetera. And there's also aggregators of this, like Origami is an example, like aggregates this waterfall for you. So you can just send them a LinkedIn profile and it's going to like waterfall through the options that are available. Um, okay. So the other, other thing to throw in here that will be valuable to your team is a software called Lead Magic. So this is one that I, we use a lot for like mobile phones in particular. But same strategy here, it's just basically, you know, another enrichment tool, but specifically on the phone number side, we've used a decent amount. So, once I have that contact information, I now need to go and actually build this outbound motion.

• 21:45Speaker 01

  So on the cold email side first, how do we go and do this? We need to buy inboxes. So a couple of different ways to do that. I can use a tool called InboxKit. I can use Instantly AI's pre-built, uh, like emails that you can buy from them. Or I can use a company called Hypertide, which is the partner that we use and we work with. They are some of the best in front, in my opinion. So when you're buying these emails, uh, you're buying, or really what you're doing is you're buying inboxes and domains that are burner domains that enable you to send cold email, not from your core domain. And the reason that you have to do this is so that you don't burn the deliverability

• 22:30Speaker 01

  of your core domain. So what do I mean by that? If you send from, you know, your exact domain, um, and, uh, you know, say we send 10,000 cold emails from that, um, we will nuke the deliverability of the business URL, the actual domain that we use to, you know, run our company, right? You don't want to do that. So typically what you want to do on the marketing side is have this, have this separation. So you have domains that are for your cold email. You have domains that are for your email marketing. You have domains that are for your transactional marketing. So this would be, or sorry, our transactional email. So this would be email that's being sent directly from the product to a customer. Imagine like a password reset as an example. And then you want to have your business, you know, domain email, which is what your

• 23:14Speaker 01

  team actually uses to run the company, et cetera. Um, so with Hypertide as an example, um, we, we have a partnership with them. So it's a little bit different, but, uh, we can send about 10,000 cold emails, uh, just to give a, uh, you know, kind of the cost breakdown here. We sent about 10,000 cold emails with them for about a hundred dollars a month in infrastructure costs on the inbox side. Um, it's about the same for majority of these. So InboxKit as an example is very similar pricing. And they also run like sales all the time. So look for those on the domain side. So you basically buy the domains and then you're paying a subscription to have these inboxes hosted for you. And then on Instantly side, uh, you can typically get started with this $97 a month tier.

• 23:56Speaker 01

  So in total, you know, out the door to get going on this, the infrastructure costs can be in that range of about a hundred dollars to get started, or sorry, about $200 to get started for the sending, uh, software and then also the inboxes. So again, just to reiterate this, because I know I've talked through a lot. I'm pulling the lead list from LinkedIn. I'm finding these people. How do I know that these are people that I want to reach out to? It's because they're engaging with content that I know my target customer would be interested in. And so these people are basically hand raising that they are, would potentially be my target customer.

• 24:29Speaker 00

  Which is insane, by the way. Right? Which is insane.

• 24:34Speaker 01

  So hard to find this, right?

• 24:35Speaker 00

  Yeah.

• 24:37Speaker 01

  It's impossible to find this. And so the, uh, so I'm finding these people. I'm then doing a waterfall enrichment to find all of their contact information. And then once I have their contact information, I need to actually be able to send to them. So I'm getting inbox infrastructure to be able to send. And then I'm sending with a platform like Instantly. And then on the LinkedIn DM side, what I'm sending with is a platform, um, like Hayreach. Another one that we like is called BotDog. Um, both of these have APIs. Um, but what these enable you to do is basically, uh, do LinkedIn DM campaigns, um, from these accounts.

• 25:21Speaker 01

  I also know people that are just like using LinkedIn DM, or sorry, LinkedIn InMail for this and seeing incredible success right now, uh, using this strategy. So yeah, just throwing out all the strategies that are available. So this is how you can build this pipeline, right? Now, how do you actually like have an agent that is managing that inbox? So looking at Instantly as an example, they have an API. And that API allows for you to monitor and manage the entire account. So you can have an agent that's literally writing copy for each individual email or person that you're contacting or reaching out to, um, and writing those variables.

