90% Open Rate: This AI Monitors Socials and Creates Custom Lead Magnets

How can we use a agent AI to create actually useful, hyper-personalized? Lead magnets that's what Max Mitchum and ended in power user and founder of Triggery. Walks me through. In today's episode, trigger fire is basically a platform, to get all these different intense signals available on a prospect that might not be coming from your own site. what is an intent signal? there's all these actions that a prospect. Or a lead might be doing that helps you inform the playbook that you apply to basically win that deal. For example, events on the website. Did they click the get in touch button? Did they spend more than one minute on the FAQ site? But that's basics. Most off the shelf tools let you do that. So if you're doing that, you're not really ahead relative to your competition. What Max is trying to do at Trigger Fire is basically give GTM teams access. Toward these social intent signals, the kinds of things that those leads are doing on their social sites to help inform what that best outreach might be, or that next step in your playbook might be. once you have all those signals, you need to do something useful with it. one thing that Max is doing is mapping out the ways that like all these marketing and sales signals and whatnot. Can inform like a unified strategy for getting back to these people. that's a whole big, body of work that he's working on, but to highlight exactly what he means by all this. In today's episode, he walks me through a specific part of that playbook, which is basically getting all these hyper-specific, signals on an individual and creating a really personalized lead magnet. to reach out to them to basically cut through the noise in their inbox. So on that note, take it away Max. The other Max, not me, of course.

Hey, Max. How's it going? Not too bad, Max. Funny saying Max, Max. How's things, man? going really well. I gotta say you are my first max guest. So we'll probably do something cheeky with that in the promo. how's it going, mate? things are busy, but, I'm excited to get into this one today. It's going to be good fun. Yeah, absolutely. don't think I've talked with a builder in 2025 yet who wasn't like, Oh, it's busy season right now. but before we get into your awesome use case today, would you mind introducing yourself everyone? yeah. Hey guys, I'm, Max Mitchum. I'm the founder and CEO of a company called Triggerfy. is a social Signal. platform. So it does, pull social engagement from various different social sites and runs social listening. So in this particular use case, powered with NHSN, it's pretty exciting So we can start to Do some really cool agents off the back of social data. Very cool. I think that's a perfect segue, Max, uh, to use case. what did you build and why? lead magnet agent. So this is very much geared towards sales and marketing, professionals. because if we take trigger five, for example, we're a small team, but my goal is, how can we continue to increase our user base, and essentially, grow more clients, In order to do that, I needed to create. An outbound campaign that was beyond any sophistications that you could get with other basic, workflow builder tools, which any time basically gave me. But it actually, like, before I dive into it deeply, I think I'm going to share my screen and showcase. an interesting, where this agent fits in and part of like a grander, vision and I'll zoom back into this particular agent. built this framework or playbook essentially called like the network effects little a concept of combining like 20, 30. In some cases like 40 different like micro automations and micro agents together The idea being that you combine marketing and sales in a way that's never been done before. every action that an individual takes. Feeds another action, I'll give you an example and then I'll zoom into the lead magnet agents specifically. But if we look at, like, an outbound perspective, and if you're posting on social media, and this is where it's interesting with trigger fire you can collect all of that engagement and reach out to these people But when salesperson takes a call. in this particular use case We can pass the transcript that we're doing right now automatically through to what I call knowledge base. This knowledge base is a trained, LLM using a vector database doing a rag retrieval without getting too complex there to pull that data out. But the idea is that every call that's ever done, every webinar that's ever done. every podcast that's ever done. feeding this beast so the beast always has context to what's going on. at the company able then can start to create so many different agents off the back of that newsletter agents, youtube script agents, linkedin agents, Blog agents, you name it, it can start to create like this web. Of agents that then starts to feed everything and it becomes this like flywheel motion But diving like specifically into the part which i'm excited to Showcase is this lead magnet agent, And The concept is simple. could I reach out to you? and create personalized. magnet for you. That'd be very interesting. for you to read. So first of all, what is a lead magnet is. a piece of a research article or, here's like 10 steps to improve X, Y, Z, Whatever it effectively in layman's terms, it's that stuff usually got to give a sneaky email address, right? yeah. Yeah, exactly. how we're doing in this example for today i've just used a chat message But what we actually do is we take data from trigger five. we scan Social media to find when one of your prospect engages with a piece of content that is topical to your business So for example, if I take emacs, let's say you're wanting to reach out to maybe a Had a growth or a VP of sales or something, but you