AI for holding companies: sales, content creation & more

[00:00:00]

[00:00:00] Jordan: Amazing. Thank you guys for coming. Thanks for having me, Ben. It's a, it's an honor to be here.

[00:00:06] Jordan: Jordan mix. I am an operating partner at late checkout help like Greg Eisenberg run late checkout. We have. holding company with about six companies, five that are formally built. And then a couple of experiments that we're running. And our whole goal is to build and buy businesses that are powered by community and really just use our ACP.

[00:00:31] Jordan: audience community product framework on building companies and future proofing businesses. I'll share a little bit more about what what we do in each one of those different companies. But as you can imagine, like AI and automation is basically runs rampant. So what we're going to do today is just introduce you to all of our different Companies and how they work together, how we think about AI, show you some of our automations on workflows, and then do a demo of a tool that we just built for our sales team internally.

[00:01:03] Jordan: And then share some like secret tools that that we use internally and just get you guys thinking about how we're thinking about AI, because we really, truly believe that this is like an experimental phase of AI. And automation and it's a really fun time to just experiment. Yeah that's what we're going to get into today.

[00:01:21] Jordan: And then would love this to be dynamic and ask questions like Ben was saying. I don't want, there's nothing worse than just a guy talking at you for an hour. And you're that's not what I cared about at all. So we want to be valued. I want to be valuable and use this time to meet everybody as, and have it be as interactive as possible.

[00:01:38] Ben: Awesome. Sounds great. Yeah. We're looking forward to. So many companies, most of us have one or a couple of things going on. So it'd be really interesting to see how they all interlink and your learnings across them all. So yeah, let's get into it.

[00:01:53] Jordan: Let's do it. All right. Can you guys share my, see my screen?

[00:01:57] Jordan: All right. Yeah. All good. Cool. [00:02:00] Amazing. All right. So let's check out how we use AI to run six plus companies. So what is a holding company? Let's check out at the top. We started as a community Agency, a studio, which was like for incubations and a fund. That was like a three, three legged stool, but early in 2023, we changed this holding company model where late checkout holdings is at the top.

[00:02:27] Jordan: We've got a team of about seven that works at the top and then a team of about 110 across all six of these. Different entities and there's some experiments that we're playing with as well that aren't like formal businesses. So LCA is our product and innovation agency. They work with Nike, Dropbox, TikTok, designing what's next.

[00:02:46] Jordan: Design scientist is a design agency with a marketing lens that gets you more clicks, shares, and cash. You probably knew robot is our AI and automation playground. We've got a newsletter as well as a membership community, and we do some. Some consulting projects with LCA as well, community empire, a membership for internet builders, basically our playbook condensed into a membership.

[00:03:08] Jordan: Where you can build a internet based business. We've got about 400 builders in there. It's been really cool. Greg Eisenberg, people probably know Greg, it's his media company, sip startup ideas podcast. And then boring marketing is our AI and S assisted SEO. And they're just launched boring ads. So as you can see, there's a lot going on and why this matters is our approach to AI happens in these three layers, and it's important that.

[00:03:36] Jordan: Communication across all six of those companies is happening. So if someone tests in one playground, we got to bring it to the other, if it worked and implement. So this is really how we look at AI. And I think it's like fundamental to just go back to first principles here is layer zeros.

[00:03:52] Jordan: You've got to operate. You've got to figure out what you're trying to do. What are the tasks layer two or layer one is [00:04:00] automate, figure out how you can improve that with just some simple automations. And then layer two is adding in AI and embedding those into workflows and automations itself. So like I was saying a little bit earlier, it's like we experiment.

[00:04:16] Jordan: See what works and implement. I know Ben, you talk about this on Twitter a lot. It's just like you, this is an experimental phase. Like you just have to be trying new tools. Like it's, you might spend three days with one and then completely dump it. But that's okay. And I think people had a little or have a little AI fatigue and it's annoying to sign up for a bunch of just different tools to use them for two, two days or never.

[00:04:40] Jordan: But I think that those reps and experiments figure out what you should implement into your system. So yeah, and then our views on how AI shows up in our life. And we think about this a lot at LCA, like where we're doing work for some of the leading AI companies that I can't even disclose, but just the best AI is going to be invisible.

[00:04:59] Jordan: Like everyone's using AI as a marketing tool right now. AI assisted AI, this AI powered. It's going to disappear just the cloud companies went or like the mobile companies have disappeared. No one's marketing like mobile or cloud. That's just anticipated. So it's good to remember that like AI will be invisible and the best day I will be integrating AI beyond the chat box.

[00:05:21] Jordan: Everybody's cool. Let's throw a prompt box on there. But we really think about that in, in several different ways, like the interfaces that AI can show up and will show up in our lives. That's like what we spent a lot of time like thinking on and helping clients with. And that's the beauty of running like an innovation agency is you get to get paid to be on the cutting edge and think about the future of these businesses at the same time.

[00:05:47] Jordan: And then yeah, learn how to drive AI or AI will drive you. little scary, but also I think everyone realizes that. So these are like our main tools that we use. Some I didn't put on [00:06:00] here, but is the stack. So it's just a lot. There's a lot, there's a lot going on. And Ben would love a little like quick hot take break here and talk about this.

[00:06:11] Jordan: I was, I'm curious about this. Cause I just left chat GPT, my pro subscription. I might get it back, but I was just mad at at how chat GPT was responding to me like a couple months ago. And so I've been using clot a lot more, but yeah, I would love to jam on this real quick.

