Name: How To Build AI Agents for Your Event Workflows
Uploaded: 2026-09-29T18:16:31.069Z
Duration: 45 min 33 s
Description: How To Build AI Agents for Your Event Workflows

Transcript for "How To Build AI Agents for Your Event Workflows": Hello, everyone. Welcome to September's Insider. So if you all are looking for how to build AI agents for your event workflows, you are in the right place. My name is Cindy Dubon. I'm the director of demand gen for Goldcast here at Cvent, and I will be helping hosting, my guest speaker joining me very soon. I'll let him do a brief intro, but thank you. Shout out to the MarketingOps ops community. Jomar Marr, our speaker, is very active in that community, so you probably know him from there. But really appreciate our marketing ops community showing up and engaging with our content. We appreciate you all. that, today, we'll be walking through the state of AI agents for marketers. Jomar is gonna lay down, what's going on for us right now, get us up to speed. And then we're gonna learn how to build an event follow-up email agent as well as a ready to to act leadless workflow. Don't worry if you are a beginner, advanced, intermediate. We are addressing how to build those through all of those stages, and then we'll do a Q&A and A at the end. We'll try to reserve fifteen minutes for that. But please do ask as we go. You can see Jomar's already in the chat. So he and I will both be working the chat. I will bring those up. You know, I will read those out loud as he goes throughout the presentation. Without further ado, though, Jomar, please join me on stage. Hi. I feel I feel actually better just responding as a GIF on the chat. I'm having so much fun with the GIFs. Yes. My personal favorite as well. Just make sure we pick the ones that tend to be safe for work because I think we can sometimes accidentally send the wrong one. Before I forget, I'm gonna open the first poll just to level set here before Jomar really jumps in. We do want to understand where you all are at in terms of your journey, so go ahead and navigate to the polls tab and go ahead and answer that. We'll keep that open for a while. But, Jomar, why don't you give us a quick spiel on who you are? I obviously know you from so many past lives. I think we met so long ago at the height of ABM, and here we are at the age of AI, Agentic MCP, whatever you wanna call it. Give us a little overview about you, where you're joining from, what you're doing, and what we'll go through today. Yes. Can you see my slides okay, by the way? I can. It looks great. Okay. Great. That's a picture of me. I don't age, so I look the same. That's a picture of me five years ago. so I've been advising clients in complex and regulated sectors, including legal, insurance, finance, high-tech. I drove digital transformation in a digital transformation company. I've architected AI GTM solutions in different companies. And right now, I'm an author, speaker, and builder of AI. And so today, my session is how to build AI agents for events. And I also published a book at the beginning of the year, and it's called Dare to Orchestrate, Taking Control of AI Agents Across the Customer Life Cycle. And it wouldn't be a AI agent book without a random AI agent lady holding my book next to an airplane. it. okay. So here's the timeline, where we current a short history of where we are and how we got there. I'm just going to go ahead and give you guys a little spiel of, you know, step one to where where we are right now, which is the goal of, you know, what the current AI agent architecture looks like. So this first step was the frontier models. It arrived. Right? You got your GPTs, your cloud, your Gemini. Everyone was excited. It it opened a floodgates to a lot of enterprise companies. They were like, just give everyone a GPT. You get a GPT. You get a it was giving Oprah look underneath your desk moments. And then there was the rapper era, which is all of the SaaS companies started to panic. They're like, how do we integrate AI into our platform? It was it was really giving panic, but also, like, how do we make sure our investors, our stakeholders are happy that we're integrating AI? That was step two. Step three is called the DIY agent billing, which is when NAN, Zapier, Make, all of these platforms that were, like, orchestrator platforms, even Recado. The in, they were all looking at, like, okay. We can actually we're we're the most popular girl in town in terms of, like, we're the most popular people in this cafeteria. Everyone come sit with us. We can integrate AI. That's literally step three. Step four right now is currently what we're also experiencing is this the Frontier Labs are moving up the stack. They're starting to put their little fingers into, you know, the memory, tool usage. That's literally what's happening in step four. And then step, you know, the goal now is agent architecture that's more agnostic and that you can control it. Because now, the Frontier Labs are trying to lock you in. They're trying to lock you in a relationship so you they can control the all your memory. They can control all of your AI ledgers for your emails. So my goal today is to show you all the different stages that, all the different stages of how to get there. And so today, I'm gonna focus