• 26:04Speaker 01

  And then that can be basically pushed into Instantly. So this happens outside the platform gets pushed in. But the bigger thing here is they also have webhooks. So when a positive reply happens, you can send that webhook confirmation back to your agent that's hosted on some type of cloud server. And that agent, you give it basically, um, like a base prompt, right? Of like your, the goal, like here's all the context that you need and your goal is to try to get people to schedule demos on, you know, this, this link, right? It can manage that inbox, answer questions, push people deeper. But the thing that gets really fascinating and really powerful with this, G, is like, it can do these follow-ups like months later, right? So it's like, okay, like also like every six months, right?

• 26:45Speaker 01

  I want to program that in to like re-reach out to these people that went cold. I can also plug it into my scheduling application like Calendly or like Cal.com. I can give the agent access to see, okay, did this person that we reached out to, can we, did they actually schedule a discovery call? Did they actually, you know, produce the action that we're, or, you know, make the action that we're trying to optimize for? And so from this, you can basically build this like SDR in a box, right? That is again, finding new people for you based off of the engagements that they're interacting with on social, finding the emails, actually writing the emails, deciding if this is a good ICP fit.

• 27:29Speaker 01

  And then sending that to the sending platforms and then managing the inboxes of those sending platforms. And again, when I say agent, right? Like when I'm saying, oh, it's managing this inbox, it's literally just code under the hood, right? It's code under the hood with an LLM attached. That is an agent, like in this context here. You don't have to overcomplicate this. You don't have to have God in the box managing an email inbox. Be a very simple setup to actually produce this. I also get asked this question a lot, like, do you need, use like some agent framework under hood? It's like, a lot of the times you don't need it. It's just bloat. You can just have a very simple, like, a very simple solution for these finite problems, right? It doesn't have to be this overcomplicated or over-engineered thing. So anyways, happy to answer any questions about that or dive deeper on any of this. Again, it's hard to show code.

• 28:13Speaker 01

  And so I didn't really do that today of like, this is how you do it. But what you need here, basically, the final piece is you need to set up a server. So use something like a railway or this is what we do at like Graft, right? It's like we have the data pipeline warehouse and then the server to deploy these agents to that's like off of the live data streams. But yeah, happy to answer questions, G.

• 28:31Speaker 00

  I mean, to be clear, you're, you know, you're using a harness like Claude Code or Codex to actually build out all of the thing. But this, the hard part is the strategy around, you know, who you're going after, why you're going after them, what's your tool stack? That like, what's amazing is you just like outlined, here's all the tools that you need to get like set up. Then it becomes, okay, I have to go into, you know, that's what people are talking about software factories. Like we're all in the software factory business now, right? Because we're just going and we're spitting up stuff like this, the software to actually go and complete these tasks.

• 29:11Speaker 01

  Absolutely. I think the thing that we are like focusing on, like, so to say, like, a good way to think about this is like, if you can build it in Claude Code and like have some type of local system that you're running, you can probably deploy that to a server somewhere, right? And have that run on an hourly cadence or a daily cadence or whatever that ends up looking like. The challenge ends up being, how do I set up the infrastructure that's necessary for the agent to be able to do this, right? And the, the solution is like the open source solution as an example, like we talked about this on the last call, use something like Airbyte, uh, with ClickHouse to get like create your data pipeline and your data warehouse. So you have that data stream for the agent to make those decisions. And then you have to have some server and like, when I say server, what do you, what

• 29:56Speaker 01

  is that, right? For the uninitiated, it's just a computer that is on all the time somewhere else that you're putting code on to, right? I think the software factory thing is super fascinating as well. Like, like really, it's funny, this is how I'm thinking about marketing now. Like marketing is just code. Like, like when I generate an open banana image, like that's just the JSON prompt under the hood. Like when I make like, you know, SeaDance AI avatar videos, that's just like an LLM that like scraped Reddit, like read some things, wrote a script. And then we, it's just an API call that's happening to Kai AI to generate that image with like, okay, here's how you chain this together to make it into 30 seconds.