[00:06:28] Ben: Yeah, I think you made a mistake by getting rid of it.

[00:06:31] Ben: I'm getting it back after this tweet. Yeah, I guess it's I don't know, it's one of those things where they don't have the same release schedule, so for some period of time, one might be better than the other, but then when the other one takes over, you're just going to not do it because you have some brand affinity to Claude or whatever.

[00:06:50] Ben: I don't know. But in how we've Tested it a bunch, I think, especially recently, obviously, I think there's been more model updates and the reasoning seems to be better with ChatGPT. So the way it thinks through steps and like it can define a structure of something. So if it's a tutorial, if it's a blog post, the structure seems to be better with ChatGPT and I think it's because of its reasoning capabilities, whereas the vibes of the writing and how the writing looks and feels and sounds has always been better on Claude, I think the whole time, like for however long we've been using these tools. So I feel like this is. You can use one and just not the other but we don't like we do use both and actually we had Barbara who runs a content agency on a few weeks ago and she does this in her workflow is pasted the first step is to paste stuff into ChatGPT and get the structure out then send it to Claude to make it better And then it might be taking it back to ChatGPT to then look through this version of what Claude presented so Yeah, it's hard [00:08:00] because It's two subscriptions, but yeah, like it's difficult to not

[00:08:04] Jordan: have one totally.

[00:08:07] Jordan: Yeah. I've already felt I've already felt it to be honest, like I'm like six days and I'm just, I would like to double check. But yeah I'm open to two subscriptions. It's just, it's more annoying just all the copying and pasting. So anyway, if you had to pick one, what would you pick?

[00:08:23] Ben: Right now, it'll be Claude, mainly because of the artifacts feature, I think.

[00:08:28] Ben: That's a fairly killer thing that ChatGPT hasn't got yet. Yeah, who knows? It's difficult.

[00:08:34] Jordan: The projects and Claude are I'm loving it. All right, cool. So let's jump into some of the behind the scenes of our favorite workflows, and then I'm going to do a demo a LinkedIn like scraper that we built that I'm pretty pumped on.

[00:08:48] Jordan: So this applicant tracking system as we have a ton of companies and a ton of people that are applying and it was really becoming. Like hazardous on how we were keeping up with these these applicants. And we were running everything as no notion as the central database.

[00:09:06] Jordan: And so applicants would apply, we'd get notified via Slack and that basically started the whole flow. But the review was really clunky. The resumes were inconsistent. There was a ton of looms that we had to review and there was no like it. Who should we prioritize? Who is a good fit for the open roles and who's a future candidate?

[00:09:25] Jordan: Cause we have a ton of people that just apply and it's not a right fit now, but in six months, like it could be that they could be like the perfect person to collaborate with. So what we did was just like, get this kickstarted and turn HR more into filtering and review and final decision maker.

[00:09:42] Jordan: So we didn't have to do all of the the editing ourselves. So this is a an example screenshot, but I'm just going to walk through how it works. So like people apply via the type form that sends it to notion. And that's like notions like the central brain. [00:10:00] And so we get used Claude and chat duty actually.

[00:10:03] Jordan: To prompt and parse and send back to notion. So it basically scrapes through a prompt reads the resumes and gives a back like the very structured prompt, like any gap. Oh, yeah, we ask it basically. Are there any gaps in the resume? What's applicable to one of the to, to our projects?

[00:10:21] Jordan: And basically create fields in notion. And then the, we use sauna for meeting transcriptions. We've tried a lot of them, but it's just, that's the one that's been sticky amongst most of the team. And so we pump those transcripts and action items back into the notion database.

[00:10:37] Jordan: So we have one central ecosystem and now. Notion is like the brain basically of this whole thing. So that's just an internal pretty simple. The most complex thing was just drilling the prompts and it was like prompt sparring basically to create a template And we were just going back and forth with chat GPT and Claude a lot to get like a really tight prompt that was delivering results.

[00:11:04] Jordan: And it goes back to that experimental thing is just it's, it feels like a lot of sparring in, in today's world of automations and AI

[00:11:13] Ben: and is that built with, is that make. Workflow that does that was

[00:11:18] Jordan: me. Yeah.

[00:11:19] Ben: So that's what that is. Yeah, makes sense.

[00:11:22] Jordan: Yeah, that was make we use make and zap zaps.

[00:11:25] Jordan: And yeah, just depends really

[00:11:28] Ben: yeah, which one would you pick then if I put that question to you?

[00:11:31] Jordan: Yeah, my like automation engineer that we work with, who's an absolute killer is all about make just because you can do a lot like I'd say yeah, Make if you have coding experience and can do backend stuff and zap.

[00:11:45] Jordan: If you don't makes a little annoying to where they don't have a ton of instant stuff, like when X happens, trigger it. It's all wet. Like you can see this clock. It's always running for most of the, for most of the nodes. It's like checking every 15 minutes or [00:12:00] every hour or every two. So it just, if you have a lot of inbound, that's fine.

[00:12:06] Jordan: If you're using tasks when nothing's happening, it's it's frustrating. But make is like significantly cheaper. I will say it's probably like a 10th of the cost when you like get into the credits. And you can also do, you can also do like multi steps apps with Jason. And get things to have these components.