on really two different types of agent. One is an event follow-up agent, and one's, like, ready to lead agent for list building, which is like battle cards. And I know a lot of people are take a lot of time building those battle cards. And so I'm gonna go over those two different types of agents. And so the GTM AI maturity model, like I said, with that time line, the phase one right now, which I see a lot of people are in. And there's no like, I'm not telling you, hey. Move all the way to phase three. Like, take your time, experience it. You really have to socialize this with your team, your department, as well as the IT team, the security team, the data scientist team if you have some. So it's really you know, part of it is the socialization part. But if you're in phase one, no worries. We're gonna go cover all the three different phases. Phase one is more of the native in app automation, which is like the automations inside HubSpot, inside Marketo, where they're pulling up APIs. They're not really pulling up MCPs, but they're pulling up APIs. And phase two is more the composite mo modular agent, and that's more like, hey. I'm inside NAN. I'm inside trigger dot dev. I'm inside Zapier, and I'm working outside of, you know, the native in app applications. And then phase three is the governed agentic operating system, and that's more of like a headless model of all your different types of agentic tools, memory, you know, data plot data hosting for, your data lakes, and that where we'll also cover as phase three. So the first one is the native automation, which is low friction, low context. Everything, like I said, is the architectures inside, you know, either a HubSpot, a Salesforce, a Marketo, any form where you can pull in different types of workflows inside your native system. And so the data flow is gonna, you know, still pull from your CRM. This execution is still gonna be effing done. You're still gonna have trade off. What a lot of people are doing is also plugging in a bunch of the knowledge bases for the different types of, marketing automation that they're currently in or CRM, and then letting, you know, an agentic platform basically help build their different workflows using AI. That is a very popular one that I'm seeing across the board. And so for the webinar follow-up email for this particular phase, phase one, the event data is still gonna be pulled in from Goldcast. And you can either do an API, and I know they're working on MCP, but they're currently doing an API from Goldcast into, you know, your CRM, your HubSpot, your Salesforce. And then you can draft it using another API that's pulling from any Frontier model, any secondary model, any self hosted model. I don't care. It's a model. It's there. And then you still have the review gateway. You're still verifying, hopefully, verifying all of the emails before you send it out to either your sales engagement platform or your Marketo or HubSpot platform. Here's the architect, here's the stack, and here's the limitations that you might have. I'm going a little fast, but if you guys have questions, I'm sure just put it on the chat. And there's still the human in the loop part, which is reviewing all the emails. The account intelligence and the dynamic battle card for each of the individuals going into your events that you're sending to your business development rep account rep. Right? It's going through a batch sync, and it's gonna still process to the CRM. It's gonna get a static alert, alert generation either through Slack or through an email to your team, and you're still they're still gonna be a review, which is more internal. So this is like an internal battle card that you send to your team, and it's gonna create one for each of the new members that are actually sellable for your ICP. And then there's gonna be a manual thing for, you know, the inside sales team to review, make sure it's correct before, you know, they start speaking to the actual prospect or customer that you're trying to upsell or cross sell. Here's the architecture for that. And this, you know, very simple, still native to Slack integrations, your stack, your capabilities, and values. It's all here. You guys will all get the slides later on. I just wanna make sure we have enough time to walk through the process. And this is the event driven connected workflows, which is high agility and high maintenance. So the this is the phase two of the, one to three phase, and this is more going into the actual workflow. So this is gonna go either into NAN, Zapier, your cloud routines, your OpenAI task, your Gemini triggers, anything that's going to trigger the workflow. And you could use any type of execution LLM. So I'm saying whatever fits I mean, a lot of people use the highest frontier model. I people are starting to realize sometimes the best frontier model isn't always the greatest or cost effective way to run these workflows. So the first one is the webinar follow-up agent. So once I, I walk you through this one, I'll then show you what it looks like in the actual platform. And so an event signal triggers, and so usually, this will come either through an API. For the course for this particular session, I just put it as a, like, Google Sheets. And if the middleware is gonna be n a n, it's going to