want to wait until they engage with some content around like automation or workflow building or something like that, Triggerfy picks up that it's happened and then sends the data through to this lead magnet, agent saying, Hey, this max person has engaged with this piece of content. And what then happens is it goes into our first agent, which essentially reads the post, the topic that you're interested in, automation. And so if we zoom out a sec on this, Agent it's like we're combining, one two, three, four five different agents together all doing various different parts of this, lead magnet slash research paper research, paper, et cetera. a bit I typically find that when an agent is doing a soul task. The outputs are so much better than when they are like trying to run everything altogether, just from trial and error, like I've done it before, I, I have one agent, compile everything together. You can get good level of results, but nowhere near that complexity that you could get to if you started to combine agents together. And break down their tasks into micro tasks inside of the larger workflow piece and you can get more nuanced in the prompting and it becomes really hard to prompt one agent to do everything versus individual agent and individual prompting on an agent and then combining it together at the end. England, China, and maintain it, it's probably easier because it's more modular. with the system prompts, you change a little phrase here. It could change the whole output. It's going to be isolated, To one

correct. Yeah. You can see I've pinned all of the data, Because what I do I just reiterate multiple times with a specific agent, until I get the data that I want. also, it's also quite helpful from a cost perspective from that essence, because this thing, if I start running it, this thing will take half an hour. To run like it's a long time to run it. And so when I break it down into sub agents as well, like I can actually tinker with it rather than, Thinking, oh crap, does this work? And then, wait a half an hour for it to run and then go check it again, It makes a lot of sense. how it's working in principle is we'll, dive into these, various different elements here. we take the data, the social data initially via a web hook, right. And through into what we call this, query builder. is taking the, topic this is doing really high level is this is saying, okay, let's take, this was a topic which was like, build me a lead magnet, which shows how you can use trigger fire to book 30 plus meetings per week. it takes that subject area and breaks it down into four questions, this is basically saying Hey, here are the things, if I was to give this to another agent that I would research. And so this is where it's really handy to break down Into sub agents because I can almost use an agent to prompt another agent. And that's effectively what's happening here a few things these lines get a little bit confusing, but everything that I, have all these agents, are hooked up to two These tools are perplexity And we're using their deep research capabilities. with inside the perplexity tool and this thing called like the trigger fine knowledge. hub. So That's what I was talking about. When we looked at that original, grand, like, Miro that is the LLM that I'm using that has, using a vector database, inside of that. It's another tool that we basically using, and we're just querying on that tool. comes in, it queries the, agent it creates the four different topics for it to start searching and then it goes through, to What we call our research, agent. And so here, it basically, feeds the topics through to this particular agent where it then starts to use, the knowledge base, which is trigger phi or perplexity So inside of this, lead magnet agent or inside of this particular research agent, I should say, There is context here that says, well, hey, like, if this is a trigger fire related question, go speak to the trigger fi. knowledge base if it's Outside of trigger fives domain. use perplexity to build the research capabilities And always use citations as well. So it's going to start to keep the citations as it goes through So that it basically. Thank Okay. So you basically got a trigger file for that really hyper specific, information, Your primary source. And you've got the fallback of deep research on perplexity to fill in the gaps when it's not available basically in your correct. Yeah five. Correct. Most people wouldn't necessarily need to use a knowledge base. I just like to use it because I can create like quick Hooky lead magnets, which I can post on linkedin. I did one the other day And it got three and a half thousand comments people wanting to see the lead magnet on linkedin, which was just like mind blowing and that came through this, which is pretty cool. Max, you may have to help a max out here. Get some leave magnets for and, but anyways, I digress. Please keep going. so what then happens is that research then starts to run on this particular, agent here and it researches those topics. So you can see it passes it through into this like one large, research elements. it does the initial research, which structures it. But one of the key things is it then feeds it through to this, data set, which then starts to basically. run the actual, prompt. picture of this, Cause it gets a little bit complex here. And this is where we combine the agents. We run the query to find the four subjects. It then runs research on those four subjects and works out the prompts that it then needs to use based on that research. to give to the project planner. planner then takes the research plus the prompts and then creates, the table, or the format, i. e. chapter 1 will be about this, chapter 2 3 will be about this, but it still maintains those