[00:12:24] Jordan: But one thing we're looking at is gum loop. I don't know if you guys have seen gum. Yeah, we had a

[00:12:29] Ben: workshop with Max Kovana. Yeah.

[00:12:30] Jordan: Oh, nice. Yeah. I really like what they're up to. Basically Gumloop, you can picture it like, if this is like a, just an automation flow, picture this like JSON file being like a huge other like component.

[00:12:46] Jordan: that did a whole separate thing, but that automation, you could take like a puzzle and put it other places in your organization. So like for this applicant tracker, for example, it's if this was like boring marketing's applicant tracker, we can just take the guts out of it and give it to LCA. Yeah. So I really like what they're doing.

[00:13:05] Jordan: Yeah. It was a

[00:13:05] Ben: really good workshop. Actually. It looks really powerful. I guess you've got an automation engineer, so it might be easy to just say, run with this and experiment, but it looks, and I think Max was honest about saying there's a lot to it, but it does take some learning. So yeah, there's that in mind, but yeah, there's loads of stuff.

[00:13:22] Ben: And then that's the sub. Node automations and workflows that look really powerful as well.

[00:13:28] Jordan: I would say keep your eye peeled on them for sure I would say it's a little early. To move over, and just ditch everything but that's the world that we're in right now, It's very tool subscription heavy, so cool.

[00:13:39] Jordan: This is a a podcast archives database, we're using Greg's podcast, they, he talks about four to five ideas twice a week with different guests. And he just re we just rebranded from where it happens to the startup ideas podcast. And it's really been awesome for growth and people are just like fired up about it, [00:14:00] so we're, what we're doing for our own internal use outside of like show notes, we're using some simple automations to read the transcript, give show notes chapters on YouTube are still quite manual.

[00:14:14] Jordan: I'll show what we've been using for the show notes. It's called swell AI. I'll put it in the in the chat here. But anyway, I think this database is a lot more cool. So basically what we're doing is we're watching YouTube and basically putting the YouTube transcript into. Appify.

[00:14:34] Jordan: And so we watch YouTube, get the YouTube, transcribe it, get the data set, run it through chat, GBTS script, parse the text, and then dump this into an air table that updates dynamically. So we're doing this for a couple of reasons. One, like this could be a cool database that anyone can tap into. We've put the top 30 ideas into a softer air table plus softer.

[00:14:58] Jordan: It's basically like softer is just. a really easy way to do front end stuff with air table as the back end. And I can show this stuff too. I just didn't want the demo gods to go back and forth. And I wanted to do the cooking show, make sure the the chicken is perfectly cooked.

[00:15:14] Jordan: So yeah, this is what we're doing on this. Like two, two reasons, like one have a database of all these ideas that we can start training. A model and using artifacts to basically like when we're sourcing an acquisition or wanting to start a new business, can we run this through like our startup ideas database of what we like in our ethos using our audience community product framework?

[00:15:38] Jordan: So that's one to other people can tap into this and use it as well. Three, it's a really good lead magnet. I think Greg's grown his audience his newsletter subscribers, like 12, 000 subscribers have downloaded this specifically in the last 60 days or something. So that's like all organic all organic content too.

[00:15:59] Jordan: So [00:16:00]

[00:16:00] Ben: that's

[00:16:00] Jordan: awesome.

[00:16:01] Ben: So with, and with some of that, sorry, that you've got like a rating system there, and when you're looking at. These ideas, are these ratings coming from ChatterBT assigned ratings, or are they, is any of it manual to say oh, we like this one specifically?

[00:16:16] Jordan: Yeah. So we, it's a, there's a swag that chat GPT gives.

[00:16:19] Jordan: And then we, when we come in and say this is manual, this viable tab here. And that basically means like, when we say yes, it's viable, it triggers to go to the softer. If it's no, it's okay, cool. That was like just a kitschy, like idea. That's like silly, like Apple vision pro bars was like dumb, by me.

[00:16:36] Jordan: So I can say that. But yeah, so the rating is we, these are suggested and we go and change them, basically, I'm, I pretty strongly believe that there still should be like manual processes in even automated systems. Like with this cold outreach tool that I'm about to show you I heavily think.

[00:16:55] Jordan: That the final step should still be like a human. Like you can use AI to draft, but I think like humans should be the final check and make sure that everything is good. And it should be a tool, not like a replacement in my opinion. So as of today, at least cool. Sales prospecting tool. This is where, what have I told you?

[00:17:19] Jordan: We could, I could send you a thousand leads directly in your target without being a scamming sales guy. And I also told you, you could look at their LinkedIn directly, rate the quality, get email and phone numbers and draft an outreach message. We built this with LinkedIn sales navigator and air table. So why we built this is we do a ton of content.

[00:17:46] Jordan: People like. A lot of the content, but they might not follow or might not book a call. And we're looking for leads, just like when we were doing partnerships or doing just natural sales motions for some of the agencies. [00:18:00] And there was like the content. Which is maybe prospects are in there.

[00:18:05] Jordan: And then there's the CRM cool. They booked a call, they're qualified. And we're like working them through the sales process. There was a huge gap between those two things. And we were using LinkedIn sales navigator, but the interface is just. Just not it at all. And it's just like very hard to even know who you've contacted or know, like where they are in the process.

[00:18:27] Jordan: So we built this prospecting tool that could basically live in the center and kind of be like the holding cell of potential leads or potential outreach. So how it works is you run LinkedIn sales navigator as a URL. You create this list was American. Design leaders who just changed roles.