get enriched using, you know I currently use some people use Clay. I use, Bitscale right now. Whatever is you guys are having to enrich. If you guys have an enriched waterfall, that is something you guys probably do before you even, you know, get the lead, back into this type of webinar email follow-up agents. And then you will review it either through a sales, loft outreach beforehand, either through a Google Sheet of all the emails that are coming up. I know Clay has a direct Google Sheet or not direct way where you can see a column and has all the sheet before you send it. So what does that look like from an actual setup is once you get into NAN or any other type of platform, most of these companies are going to have an MCP. So I just for for this session, I basically just signed up for, like, one of these NAN agents a lot of people use. I very much went into more of a developer, session. So for NAN, they have all of these enable assistant. A lot of people use those enable assistance because now, back then, like, I don't know, like, since last December or October, a lot of people were having spaghetti fights with n a n because it looked like literally SpaghettiOs. But you're gonna have this, connect to client. You're gonna have this MCP server, connection. So once you connect that, that's going to be your way to basically, you know, prompt Claude or prompt whatever LLM model you do have to make sure that it creates your model. So once you create that, you're going to start giving all of, that particular agent types of permissions to different platforms. So once you have this connected, you go here, you go to MCP connect, then you just go on to your either Visual Studio or Quad Code. And I already predid it, but we will give you the workflows after the session. And the first one was a webinar follow-up. So here's the webinar follow-up. It's gonna run that batch. You click it. It's gonna normalize the attendee. It's gonna look up pull up the LLM. It's gonna do the follow-up draft, and it's gonna put it right into your Google Sheet. And sometimes people do that Google Sheet either in, you know, Clay or whatever sales engagement platform that you're sending. And so for me, this was, like, the easiest way to do it, especially for the session. They'll have the follow-up body here, sorry, Drift. And you will be able to add your personalized oh, sorry. Where am I? Your personalized, brand guidelines, etcetera, into this Claude LLM model. And so I don't know what your brand guidelines is, but I didn't include any for this session just so you can include them here. And then once that triggers, you're going it's going to go straight into Slack saying, oh, this person went and did a follow-up, for this webinar. Here is what it looks like, and then you could it'll show you where it is. It'll show you where the sheet is, and you click it, and this will then be the sheet. So a lot of people are using Slack as a way to communicate with your agents and as a way to get feedback from, like, the different types of, you know, orchestration agents that you're either using, either Clay, n a n, or Zapier. So and then the second one we're going to talk about where is my stack? Is the account intelligence and dynamic battle cards. So before, a lot of people take a long time to create battle cards, especially for, people that are finishing up a webinar. And so there's gonna have a high event webhook. It's going to go to NAN. It's gonna get enriched, right, through either bit scale or whatever enrichment platform you do have. And then it will, you know, put that into the right deck, and then it will also put it into GitHub and as hosting for the HTML. And then you will have a review gate where you get to look. Okay. That looks good. I'm gonna go ahead and send that to the inside sales team or business development team. So this is the architecture I use for it, you know, NAN for for this demo. You could use whatever NAN or whatever third party platform. It has all the capabilities. And this is what it would look like in NAN so far. Okay. And this is what it would look like in a n a n. The event trigger would come up. It'd look up in your CRM. It would look up a battle card slug for that. Like, okay. What does it look is what is this supposed to be formatted as? It will then get enriched by bit scale. The the battle card synthesis will be then put it and organize it together. And then it will then I think this one doesn't yes. And then it will basically go through the battle card blocks, put it into the, HTML, and then it will send it to a Slack. So what does the Slack one look like? It looks like this one. And then you get to basically just view the battle card, and it will look like this. And it was from a live webinar signal. Here's account summary. Here's the competitor's flag, objection. And so that will automatically get sent to your team. And you could, you know, customize this to basically have your brand guidelines have more of your knowledge database. This one, I don't know anyone's knowledge database. It's just a sample of what a battle card would look like. But then it would get sent to the Slack bot, and then you could either push that, you know, to your inside sales team or business development team, and that will give them the right information