prompts, which is the key part. because now what we've done is we've got the four subjects, We've run the initial research on that, We've passed it through to there, which has created the chapters, and et cetera, and the structure of the magnet, which then essentially is, giving it the prompts which we then feed through to this team of research assistants. if we have a look at one where we actually did, an execution. I'm going to Share this tab instead. the data from here and it's split out into 10 different items. one agent, but it's like 10 different research parts going on here. what effectively is happening, and this was a bit of a pain to build this particular query, we are saying you're writing this and this updates as the agent's running It's like you're writing this chapter of this part the previous chapter, was x and the next chapter is x so always has context about what is previously written what it's about to write essentially. I'm really glad that you're showing this because my own experience in building multi chapter stuff, it's so important that it knows what it previously wrote, what it might write in future. Cause otherwise it's going to repeat itself. It's going to look sloppy, Like a human wouldn't necessarily do that. This took me ages, for some reason, I just could not get it to work, for a while, the actual expressions inside of here, and it was a good feeling once we did it, the indexing of it was a pain, so then you can see what it's also given it, Current chapter, the chapter title, but also the prompt that you need to use to run that research. and so that's what I was saying earlier. Like we created the prompt based on the research, which then it like then starts to do. So in here on the bottom guys, you're seeing like the rendered result of, the expressions that max put in there. this is like an example of what the AI agent is getting for this item, it's really just focused the research for that one chapter, Yeah. So you can see here you got, write a chapter on this article. it has the context. Write a chapter for this. That in this chapter explain the fundamentals. So this is like the prompt. It's fundamental shift from Traditional prospecting methods to signal based prospecting. Start by highlighting the limitations, et cetera, et cetera. And then it gives it like a breakdown of what it should do. And then it gives it like a conclusion, this is how you finish it make sure to use trigger fires knowledge hub. and so then you can see for everyone it is then talking to, the knowledge hub which will then like, fire through the questions to, The trigger find knowledge hub and fill in any gaps using the perplexity tool, that we wanted to do it's weird, it's like one agent, but it's actually 10. because it's 10 different agents, Writing 10 different Subjects. in the 10 chapters I should say in this case where it then feeds it through to just a simple merge we merge the previous data, which is effectively the title of the research paper essentially, and then all of the, chapters that it just did all of the, of the actual, lead magnet, and the titles, like the subtitles that we're adding. So like the chapter titles, then it's taking the research that the actual agent basically spits out, and basically merging it together to format. So You have this, you have title, the prompt and the output itself. So it structures it in like a way that we want it to be structured Before we send it through to a code node, where we've basically just written some, JavaScript to merge, the parts that we want to merge into one long, form of content, which, you can see now it's like combined data coming in the different of forms, Which is the different chapters engagement data. And let's see. Is it HTML formatted at that point No, not at this point. It's all marked down still. you can convert it. here is where you would convert it into HTML, um, okay. So the inputs coming in as HTML and you're having it as Markdown. you can, specifically told it to output it as markdown, the last agent is the simplest agent, It's like the editor. it takes all of the work the previous agents have done, which is coming through now from this code node, here, and then just starts to write the particular magnet together in a style that you would want. it to style it. So some cool things that you could do that I haven't done if you didn't want to use this as from a webhook if you wanted to tap to it You could get the agent at the very start to say what style do you how long do you want? it? And then it could this is where you would add that part in here. So like, I've done stylization guides, Content review instructions, you can really flesh this one out. But it writes it Um, it's just like, I guess also localization probably. if you want it to just to have this out in Portuguese, that should be pretty exactly. do. Yeah. Like you can create it into anything. which is quite cool. Like you had different languages, you could split it out and you could have an English agent, Portuguese agent, whatnot. And it would like feed that data through to all of those agents and then send that to wherever you'd want as well, which would be quite cool. one thing that i've not talked about the actual llms that I am using in here I use claude. for Essentially. Okay. I was gonna ask, because I was seeing the philanthropic chat model, then I assumed it's because you're doing creative writing and stuff as well, yeah. I find like anthropic is just performs so much Better, but the other thing that I notice is I am requiring it to be, as you've seen, everything is like a structured output that I have coming through from these things. Now, I'm actually using I'm using the thinking model, because. I've just like literally just from trial and error. I've used oh 301 Just the 3. 