[00:18:51] Jordan: So that's, this was specifically for design scientists are design agency that, that is like powers marketing teams and creates assets that not only look good, but also like convert. So our hypothesis there was if someone's been moved into a new role, they're looking for a win and they might be behind the eight ball and might need some support from our incredible design team.

[00:19:16] Jordan: We run this, just grab the URL. So anyone can run this T like form inside of our holding company, and we want to make this public. So if anyone wants to like, beta test this, please DM me on X or email me. I can. I'll drop my email in the chat, but we think this could be like a cool micro sass for people to use.

[00:19:39] Jordan: And so basically how you do it is just name the search, drop that URL in, put your name and then the number that you're the number that you're trying to get. And then I'll go into the demo here. So are you guys seeing the LinkedIn, the air table? [00:20:00] So cool magic. Oh my gosh, fresh chicken. It's amazing.

[00:20:04] Jordan: Perfectly seasoned. This is what you get. It's terror. This is terrible. But it's just, it's incredible data, but it's terrible to look at. So we use air table interfaces. I'm going to show you what it looks like in a pretty view. So this is, what'd you get dropped into? You can filter all of this stuff still.

[00:20:26] Jordan: So you can say like in, in ICP or like job title contains country, this was all American. You can filter stage. So say you've connected. And I just ran this the other day. So I haven't clean this at all, but you can see which, which stage it's in. So basically like how we've been using this is.

[00:20:50] Jordan: Over here, you can see like the name and like the company. So I just we just like skim this okay, cool. All right. What's like orchestra AI head of process design. I wouldn't say that's in our ACE ICP. So let's just keep cruising. Blue bottle coffee, global creative director.

[00:21:08] Jordan: That's pretty cool. We got Remy here so I can visit their profile directly just to check it out. We got a little mutual connection, blah, blah, blah. Okay, cool. I would come back here. Let's say I like, and I want to reach out and then I want to enrich the data. So that's running an automation in the back ends.

[00:21:28] Jordan: That's pulling from an API and we will get, if it's available, their email phone DM suggestion, an email suggestion of what we should do for cold outreach. So we've got a prompt running in the back for the DM suggestion, which is 300 characters. And you can see didn't land Remy's email, but we got his phone number maybe?

[00:21:53] Jordan: And a very like DM that's been tested to work. It's [00:22:00] very very casual. And so this was specifically for LCA. And we're ready to rock on cold outreach, this is just a suggestion. Let's see how good it is. A blue bottle caught my attention.

[00:22:18] Jordan: Yeah, I think there's some tweaking here based on the role And what we're like trying to do. I think like how well how we're gonna level this up is

[00:22:34] Jordan: Probably add some fields in here for what company is it? So that will pull from a specific prompts like what the campaign or outreach is for So is it sales? Is it collaborations? Is it that will tweak the messaging a little bit. And so we can refine and have the book of prompts, but really this is just, we're using this to like surface, interesting people and just connect with them.

[00:22:58] Jordan: So if I was actually going to connect with Remy right now and I can do it, I would just send without note and then see if he hits me back. I really, that's worked really well for me. And in my past, I built a mid seven figure agency, literally just doing this in the med tech space, it's just reaching out to people if they hit me back, then messaging them.

[00:23:21] Jordan: So it's a nice little outreach cadence. It's all, it also works for just like meeting interesting people on the internet. So if you want to do that, you can try that on Ben or me try to say what's up. But yeah, so this is. The tool any questions or anything we want to jam on before we go back to the press?

[00:23:40] Ben: Yeah, I guess just there's some like manual pieces to this as well as automated, like you want the stage to be updated automatically when you've done certain thing, but might not be possible knowing when you've sent a DM or whatever. And then, yeah, it seems like there's going to be [00:24:00] some rooting in your make workflow, where if it's collaboration, it goes through a different prompt and a different thing.

[00:24:08] Ben: Yeah. We've got a question here from Alex.

[00:24:11] Guest: Yeah thank you so much for this workflow. Looks really impressive. I have a question regarding the Airtable. So you mentioned it several times during your presentation. And as I understood, you use it a lot in your workflows. Maybe you can share a couple of your favorite features of Airtable, why you prefer this particular tool.

[00:24:32] Jordan: Yeah, sure. Airtable is the most user friendly database That, you can build interfaces on like this. You can build forms very simply. And there's also baked in automations that you can do. So it's just Google sheets, a lot of people use a simple spreadsheet, air table.

[00:25:01] Jordan: I, we just like it the best because it's the most like flexible. And modular place to put data. So people can add it, you can add it yourself. People can add it with a form. You can have things happen that don't require make or. Or zaps so we got workflows inside and it's just very malleable.

[00:25:26] Jordan: And you can build like micro SAS little internal systems. Like I've seen people build like incredible dashboards. Which is like for either sales or company tracking or let's just get like air table dash or and see some examples I don't know if this is an air table.

[00:25:47] Jordan: Yeah, I guess it is. Like the interfaces layout, like you can say like number of DM cents. So it's a good, like way to track your sales team or like proposals sent. And this all happens like within air table. So pretty [00:26:00] powerful.

[00:26:00] Ben: Yeah. Do you use any of the air table AI features within your tool there?

[00:26:05] Jordan: No. I haven't explored much on the Airtable AI. Honestly. I haven't heard

[00:26:11] Ben: many people have. Yeah.