• 30:38Speaker 01

  Every, everything now, like in, in my co-founder, this is his firm belief, like Max always says this, he's basically like, The only agent is a coding agent, actually. Everything else is just software that's being made by the coding agent. I think this is like this paradigm shift and like something that we are obsessed with. It's like, why are you paying tokens for things that can be just code? That is running on super cheap compute. You don't, you don't have to have like inference every time that you're doing this action. Only use inference when you need it. And this is kind of this like differing viewpoint that I think, you know, everybody's just like, oh, token abundance. I'm going to token max. I'm like, I'm actually totally like probably the opposite of that. Like, why? It just, it is wasteful. Like, do the thing that is the simpler thing that has less likelihood of breaking.

• 31:22Speaker 01

  Like, if you have Hermes try to run your Facebook ads, high likelihood, it might just like absolutely nuke the account. But if you have it run based off, you, you build a piece of custom software for yourself that's running based off of a system that a normal hu, like a real human run, totally different, you know, outcomes that you're going to get from that that are probably higher quality. So.

• 31:41Speaker 00

  Okay, do we have time for a second marketing agent demo flow?

• 31:48Speaker 01

  Yeah, I can talk through, um, I just did this.

• 31:54Speaker 01

  I just did this for my team. Um, I, I don't know if that'll be super interesting, actually. I mean, you tell me, we basically were like, okay, how do we at scale make social content on LinkedIn? For like the entire team and like, so we have them like, basically we're interviewing them. We take the transcripts, we pull out the insights. The insights get written into posts, the posts automatically get scheduled to their LinkedIn accounts using a tool called ordinal MCP.

• 32:20Speaker 00

  Yes, stop, like, yes, this is interesting because a lot of people, I mean, a lot of people might have heard, you know, listened to this cold, cold email approach or cold reach out approach and are like, hmm, I want to go the organic route. So like, what's, what's an example of setting up a marketing agent in our organic route? And can you break that down for us?

• 32:40Speaker 01

  Absolutely. Yeah, I'll do LinkedIn one because it's very super topical and like, we've had a lot of interest in this lately by companies, which has been pretty fascinating. They're using this with like their sales teams, like they want, you know, their seven person sales team to be posting daily. How do they actually do that and make unique ideas? So, um, this also pairs with the cold email. I'll talk about that as well. Um, but yeah, just to run through the process, uh, super simple. Um, it's like literally record a conversation like this, like I have a, a, a weekly call, like one-on-one with like the people that we're doing this for in the org. And I'm just like, tell me everything that like you've learned in the last week. I just basically interview them, have a conversation, right? It doesn't have to be anything like, you don't have to have any focus.

• 33:24Speaker 01

  It's just like, what are the things that, that jumped out at you after being in these sales calls or whatever your job is? You can do this for like technical people as well at the organization. You can do this for everybody. And I imagine this is how the large, like the real companies are doing this. There's no way that like everybody at like a lovable is running the content that's going out across all of the accounts. Maybe that's happening, but, um, I think what's more likely is that there's somebody behind the scenes that's orchestrating this. It also doesn't have to be an interview. It can just be sales calls or internal comps, like Alex Lieberman, as an example, has been talking about, about this a lot where they're, they're basically sourcing, like so much context is happening within their notion, within their code base, within their, their Slack. We see this as well, right?

• 34:04Speaker 01

  You can use one of these agents to query those data sources, right? Like query the sales channel, um, or query the Gong transcripts. And that's where you can pull this insights from. And honestly, a lot of the times you find that it's really inside, like it's really good content that's trapped in there. Like these ideas, like for example, a customer had, uh, you know, a customer said that, or a potential customer said this, and it was like why they didn't buy the product. And that can turn into an unbelievable piece of content, um, that you can extract from. So you get source material. Why do you have to get source material? The reason is because if you go and you try to just have the agent like, think about this, you're like, write good LinkedIn content.

• 34:46Speaker 01

  It's going to be the most mid thing you, I mean, it's, you're going to waste the person's time on the other side, right? Um, or you're going to get flagged for AI slop by LinkedIn's new feature that just released this morning. Um, the, the better way to do this is source this from real human conversation because that's where these original ideas are coming from. Um, the, another example of this is like literally this podcast. You could extract all the insights from the transcript and that can be used as social content. And this is like a strategy I use for myself, but it doesn't have to be just your own. It can be somebody else's as well. It can be, you know, a podcast with Naval. It can be whatever, it can, the source material can be anything. But the system that you create is some type of source material that's happening on, you know, some type of cadence. And then from that, I'm, I'm building basically this writing and scheduling process.