7 thinking gives it to me pinpoint the structured output that I want every time without it. Like hallucinating right model for the right job, right? I think some people, especially some engineering folks I talked to at a moment like this might start talking about costs. And obviously that is a concern. But I think it's always consider the context. How often is it going to run? How important are these outputs? If this is the lead magnet to try and convince a human decision maker to take, action that costs money. Is 0. 02 really going to make a difference? Even at scale, hopefully you have a pipeline where, the ROI is there. So this isn't the case. We want to, cheap out or use, a small 7B, model or something like that. Yeah, exactly. you could do that, but the cost, like it really isn't, it's not that astronomical. this thing doesn't run like 24 seven. like you said, you need the output to be pinpoint because otherwise. when you read these things, it looks crap and Hey, what we've done here is we've basically created deep research, Which costs you 200 bucks a month to go purchase with open AI, this is our own version of a deep research And one that you can control way more, right? And make it absolutely perfect for yeah, give it your own Stamp onto it one thing that i've not done yet But i'm going to do is i'm going to feed it all of my social media posts learns my style of writing and so the editor then has Stylation guides based on how i've written 20 000 words in the past. Literally looks like me doing it. like the easiest part, but I love it. I just, it's it adds bow to the cap, because what then happens is it creates like a Google Sheets document. It adds the text to the then turns that Google document into a shareable document and then gets me the share URL. then on the other agent, which I have which is in live production then sends it through, to a, either like a LinkedIn automation tool or a, email automation tool. So it gives them that shareable link. if we zoom back out a second, just have a look at what's happened here. I've noticed that you max have engaged with, automation content on social media. I've prospected you automatically through using Triggerfy, which then creates the webhook and triggers this lead magnet and says, Hey, they've engaged with this post, this post is about X, Y, Z, topic, create me a lead magnet around that. It's then gone and ran this it's then created a share link, which then sends that data back fire HTTP to one of these particular tools, which then basically says, Hey, Max, can see that you're really interested in automations. pulled together a document that I've spent the last couple of weeks working on. It's about like, how you can leverage automations and scale your teams, whatever. here's the link, ps go do xyz or whatever. it is, and here's the fun thing is you could get the magnet I've been playing around with this now, but it's have it like, research the company So how let's pretend you don't work for an a10 apple or a random fintech company you have research that company and work out ways that they could apply Automation a company and so then it does this whole document contextualization around that as well And just it looks sensitive user crap. their team just spent like a week trying to get my attention. wow. Yeah. Fantastic. exactly. It really the question is can you get in front of someone because if you can This stuff for me testing it so far converts it 90 percent of people Want to see this magnet and turn into Really. That was going to be Cause you see a lot of influences, let's say build some basic template and they're like, Oh, it's super effective. yes, but has anyone actually seen this in driven action? So you're saying out in the wild people are clicking this, and it's driving action. Yes. 90 percent positive reply, And so the question is then can you hook them on the initial part? the campaign that I'm using to run this, one in four, people that we reach out to are interested because there's the context of we've seen their engagement, social media. So it's a really high, response rate and then a very high positive response rate on top of that. how much would you estimate the cost is like ballpark for a generation? it's really hard because it depends how much research it depends if it needs to use perplexity much or not if it doesn't It's incredibly cheap because it just taps into my Triggerfy knowledge base. If it needs to go and run a load of research inside of Perplexity to add flavor to it, then it can get a little bit more, expensive. But we're talking, cents, to 20 cents maybe to run this thing start to finish. it's not a lot at all. very viable for so many different use cases and whatnot. yeah put it this way I uploaded 10 bucks onto like open router play around and test with this one and I ran 50 different magnets now and we're at eight quid 50 remaining basically it really doesn't touch the still some scope, Yeah, exactly. Yeah, I'm guessing you've thought about the different applications of, this general pipeline, any first ideas from you, what you'd do maybe next with this or how people could repurpose this. one thing that I do have in another one actually, but it's more general, this one is very prompted towards trigger five because it's like a trigger five magnet, but I've got another one, which is more of like a generalization. but I would basically add, prompts to create images, or retrieve videos from a drive or wherever to then put into the document itself. we could have. some form of LM create an image to put emphasis on what it's talking about. and put the image inside of the document, which is really cool. and then The other option is, if you've already got an arsenal of videos that you've done in the past, shorts, you could label that