[00:26:15] Jordan: They did a really good job rolling out interfaces where they gave everybody access to it for a month or something. And I think everyone got really like sticky on it and then they made it in the paid plan.

[00:26:27] Jordan: But the AI stuff, it feels like it's got the beta tag still, which I never, it's never like a, Ooh, I want to check this out immediately type tag. And I think it's in an additional payment tier. So anyway, like that's the, that is the downside with air table is like, just like Figma now and like all of these things, like your collaborators can stack up and you get a bill and.

[00:26:52] Jordan: It's in a whoopsie daisy, like we just spent like a thousand dollars extra for no reason. So it is, you gotta have a, you gotta have budget watch when you're like sharing stuff out and doing some of the automation stuff.

[00:27:06] Ben: Yeah it's, it gets very expensive very quickly. We're always pulling what we are like spending on there.

[00:27:12] Ben: It's crazy.

[00:27:14] Jordan: Yeah. This was the swell AI thing that we were using for the podcast transcripts. They also have a little editor in there too, but. The founder Cody Schneider is if you follow him on Twitter, he's great. Yeah, he's a legend. You're really good on

[00:27:30] Ben: the podcast with greg.

[00:27:32] Jordan: He's been

[00:27:32] Ben: on there three times, right?

[00:27:33] Jordan: Yeah. Yeah He's become like the golden sipper. As I joke with him he's the co host and residents. Yeah, Vegas has DJ residents and the SIP podcast has Cody residents. But yeah, so this tool is like probably one of the best for we've tried to do like manual show notes and transcripts through Chad, GBT and Claude.

[00:27:55] Jordan: But. Swell has been significantly better at, at [00:28:00] parsing the transcripts and getting like show notes that are formatted in the way that we can plug and play them into YouTube.

[00:28:08] Ben: I might have to try that out. Cause I'm with these workshops. I'm trying to get better. Like I'm trying to use the script, edit and cut off the bit where you and I just talk before the thing kicks in and get the transcript and get the summary and.

[00:28:19] Ben: Timestamps. And it's not perfect yet. Adam's asked how are the API AI API costs for some of these use cases you've got?

[00:28:28] Jordan: Pretty like minimal, honestly, they're not crazy. They're not crazy in the, under the hundreds a month for sure.

[00:28:36] Guest: Yeah.

[00:28:36] Jordan: Cause we're not part of it. We're not like.

[00:28:38] Jordan: Parsing like mad stuff. Like we're not blowing a ton of tokens. Like I think last month's like open AI bill was like 52 or something. Super. I think air table is more expensive. So yeah, it's like very, and they're like, the token cost is dropping like significantly. So is this like Descript or Opus?

[00:29:00] Jordan: They're trying, I think. I wouldn't say it's like a competitor. I would use it more for You can, but just it's mainly like scripts, socials, newsletters. It's mainly like taking video to text. But yeah, you can definitely you can definitely do. They're trying to add some of these features, but yeah, I don't think Cody, Cody would say himself that they're mainly like they're hyper powered on the trans transcriptions and stuff, clips and show notes, et cetera.

[00:29:30] Jordan: But yeah, I think their editor it's okay. I think their editor is okay. But yeah, Descripties is swell. I'm pretty, yeah, so cool. Yeah, just I like jamming on this stuff. Would love to hear what you guys are like, what you're thinking about or what your, um, What's your, what you're working on.

[00:29:50] Jordan: So I'll share a couple of like more secret tools that I can't really like show or talk about, but this is just like to trigger some ideas. I don't want to show like how we actually have [00:30:00] built them because they're like proprietary, but for DSI, like we have a vision model that grades designs that we're really trying to work on and make better for internal use.

[00:30:09] Jordan: So designers can more or less give their blessing and also attach a quantitative number. To design. So it's here's my recommendation and here's our models recommendation. It's 80 percent confidence that we think because of your inspiration, because your mood board, et cetera, that you're going to like, like this design.

[00:30:29] Jordan: And that's going to, it's going to likely convert based on your outcome. It's not perfect yet by any means. Like it's a huge, that's a huge task to be like, yo, this is going to, you're this ad's going to convert with a 1. 5 CPA. It's just and then same thing, but clients have been really liking this.

[00:30:47] Jordan: And we've been like tweaking this model a lot. Is that

[00:30:51] Ben: like website design or like you said, and does it like creative design? Is it for everything?

[00:30:57] Jordan: Yeah, we're trying all of it. But we started with landing pages and using like hot jar and some of the other stuff that exists out there as like the data and like giving the heat map from hot jar to match the model.

[00:31:10] Jordan: That's a little bit of the sauce there, but like I said this is not in like full implementation mode. If it was like. We would put it out into the world and be like, yo, anyone can grade their designs. If you want us to create designs, let us know. So we're very much like in that, like free tool or like SAS to service type model for the whole co that's just more like a business strategy thing than anything.

[00:31:35] Jordan: Yeah.

[00:31:35] Ben: Like

[00:31:35] Jordan: the

[00:31:35] Ben: HubSpot website grader

[00:31:37] Jordan: doing that kind of, yeah, exactly. It's you got to store your leads. You might as well, we might as well give you a tool to capture them.

[00:31:44] Ben: Yeah.

[00:31:44] Jordan: And I see a lot of sass moving that direction too like even beehive the way that they're moving and they bought tight form.