from the people that attended the webinar. And so that is the battle card one, and I will send you an updated version of that in n a n after the session. And then the third the third one is governed tool calling orchestration, and this is the one that's a little bit more this is more advanced, and this is where I'm very much leaning into, which is like a headless model. I'm not using any Frontier models. I'm not even using Zapier or, Make. It's very much so, like, I'm using more developer friendly tools. And that's kind of what is happening across the board is that we are we as technical marketers or technical GTM people are very much leaning towards developer AI tools because it the barrier to entry is very, very small. Now, like, we don't have to you know, AI can basically guide us in how to actually use this correctly. And so the the product build of my AI GTM is that I don't own a particular model. I don't care about the model. I don't care if I'm using claw the highest the newest claw, the newest open AI. I it doesn't matter to me. It's very I could switch it. I'm playing very Switzerland when it comes to LLM models. There's no vendor locked in. I'm not using any of the Frontier models' agent harness. I'm actually using my own agent harness, which is called a deep agent, harness, which is, from Langraph. And then every run makes it smarter. So for me, every time I'm what I'm really collecting is that AI ledger for each of the workflows. So every I'm basically gathering all of my employees' attention and how they're interacting with the agents and saving that into my AI ledger. So what I get is, like, governance from day one. Like, I'm showing them exactly how to use it. I'm basically gathering all the information from my Slack bots, from anti gravity or whatever platform that they're interacting with, and then making sure I'm collecting that AI ledger. Because for me, that is the sweet spot of, like that is the new intellectual property. That's how I see it. I'm not giving that away to any Frontier models. There is a reason why Claude gave out three small business agents for small businesses because those small businesses, their workflows are and the way they're interacting with it is actually the intellectual property that has more value. Because in this world, especially AI world, anyone can code. Anyone can create a AI workflow. I made that AI workflow literally in five to ten minutes. That's how easy it is to make AI agents these days. The ones that are going to withstand are the ones that are not giving away all their information to Frontier models. There's no rip and replace, so most of this all everything I built with this headless g t like, AMR GTM AI architecture is gonna sit on top. So these are this is how I would have built this with the email follow-up agent and to ready to lead agent, which is the battle card, is that I would have a business intelligence context. So this business intelligence context and operating environment all live within this this is the new IP, intellectual property for companies. And any third party system like Salesforce, Bitscale, Goldcast, those are third party system that we can still use. But these are the new intellectual property. Like, the semantic like, the skills that you create. Right? Every AI workflow, every output you have will increase the skill of your AI workflow. That's semantics and definitions. There are specific definitions that you have in your department, in your team that cannot be copied into other companies. Right? And so that will live here. Your policies and exceptions. So that also from your department. If your business team has dip this development team has different SLAs than your marketing team, and those policies and exceptions will live there. Your commercial contacts and evals and feedback, these are all business intelligent context that is all proprietary to your company. And you guys should protect it as much as possible, especially the evals and feedback. Because when another company tries to acquire your company, they're like, oh, we can make all these AI workflows. But but we're like, oh, but we have all the emails and feedback for each of the AI workflow. Our confidence rate is this much higher than your confidence rate because we actually have the ledger that shows all of the iterations and corrections that we made to make I our our AI workflow a lot smarter than yours. And so our, operating environment is AI agent orchestration, tools and connectors. These are all going to, you know, these are all operating. These are the arms and legs of your agent. And so the webinar follow-up is you can't really see it in terms of, like, a workflow. There are certain types. You could simulate what it would look like as a workflow, but the webinar follow-up will basically pull from an API from an event stream, still very similar. I use a LangGraph orchestrator, and then that will do a runtime OS. And then the bandwidth here for an MCP, it will go through a dual review gate, and then it'll get sent automatically. So I'm still using Slack as a centralized form to communicate with everyone. I'm still using, you know, a different type of Google Sheet. It could either be a snowflake or it could be, you know, a supabase. It really you know, at the end of the