correctly inside of an air table a drive wherever you want to store it. And then the agent can go, and look through the description of that video and be like, That's a perfect one to add into this section and add that video in there. it just takes that to the next level of adding personalization and really cool, like, relevance to that. you've got your asset bank and whatnot. with a tool can flexibly build that out to where some SaaS app have that. exactly. Max, I'm curious, why did you choose the tooling that you did to build this out? versus, other tools. for me, it's the complexity and the control. that you can get down to with the agents It's number one. I also really like the, it's ironic. I only use it once, inside of here, but I love the fact that I can add my own code, into the system as well. when I get stuck on working out the what not know to use, it's just easier sometimes just to code the thing in, implementation, um, as well. But it's the agent complexity. I'm only using tools agents in this, element, but I've done, some, other agents, in the past that add, for example, like this one that has like conversational agents, which do plan and execute agents, creates plans based on inbound requests and stuff like that. Just like the functionality that you can get. you just can't get in the other tools. then I've talked about a little bit why we use Claude and stuff like that inside of there But, know, it's quite simple, I think, the sense that one I just showed. because it's really just using NATEN's Agents knowledge base. And perplexity, combining that together, effectively. there is complexity there, but if you break it down step by step, I think no particular step is beyond the comprehension. or beyond the capability of anyone, watching this to be able to build that themselves. No, only complex step is the, read the previous chapter. Here's the next chapter. the only part that gets, a little bit complex, but otherwise, like you said, broken down in isolation, it's actually quite simple. that note, is this gonna be, or a version of this, something that's, free for folks to check out and learn from? I can track this wherever, wherever you want me to put it max. And we can have all of the prompts, everything to put in there for sure. And then people can tailor it as they see fit. Awesome, that'd be great. Cause I think this is such a great inductive learning resource, especially like studying those prompts and what on being able to them. We do have an official template library on nnn. io slash workflows. If you upload on there, even without signing up for it, this is going to be a shameless show, even without, signing up for our affiliate program, it does start tracking affiliate revenue. So if your page gets signups, you can get a bit of bread as well, which I think is totally fair. but yeah, if you upload that guys, there will be a link on this video, wherever you're watching it, somewhere above below, check that out. Max, thank you so much for showing this to me. I think it was really nice to see a real use case that you are using that is creating real value. I think that's important because a lot of us, we're trying with our little personal agents at home and stuff, but it's nice to see it, doing real, creating real business value. and truly any other cool, neat use case that you have. Send me a DM. I'd love to have you back on the show, mate, and you show me, what you're working on. I got it. I just got a ton of agents man Like I get A little bit lost sometimes and then I realize i've just built something for the sake of building it and then actually Use it but uh So we can have a meeting of the Max's and we show Yeah, exactly. Yeah, just nerd out for a while exactly. sophisticated pointless thing that you've actually built Well, Max, have a beautiful rest of your week, mate. and I'm going to hold you to Uploading this online because he said it on camera now, brother. Yeah. gonna do it. Don't you worry? We'll get this one out there. I think it's a cool one for people to have lastly, if people like what they see, where can they go? follow you and keep up on what you're sharing. go check me out on LinkedIn, max Mitchum, on there. I post a lot of my stuff on there. Pretty much every other posters and a 10 s. and then I just launched YouTube channel, last week, so I'm trying to catch up with you, mate, but, I will be posting more in depth stuff, on that, which will be pretty much centered around all of the Asian stuff that I do. probably those two places, man, are the best place. Very cool. it's always good to have another max making edit and content, but we might have to create a little leaderboard at some point and get a little contest going there. by the way, mate, people love in depth content. I've done some one on one stuff, which people ask for as well, but there's the longest video I've done. I was worried people like, nah, go deeper, go more. So I think this end up stuff like, You getting in the weeds here. It's going to be very interesting for people. Wicked Awesome. hear. Yeah, that's the concept of the youtube channel. mate, have a beautiful rest of your week and happy flowgramming, Max. See And you buddy. Thanks for having me on

Thanks so much Max. I love how you're thinking about this, nerding out on how to evolve the status quo and what people do in GTM today. And I think NAN is fast becoming like a dream toolbox for go to market teams, so hopefully this might inspire you, how you could apply and it and, and tools like Triggery to really up your game relative to the competition. If you want to see more, go to market. GTM use cases. Do drop a comment. because I use comments basically to inform what videos I do next. I'm Max. This is the studio. You are awesome for watching this video and happy Flo Ming.