[00:31:54] Jordan: I think that It won't be long until they're a kind of funnel builder [00:32:00] from like maybe lead magnets to websites, landing pages to capture automations, drip campaigns, et cetera. Yeah. Type dream. What did I say? Type one. My bad. That's on me. Tight dream. Yeah. So yeah. And then this other tool that we have is just a briefing tool.

[00:32:18] Jordan: Like when a client says make it pop, it's that means absolutely nothing. It's like designer's worst nightmares when someone says make it pop. So we basically had a greater. And we give that grade back to the client and saying this is a three out of 10 prompt. If you want less iterations, you should try this, and this to make it at least a seven out of 10.

[00:32:42] Jordan: So this is a, basically a Slack workflow that happens. And it's it's been pretty helpful. Some of the clients don't like it though. Because it's like basically telling them to retweak their stuff and it's a little annoying So we've had a little bit of pushback on that. In full transparency

[00:32:58] Ben: Yeah, it looks like you've obviously you're looking at the pain like the pain points you're dealing with or anything manual that you're Coming up with in each of these businesses or across them all And then trying to figure out what's an AI version of this so we can help.

[00:33:12] Ben: 'cause like with clients giving you a design brief, lots of them are terrible or not clear, or just like you said, say make it pop. But in, I thought from this, you were gonna say from that design brief, you just make it better for yourselves rather than you telling them, Hey, you gave us a really bad design brief.

[00:33:32] Ben: You should go and make it better. Instead of saying here's what we think you meant in your design brief. Does this sound clearer and better? That might piss them off less.

[00:33:41] Jordan: Yeah, good. Yeah, I think it does make suggestions. They just don't like doing double work, like they want to push it into the machine and the next thing they want to hear from us is like a sick landing page that like, Looks amazing.

[00:33:55] Jordan: Or like a sick ad or a lead magnet or whatever we're designing [00:34:00] for them. Yeah, it's it's not good enough yet to honestly just be like, all right, cool. We improved it. No, no other input needed. So it's an, it's a nice way to do like client retention and less iterations to it just the first couple, they don't love doing more work.

[00:34:17] Jordan: So anyway boring marketing has some really cool stuff that they've built internally to help with their their services that they're providing. So it's we've got this AI assisted SEO research plus content engine. So basically helps with keyword research, research, domain authority, search volume, et cetera.

[00:34:37] Jordan: I think AH refs and SEM rush meets a Claude. Content engine. So this has been super, super valuable to basically consolidate some of the disparate information that exists. There's tools like everywhere, but it's really annoying to go to every different tool to get different data. And they've been able to basically use their expertise in SEO, combine the tools and also create content.

[00:35:09] Jordan: And the clients are absolutely loving the data and the output on the initial research. And then they're producing more content because the V1 drafts are just like way better. So that's been really cool. And then. In development as this ad platform that basically combines like data insights research from the meta ad library and create and creative like a Kanban board all into one so you can scrape the ads that are working for multiple different clients where the meta ad library.

[00:35:41] Jordan: I can pull it up if you guys want. It's just like one at a time. It's like really painful to look through. So we're like really in this phase of what's a painful interface that we hate, like spending a lot of time in and how can we make it centralized basically? Let me pull up the meta ad library here.

[00:35:57] Jordan: Yeah.

[00:35:59] Ben: I [00:36:00] think because a lot of people don't know about this library anyway, and you can just type for any company, basically all the ads, you can see which ones are still running and you can basically go back and see. Have they been running this for six months and if so you can guess that's a good ad and working well for them

[00:36:16] Jordan: Yeah, exactly.

[00:36:18] Jordan: So let's just do like stanley cups like say stanley wanted to hit us up or whatever So they've got 36 ads That are live that are launched. This month I guess and so you can like click in here and see all the ad details and stuff And you can see it's just like one at a time and it's just not an amazing Oh, that's good inspiration for my ad campaigns or I really liked this static.

[00:36:44] Jordan: So if you had a cup competitor, maybe you're like simply modern, the water bottle company and you wanted to like, Understand what ads were working. That's what the boring ads tool is going to do is basically like scrape Stanley scrape now, Jean, that's a throwback. I don't know why I went there.

[00:37:02] Jordan: What else is another, I can't think of another water bottle, Ottawa or whatever, scrape all of those, put them in a competitive landscape and see which ads. You like the vest and it almost creates like a mood board. And then with the Kanban like system, you can be like, cool ad idea based on this.

[00:37:18] Jordan: And it's basically just like closing the gap from research to creation.

[00:37:27] Ben: I actually saw someone recently on Twitter say they looked through the library and saw, I can't remember whose ad it was, but it was like. It was someone's ad who they spent, I think a ton of money doing the ad and it worked really well for them. And he just copied the same video and it also worked well for him.

[00:37:46] Ben: Which I don't condone doing that, but yeah, it's just do your own versions and you can do any AI generated versions. But Kesha, who's on our team has asked how automated is the SEO and content workflow. And actually I wanted to add to that, which is there's a [00:38:00] few tools out there that do something along these lines.

[00:38:02] Ben: So what is it about those tools that, You didn't like, or made you decide, actually, we want to do our own version of this.

[00:38:09] Jordan: Yeah. I don't actually know the full back down, background on it. Why that team decided to do it. It's not a public tool at all. So they're using it for like basically to productize the agency offer and use their expertise.