day, a database is a database. And I'm also doing that exact same thing for account intelligence and battle cards, taking out the signal, putting into an MCP, reviewing it in a Slack, and sending it into, putting it in, like, a GitHub repo that has HTML feature, and then using Slack as a way to communicate with the data back and forth. And so thank you so much. It's was really fast that I went through, but, hopefully, I was able to cover all the different phases of how to make AI agents. And there's levels to this just like everything else in life. it. Awesome. Thanks so much, Jomar. It's I see a few questions. Before I get into it, I am going to actually close the current poll, and I'm gonna open the third the second poll. And it's for those of you who actually want a deep dive into Jomar's phase three. Just TLDR, when he and I were prepping this, I had asked him to hold back on phase three, because I wasn't quite sure how the audience, how savvy the audience was. But if you are interested in phase three, we can kinda work set that up for y'all, but, would like to know if anyone would be interested in that. Jomar, before I get into to Q&A and a, specific q and a's, I'm curious, where you think, like, most marketing teams most marketing teams, not the leading marketing teams. Where do you think most marketing teams should be six months from now given the landscape of AI agents? Like, the ones that are gonna be left behind are probably doing this, and then the ones that are job secure are probably doing this. How would you answer that? Well, I feel like right now, there is a lot of you know, there's still stragglers from Clay. And I like to think that they're stragglers because, like, Clay did make the GTM engineers kind of, like, popular across the board. They're but so far, they're like, oh, I can do GTM engineering, or I can AI your entire customer life cycle. But at the end of the day, when you hire them, they're just setting a play for you. And I feel like now we're moving a little past that. So when you hit that six mark, I think people are starting to look at this from a headless perspective. Right? What is headless mean? It means being agnostic. It means not relying on a Frontier model. It means, you know, owning your own emails. It means owning your own proprietary information. It means having a semantic layer with each of your AI workflows. So for me, like, that's how I see six months from now. Everything is moving so so fast, and people can't keep up. I, for some reason, can keep up with this kind of stuff because I've been trained through telenovelas since growing up how to keep up with various degrees of scenes and scenarios that, like, I can keep up with the flow and be like, hey. I know how Sebastian got here, you guys, and it's not through his stepmother. So for me, like, that's how I can able I'm able to follow this drama and also follow the amount of intricacies that is happening from a technical perspective. So, yes, I mean, I think the the six months from now is gonna be agnostic, and it's going to be, like, every there's Frontier models are gonna be, like, a dime a dozen. Anyone can make a Frontier model these days. I'm not even joking. Like, there is every like, it's not anything revolutionary anymore. So, yeah, I think everything's gonna be headless. Everything is gonna be agnostic. I'm already setting this up. But the great thing is, you guys, let me tell you the tea, is that all these AI companies that are supposedly AI, their product is great, but their internal team running AI is not that great, hon. They're trying to, like, operationalize this inside their internal system, and they're having a hard time because their product is great, but their internal architecture isn't that great inside yet. So that's where we all come in and be like, hey. We actually know how to do this now. Let's pick up. But the thing is everything is moving faster, so the people a lot of there's a lot of AI there's a lot of noise. Right? There's gonna be a lot of AI agent influencers or AI influencers telling you the next greatest thing, like, really feeding off your FOMO. And they re their whole bread and butter is literally just attention. Like, no architecture built, nothing. It's just attention. So there's gonna be a lot of noise out there. And by the time you finally waste, like, five to ten hours of your life trying to figure out what they were showing, it really turns out to be nothing. And you're like, well, that was nothing, and you're just, like, pissed. So, yes, it's long winded answer. I don't know if you wanna actually share, like, specific logos, but are there it sounds like there's a lot of noise. Who is catching your attention but in a good way? Anyone out there we should all follow, Who's pay attention to? I mean, there isn't really I feel like everyone kinda stabilized, and everyone for me is kind of, like, in this, like, plateau where they just don't know where to go. Like, they're like, how do we even build this? They're hiring, like and everyone's kinda failing and then rehiring, and then it's just like, I don't think that there is a particular individual that's like, hey. I'm putting my neck out there. Because the people that are really doing well from my perspective is that they're building so much. They