• 35:31Speaker 01

  So what I, I'll walk through now how to actually like do this. Um, so take that source material, you're going to do an API call, um, into, uh, you know, some LLM, uh, as an example, uh, for this, like you could, I mean, we've even used just like, uh, Claude Sonnet as an example, and it's probably good enough on the writing side. Um, and then once you have that, those written posts, you're then going to go and use scheduling tool. We like ordinal for this, um, they're a partner of ours as well. Um, but it allows for you to have multiple LinkedIn accounts connected to it. And then they can also interact with each other, which is amazing. Um, but you can through their API or their MCP, schedule these posts to each of the individual accounts.

• 36:17Speaker 01

  And then ordinal also has, and I could just go into this to actually show you. Um, ordinal also has, uh, the, uh, analytics data that pulls in from your LinkedIn posts there as well. So we can see the breakdown of like which content is actually performing well. So it has the analytics of the multiple accounts. You can actually see the breakdown of the individual posts. And that data stream can go back to the agent so that it understands, okay, this is what's getting impressions. This is what's doing well. Let's go do more content, like when it does its cycles of writing, that can influence the next round of creative. So topics like this perform better based off of the source material we pulled. How can we snowball or remix? Use those specific words, snowball or remix to have it go further, right?

• 37:02Speaker 01

  And this is where the LLM is thinking on top of that data stream. And when you look at like what is happening here, like what does a social media manager do? I actually think the social media manager job, like full stop. It's, I think it's already dead, but that's, I won't get into that. If you're listening to this, please learn how to make and manage content at scale across multiple accounts, um, it's with agents. Cause that's going to be, I think that's the real meta now is like, how can a single person manage, you know, 10, 20, a hundred accounts across all of these different channels. Um, but when you look at what a social media manager did previously, like a good one, that was actually excellent, excellent at their job, is they would prospect for ideas, they would make content about those ideas, they would publish it.

• 37:45Speaker 01

  They would look at the data to see which got the most impressions. And then they would turn that into a recurring content calendar where they're like, okay, I'm just remixing this, these same ideas over and over again. If you look at my Twitter, like, uh, posts, so that's an example, or even my LinkedIn, it is the exact same thing, remixed every 90 days. Like full stop, that is all that's happening. And that, when you get enough information, like a bigger enough corpus, you have, you basically understand what's already going to go viral. Like I, I have these posts that I've literally used for the last two years. Every time I post it, I know it's going to go viral. I can't post it every day, you post it every 90 days, right? And that's how you can go back into this cadence. And so again, have this mentality of I'm prospecting for ideas, I'm prospecting for winners.

• 38:28Speaker 01

  Once I find those, I'm trying to use those as, as often as I can, because I know that that's what's going to work. That is what the audience is resonating with. And this is, this applies to product as well, right? Like when, I, I think that a lot of first time founders, they, they spend time thinking about like, I'm trying to get the market to buy this. And in reality, it's like, I'm trying, the, the, the, the pros of this is like, what does the market want to buy? Can I build it and can I sell it to them, right? Like that is actually how you start a business. And it, it, for some reason, it's this, this flip thing where they're like, oh, I'm trying to invent a new idea. I don't want to invent a new idea at all. I want to be like, what do people want to buy that currently like they can't buy?

• 39:10Speaker 01

  And can I go and figure out the, the way to build that thing? And then I know I can sell that back to them. I know the market is going to be receptive to, and you need to think about content in the same way. We're like, what is the content that the market is currently receptive to? And by mining that content from other sources that has already had a viral moment, this is a way to leapfrog that to identify that. And then you're going and you're putting your own spin, you're putting your own, you know, angle on this. So anyway.

• 39:34Speaker 00

  Lot of thoughts there. Um, agreed on the social media manager is like, that role is dead or it's evolved. It's going to evolve, like it's going to evolve into the social media agent man, manager. So you're going to need to be able to spin up agents so that you can create a bunch of accounts on the fly that systematically creates content like you have, like you get millions of impressions a month, free impressions. Actually, the platforms are paying you, which is insane to do it. It's insane.