[00:38:25] Jordan: And it was also built in like February of last year. That's when we started it February, 2023. So it's it's been a minute, it's been a minute. But yeah, I would love to hear what those other tools are.

[00:38:40] Ben: Yeah. I think I don't know if you remember seeing on Twitter, there was a guy who talked about stealing loads of ad traffic or search traffic, and he did it with Causal which I'm an investor in Causal, and I don't think they liked that eventually, and everyone like went crazy, but tools like that.

[00:38:57] Ben: And I think, did Cody have something like this as well at some point? Some sort of way to generate AI SEO content? I thought he did, but maybe I'm wrong.

[00:39:07] Jordan: Yeah, it was called Draft Horse.

[00:39:10] Ben: Okay,

[00:39:11] Jordan: that's what his was called. That's right. I forgot about draft horse. I think they just had To focus honestly, I think it's still running.

[00:39:19] Jordan: But yeah, this is what he had.

[00:39:22] Ben: Yeah

[00:39:23] Jordan: This is actually how I found cody originally. Yeah when he was tweeting about this and you tweeted just recently about I got a little I think I got it saved in my bookmarks That seo. ai You Yeah.

[00:39:38] Ben: Is that cool? That's, I think that's one that Barbara mentioned as well.

[00:39:42] Ben: So I think there's quite a few of these and that's why I don't know why, what which tool is best and why, and what point do you get to where you're thinking actually I need to build my own version. And actually David has asked, [00:40:00] do you know if there's any punishment from Google for using AI driven content for SEO?

[00:40:03] Ben: I believe that's. They don't know,

[00:40:07] Jordan: but I agree. I agree. If it's like straight up generated, maybe, but I don't think they know either. I'm in the camp of the, of that as well for now, at least. Yeah. This is pretty interesting to me. I just saw this guy, Ian has launched this. I'm just like really curious about it, but basically like you connect your data source.

[00:40:34] Jordan: And this is just a really nice live demo on the website as well. Like just add this to the swipe file for sure. But you can customize your template and that's basically what this is. And so you can say rating one and it changes this to like rating two greater than, and basically just creates these articles based on your data sources.

[00:40:59] Jordan: That's interesting. So like ClickFunnels versus Builder all so this is like a programmatic play. I don't think it's for everyone, but wasn't that a cool demo on the website? Yeah, that's a good way to sign up.

[00:41:12] Ben: Yeah. Yeah. I think there's something we're doing at the minute with. If Zapier blog posts, a lot of them talk about like how to do a certain workflow or something, and then they basically embed the zaps to do that thing within their blog posts.

[00:41:26] Ben: So they've been doing that for years. And we thought of how we've got all these tutorials, but like almost 300 tutorials now, how do we write posts about them and then resurface tutorials around certain things. So if it's five different ways you can build an AI chatbot, how do we, cause everything's a notion for us.

[00:41:44] Ben: Notion AI sucks, actually. I found it really bad and just doesn't do very well. But yeah, being able to connect your data source and say, we're writing an article about this, find the best tutorials that link to this topic. [00:42:00] And then also we need to then write copy for that, but it's interesting.

[00:42:05] Ben: Really cool. We've got a question from Jacob. And I will unmute him. Jacob, go ahead.

[00:42:17] Guest: Hey, Jordan. Thanks for presenting. How often are you seeing your companies move away from some of these automation tools and then just build custom stuff in Python or with the LLMs? Because I know a lot of these automation platforms are really configurable and nice.

[00:42:32] Guest: But as soon as you hit a barrier or there's a gap, instead of just saying, okay, I guess we can't go forward with this. Typically people will just go build it themselves. I'm just curious how often your companies are having to make that decision,

[00:42:44] Jordan: if you're aware of it. I'm not a ton. We really like. Yeah, for most automations, we haven't hit like the threshold.

[00:42:56] Jordan: Like even that, like the LinkedIn scraper is all like low code the most sophisticated stuff is really like how you get the data. And like the API and stuff, but other than that, it's just like air table. If we, maybe if we put that into the market as as a software, we'd probably like.

[00:43:16] Jordan: Remove air table out of it because air table starts to get constraints at like 20, 000 rows or something So we'd probably build from scratch there. But yeah, I would say like very low likelihood that you'll ever like bump into to these like things on the routine stuff if it's like you need a lot of user interaction and like user permissions that can get a little trickier

[00:43:41] Ben: yeah and it might be when you're in agent territory like we saw with Chris last week he posted about if you enter, he built something on Slack, on Google sheets that if you put a company URL in, it'll go to LinkedIn.

[00:43:57] Ben: It'll search people, it'll pull in data that way and [00:44:00] find the right people in these jobs that you need to find. Like doing that, he needed to write an app script basically. So again, it's not no code, but he used Claude. He can code, but he uses Claude to do that. So I think there's loads of ways to get really close to needing a developer, but not quite needing to know slightly how code works or being patient enough with Claude to go back and forth on, look, this is exactly what I'm trying to do.

[00:44:26] Ben: And there's API providers. Like he mentioned rapid API. You mentioned you got appify. I think a lot of them with Claude, some custom code entered into a workflow and using these API platforms. There's a lot of that you can cover.

[00:44:45] Jordan: Totally. Yeah. I think it's like the classic, like 80, 20 rule or maybe 92, eight. You can get like 92 percent of the way there most of the time. And then I like, yo, Tony, we need some help. Let's bring in the powerhouse help us out.