have no time to go on social media. They have no time. They're like insiders. They're like, I got my bread, hon. Like, I got my bread. Like, I'm collecting that bread. I'm don't have to talk to anyone. It's, like, very much so in and then the people that are kind of a little bit, they try to over hyperbole what they're doing to get that attention. Right? Attention currency is very much so alive and well, especially with the AI agent hype and People are scared to get left behind at the at this moment in Sage. But for me, I'm looking for how do I build an architecture where I'm not reliant on a frontier model. I'm not reliant on anything else. I'm a Switzerland. I'm building Switzerland infrastructure because I don't really care how everyone else is building it. As long as the companies that I'm working for and the company that I'm working for is good, and I'm making sure I'm able to swap, I'm making sure my confidence rate for each AI workflow is really good, my evals are good, my runs are good. I'm increasing EBITDA workflows. And if you guys are interviewing and need to know, look up EBITDA workflows because that's really gonna catch the attention of recruiters. It's gonna, make sure you drop that EBITDAO AI workflow high, and making sure you're, like, building at the correct stages is what I realized. A lot of people are not building AI workflows for the stage of their company that they're currently at, and they're building something that no one even wants. And make sure when you're building AI workflows, it's through phases. Like phase one, get it to work. Phase two, get more people on working it. Phase three, then let's add the let's add a little bit more, like, you know, to your AI workflow. Don't go straight to three, and then you're wasted three months of your freaking time at the company, and no one is using your AI workflow every and it's just like a dream. Like, that is not gonna increase the hype of your AI workflow. So my thing is make sure you build in phases and make sure you're teaming up and you're socializing all your workflows. Just so no one feels alone here, I just wanna share that how much our first poll question, how much AI are you using in your event workflow? The bulk of us are in early stage use or moderate use of AI with very few advanced or not using AI at all. So, we're definitely all in the same season, I would say, of this journey, and I would definitely double down on what is saying, which is that don't feel like you're behind and start building for practicality, versus for catching up. Alright. I'm gonna jump into some of the Q and A because I see a few that are upvoted quite a bit. Jomar. Do you have resources for how to learn this lingo around building the agents? As a marketer, how do I get up to speed so I can build the stage three? Because I think a lot of us are in, like, one and two of the journey. So what. resources do you recommend? What things are you besides reading your book, obviously, and joining the marketing ops community, how can we best get to stage three? Well, to be honest, I got to stage three because I was working with sounds crazy. I was working on a bunch of projects, and I interviewed, like, 50 different types of AI, GTM, or AI engineers. Not even GTM engineers because my knowledge is customer life cycles in GTM. Like, I got AI to like, I got customer life cycles, like, wrapped up. Like, I know how this works. I know all the different, like, teams that go into the customer life cycles, and that's my strongest point. But I'm also extremely technical. Like, I'm really good at figuring things out and taking things from different expertise. So when I interviewed the 50 AI engineers, I did to, like, hire in different type of enterprise companies. I interviewed them. Like, how do I scale where because a lot of the companies that I was working with, they were basically given the floodgates for, OpenAI GPTs. Right? And that's gonna expire by the end of December. And everyone was building it. Like, everyone in all these different enterprise companies who built in GPTs for their company. They didn't know how to use it but they were building it for random reasons. And so a part of me was, like, how do we make sure we transfer these GPTs into agents that is not gonna get wiped out by a sunset of a frontier model? And that's kind of how I went about it. My strategy isn't, like, you know, golden it's not like best practice. I think everything is different for each company's best practice. This is my interpretation of an agnostic Switzerland because I was tired of switching different models, like, different models for different companies. Like, now I'm fully agnostic of all LLM models. I'm making sure all the proprietary data is in our company's, Snowflake or Databricks, and I'm making sure that, like, it's built to endure whatever fight this Game of Thrones frontier model is happening currently in the in the ecosystem. Like, I'm gonna just sit down and watch everyone fight it out, and whoever comes out victorious, that's cheapest, that's the best, that's the model I will have my AI workflows, and that's the model that I architected. Right now, I don't see that many people doing a deep agent harness, for Langchain. What I and using neurosymbolic AI pattern to basically increase, you