• 40:07Speaker 01

  I get paid to build lead pipeline, like think about that.

• 40:10Speaker 00

  It's crazy.

• 40:11Speaker 01

  And like, I, I, it's so funny, man. I'll talk to like founders or like, you know, large, like people that, that run bigger companies and they'll, they'll be like, why are you, why would you, would you invest in social? And I'm like, look at the earned media. Like if you were paying for those impressions on platform, for example, on LinkedIn, it's like $22 per thousand impressions is the average, right? It's like every post that you get, even with an account that's like 500 followers, you can get a thousand impressions. That's like $20 that you just like put into your pocket for free, right? But, but it's, so there's the earned media side. And then there's also like the platforms pay you, like YouTube literally pays you to do marketing for like checkout. Like what the, what the hell?

• 40:54Speaker 00

  It's crazy. And then, you know, for the people who are like, well, I don't want to do a personal brand, makes sense. What Cody is suggesting is like have people on your team have these personal brands. And if you don't, and by the way, I'll give you a piece of sauce. If you don't want to do that, another really, uh, smart thing to do with agents creating content for you is creating theme based pages or topic based pages. So for example, my good friend, uh, Julian Shapiro, you know, he had a company, a growth agency called Demand Curve.

• 41:25Speaker 01

  Absolutely. By the way, his blog is incredible and that's what I came up on. So I'm just like.

• 41:30Speaker 00

  Yeah, I actually grew up with Julian.

• 41:33Speaker 01

  No, did you really? That's amazing.

• 41:34Speaker 00

  Yeah, he was like my neighbor.

• 41:35Speaker 01

  He owns like a farm now or something, right?

• 41:36Speaker 00

  Yeah, yeah. So I need to get him on the pod, but, uh, Julian, being the smart guy he is, it's not like he created a, uh, X account that was slash Demand Curve. I mean, maybe he has that, but he actually created an X account called at Growth Tactics. So he's creating content on this growth tactic page. People interested in growth tactics follow it and then they learn about his agency and his products through that, right? Exactly.

• 42:08Speaker 01

  Media company. And like, again, it doesn't, it could be, I mean, there's the ones that are my favorite are like Chase Passive Income. I don't know if you've seen this.

• 42:14Speaker 00

  Yeah.

• 42:15Speaker 01

  Um, they're doing it more as a meme page, but like you can use like this attention that you can garner for free as a way to drive inbound for whatever, whatever it is that you're building. It doesn't have to just be you. It can be this like anonymous thing that is still providing value that you're aggregating and, you know, organizing for the internet, right? So I'll leave it there. I don't know if there's any other questions.

• 42:34Speaker 00

  Really, uh, you know, impactful marketing agents that you just broke down. Um, I wish we had 40 hours together and we did like a crazy.

• 42:47Speaker 01

  Comment below. That's the only way I come back. That's the only way he'll have me. All right. So you have to do this. You have to comment what you want to learn. I'll tell you, I'll, I'll teach you whatever you want. It can be how to build social media agents, like for TikTok clouds. It can be like, how do I actually run and pay to ads account? It can be anything that you can imagine. It can be direct mail. I'll literally walk you through, how can you send direct mail at scale by scraping Google Maps? You name it. How do you advertise on TV and what's the meta there? Uh, like, how do you get cheaper clicks on LinkedIn? I, I can break down any of that, so.

• 43:22Speaker 00

  I appreciate you, Cody. We'll see you in the comment section. Like always, I'll include links for where to follow Cody on the internet, in the show notes, in the description.

• 43:30Speaker 01

  Can I shout it out? She, give me, give me the opportunity.

• 43:34Speaker 00

  Go for it.

• 43:35Speaker 01

  Hell yeah. Go find me on Twitter, LinkedIn. That's where I'm the most active. And if you want to deploy these exact agents that I talked about today, go to graphed.com. Uh, we have both the platform solution for this and also we forward deploy software engineers to do these actual implementations on our platform. We would love to help you. If you're a fast growing company, that is who we're seeing the most success with. So thanks for having me, G.

• 43:56Speaker 00

  God bless you, Cody. I'll see you next time.