[00:45:01] Ben: Yeah. We just got one more question from Adam then go to things you're thinking about.

[00:45:06] Guest: It's a little business idea to finish this up. Thanks guys. Just had a quick question. On a high level as a holding company, do you have. Or do you encourage the companies to have a AI governance policy? And is there any like specific framework that you have for how you encourage or guide AI use within the sub companies?

[00:45:29] Jordan: We don't, honestly, we're just too small in the grand scheme of things to allow that. That's a blessing in my opinion. So if we have no governance at this time other than like the NDAs that we have with. Some of the clients at LCA, the innovation agency. We don't have much governance as far as a direct policy.

[00:45:51] Jordan: We don't really use a ton of this stuff for like client projects. Cause it's mainly like [00:46:00] design and ideas that they haven't even built yet. So we're not dealing with a ton of it's mainly we just have to be quiet about it in the future. So there's not I don't know. Liability wise business idea I'm thinking about like bundling these model, all these models into a wrapper.

[00:46:18] Jordan: Like I saw this from Roberto Nixon, who just does incredible videos. And he's always playing with AI, but like you have to sign up for one, two, three, four, five, five tools to do this 15 second video. And I just thought is there an opportunity to like basically create a like simple front end interface for some of these that goes and like leverages these models. I haven't explored this idea at all. I just think it's very prohibitive to 90%. I'd say all of us in this room are on the like edge case of testing this stuff, but it's Very.

[00:47:01] Jordan: Prohibitive to like my parents, for example, or like they're not signing up for all of these things to even do a routine thing. So I think there's a really big opportunity in leveraging some of these things and like wrapping the rappers basically.

[00:47:20] Ben: Yeah, I think it's interesting and feels like what, so like TypingMind and Poe both have the ability to choose from like a list of any of these models, so Claude or Chatubitty and all that kind of stuff.

[00:47:33] Ben: And Stripe actually built this internally where you can just choose which model you want to talk to and you can change it and you can save your prompts and all that kind of stuff. And I wish Stripe released it because it would be a better looking product and a better user experience. Interface product that I think Type in Mind or Poe is.

[00:47:50] Ben: I think I don't know, there's something about them both and have both subscriptions, but I never use either of them. So I like this idea [00:48:00] with The visual models, like being able to do visual stuff and plugging into actually, I want to use runway. Actually, I want to use Pico labs. Actually, I want to use this new one that just came out.

[00:48:10] Ben: I want to use flux for this. I haven't actually seen that. And I think that's pretty interesting to see, but I wonder, it feels like a different, like a difficult, I don't know, maybe it isn't a difficult interface to build. Cause I think they're all fairly similar. I was thinking the LLM one is easy because it's just a chat prompt, but maybe it is easier.

[00:48:31] Jordan: Yeah, I don't know. I haven't looked like a ton into the API's and stuff, but Linus was like it was basically doing like an ecom try on trying to find it of where he was like putting on these sunglasses and Maybe here it is this is all like AI, try on stuff with what is he using with let's say I, so he basically like you load this product in and then load context images and.

[00:49:05] Jordan: You, you can basically create train models, but you have to train a model on yourself. If you want to do like self try on and you have to train a model on the product. And so it's a little bit of a lift, but you can see if you created like a mobile app that was literally like take five pictures and leverage the let's AI API, like you could do like in real time and just there's a really good chance of that, like creates a viral loop because it's just like of how the user interface shows up.

[00:49:35] Jordan: So we think about that a lot of just all this cool stuff is quite prohibitive to the mass population. So like, how do you wrap this stuff up in an interface that is approachable? And that's like what we get pumped about.

[00:49:50] Ben: Yeah, I guess it's what Greg talks about all the time, which is like the bundling and unbundling, like it doesn't feel like things have been unbundled, but if you think about it, [00:50:00] there's a model and a product that helps you create images of a headshot of yourself.

[00:50:06] Ben: There's a product of one that does like interiors. There's all these like individual ones that have often built off similar models or have their own training or fine tuning on top of them. But yeah, for someone who trying to make. I want this mid journey image, but I want it to be video generated.

[00:50:23] Ben: And then I want like this text to come in at the end and I want to do whatever it is. I haven't actually seen anything like that. is an interesting rapper product. That probably is

[00:50:32] Jordan: as well. Yeah. I think wrap the rappers is as a takeaway for me this week.

[00:50:37] Ben: New bundling and unbundling

[00:50:39] Jordan: is wrapping and unwrapping.

[00:50:41] Jordan: Exactly. Yeah. It's just go down to the family tree and it's all the same. LLM. And that's the root cause. Yeah. But yeah.

[00:50:52] Ben: Yeah. Thanks so much for all of this. Yeah, it's been a great session and yeah, given us some things to think about and a bunch of different tools to test out and thanks everyone else for joining.

[00:51:03] Ben: I'll try and get this edited correctly with the right AI tool and all the right timestamps and everything as soon as possible. And yeah, I'll have it up on the event page shortly, hopefully next couple of hours. Cheers Jordan, and we'll we'll have you on again soon. Amazing,

[00:51:16] Jordan: thanks. I, guys, I just dropped my email, if you want to hit me up, and, Jordan mix on Twitter.

[00:51:21] Jordan: So thanks for having me then. Cool. See ya. Bye. Yeah.

‍