know, the confidence rate and decrease hallucinations. What I do see a lot of in the industry, which is weird, is they're using the Hermes agent and then using that harness to then sell to small or medium sized companies. It's great for solopreneurs, but it's not great to scale for small, medium, or enterprise companies. And so I've seen that across the board, and a lot of people are, like, pushing that. These AI influencers are pushing that, which is, like, totally wrong. And I hope that, like and I've only said that because I've had to fix, like, six of those. And, it's really concerning, like, because a lot of people are wasting their time and money on different things that people are pushing that is not really fit for the architecture. But I don't blame them because the people hiring don't even know what to do. They're it's very trust bro mentality right now. And the AI agent infrastructure world is trust me, bro. I got you. I'll build it for you and see if it works. That's kind of how it is right now. The trust me, bro age. That's cracking me up. How it is what it is. Yeah. What I kinda like about what you're saying too is that as you're building, like, you can swap. Like, when when it's another model or some other tech or whatever you wanna plug in, it's like you can easily swap it because you've built it so agnostically. Mhmm. So it's gonna be hard for a lot of us to break out of that. for risk too. Like, how many times have you seen that cloud go down? A few times? Have you how many times have you seen OpenAI go down? Do you really wanna stall your business because an and Frontier model is is down like that? And it's like, do you really wanna use their harness? Because they can increase price any single day. You don't want that. You know what. I mean? You want to make sure your business is predictable. You wanna make sure everything you all your skills are also agnostic. Right? Your skills are agnostic. Your LLM is agnostic. Like, everything you're building should be agnostic going forward. Because then when you have headless, that head can fit in every body, hon. That out that shirt can fit in any person's body type, basically. I'm gonna bring up the next question because I I also had it as well. But with the ability to build to really build whatever you need, how do you decide what tech to buy versus what to build as a marketing team? Oh, for me? Okay. I I mean, going into new organizations, I see this all the time. I I don't like to fight. I'm a Libra. So, like, for me, I like, I'm a Libra. I don't like to fight. I don't like to stir the pot. I'm not a pot stirrer. I'm very much so, like, what tools do you currently have, and how can I use those tools? Because at the end of the day for me, as long as it could do the damn thing, I'm looking at like, the main architecture. I can spin up any type of data data from, like, you know, Snowflake, a Databricks. Sure. If you don't have anything, use a Supabase. Like, self host that shit. Like, everything is self hostable these days that your cost can be, like, literally zero if you wanted to. Someone has to maintain that for the IT team. So right now, a lot of IT teams are making self hostable environments for people so they can put a lot of their products as self hosted so they're not using a lot of their money. Right? For me, that's kind of, like, where I've been pushing people is that it's, hey. If this platform doesn't work, let's just self host it. And that, also, if you're a high regulated company or if they have a lot of IT infrastructure or IT help, it really, really makes people happy, like the IT security. I'm like, self host? Okay. Like, we love that. Like, our data comes safe in our server. Like, they love it. And I'm just like, I'm here for it, babes. I'm here for it. I'm like, alright. That's what they love to do. Let's just help host it. But any type of technology, I'm very open to. I'm you know, if it's n a n, if it's, like, trigger dot dev, if it's whatever they have, I specifically ask an IT person, like, what are the list of tools that you have? Get let me see that. Because you wanna pick and choose your battle. You don't want to go in there being confrontational when you first go there. You want to make sure you have a lot of small wins. So when you do add that bat when you do ask for that big thing, you're like, but I've already built all this. Mhmm. You know what I mean? You so you wanna make sure it's still a political game no matter where you go or wherever you land, but make sure you pick all those tools and, hey. Can it trigger from this? Can it access to this? That's why I I like Slack so much, and I think it's one of the best acquisition Salesforce has ever done because it really this a AI agent world is so reliant on communication with agents and communication with humans. Right? And so as long as you have, like, a way to trigger something, it could be Claude managed agents. There's so many everyone's going into this orchestration thing. Look at Clay. They're going into the orchestration thing too because they realize, hey. Like, you know, everyone's copying our product. What's different? Orchestration. And and everyone's fighting for the harness. Right now, everyone's fighting for the harness. But what I'm saying is build your own damn harness. Alright. I think two more questions. So first one up is, what are some of the most impactful event workflows that you've seen people implement? I honestly think the best one is creating, a battle card straight away. Like, making sure you have this battle card knowledge database. You have the because that usually takes so much. You have the speaking points of your product, how you're gonna speak to them. Because then once you're you get all those people that's like, watch your webinar, you could basically see, like, okay. Here is the webinar. Here are the ICPs, the accounts that I'm targeting. Here are the individuals. Here are the speaking points. Because, usually, that is a high EBITDA AI workflow for lead to speed is. And it's not enrichment because everyone can do enrichment now with, like, that. It's basically creating this battle card that you can send to your AI sales team, understand what these the speaking points, making sure everyone's in the right brand, right messaging. They know, you know, where they're currently at with the comp account. Like, are they current customers? How how often have they been interacting with us? What is their current stakeholder? Who isn't who is their boss? Like, what events are they coming up to next? Like, how should I you know what I mean? What do they even like? Some people even scrape, like, Instagram or Facebook pages to see, like, okay. This person loves golf. Let me invite them to a golf event. So it's, like, really socialized engineering the battle cards and making sure, like, you know your account and you know how to brand message them consistently throughout the customer life cycle that they're currently in. you. Last question here, and this will be a fun one for you to bow tie, Jomar. Explain a deep agent harness like I'm a three year old. Don't understand where I store an AI ledger or how that connects to LLMs or if it even does connect still. So let's let's bring it back to basics here. Explain a deep agent harness to a three year old. a deep agent harness to a three year old, I would say it's basically you making sure that I'm trying to I have a three year old niece or a four year old niece, so I'm like, I'm gonna just explain it to her because pretend like I'm speaking to her. So I would say a deep agent harness explaining it to a three or four year old, would be like, hey. When you build something, make sure you are building it so all the other items get to share this resource. So when you are building, you know, blocks, you are taking pieces that can then help build the other pieces of the agent or the block. So So when you're building a castle, make sure on using these blocks, make sure these three these three little pieces can also work with the other pieces so the square happens and you can build a better foundation for your castle. So it's like a shared resource for your blocks. And yeah. So make sure you get line all these different pieces and that they're all can work together so when you build the castle, you know, they all fit in the same order. So when you keep stacking them higher, the foundation is strong. Awesome. That's how I explain it to a baby. Yes. I I I it followed for me. I hope it did. I'm seeing a lot of reactions. Seems like all the babies and the the three year olds in the audience definitely understood it. Yeah. Jomar? It's just shared resources. It's literally like the it's a file folder, and it's a shared resource so you don't have to keep building the same tool using the same tools, using the same data lake. It's like you build it once. Hey. Can I use another agent could be like, can I use that too? And the agent will be like, yeah. It's already here. Come join. And that's basically what a deep agent harness is. Awesome. Thanks so much, Jomar. It was so fun having you for almost an hour. Hope it was fun for you as well. We will hopefully have you back in some form or fashion, especially to walk through Stage 3. I saw quite a few yeses, so we have a lot of experts in here that wanna advance to the next stage in their journey. For those folks that wanna connect with you, ask you questions, is LinkedIn the best place to find you, or Yes. email? LinkedIn is I do have an AI agent on my other phone, but, I prefer LinkedIn because then I know, like, okay. You're from this event, because I go to a lot of events. So, yes, do a LinkedIn post. Add me there. I respond really well. I don't respond that well to sales calls. Thank you, guys. Have a great day. Thanks so much, Jomar. You're welcome to exit the stage. Just for the crowd, you can catch us at our next insider. It is on October 15, if I'm correct. The LinkedIn events playbook for b two b marketers. We're gonna be asking LinkedIn all of our burning questions. Should we be running ads? Do we need a lot of money in order for our ads to run properly? And what what kind of best practices they are seeing? And then also save the date. We have our big virtual event coming at the end of October. Just TLDR, it's gonna be the rise of the headless marketer. If you're interested, just be sure to follow along on LinkedIn. We will be dropping the registration page for that very, very soon. But with that, I really appreciate y'all joining us today on Tuesday for Insider. I will see you all at the next one. Bye, y'all.