7 AI agents we built to run our company.

Cleaned and speaker-labeled to match the edited replay. Tool names and proper nouns corrected, no editorial rewriting. Recorded Sunday, July 12, 2026.

[0:00] Justin: Quick agenda: we'll showcase the seven agents, how you can collaborate with us, and then a Q&A at the end. Alec, I'll pass it to you. What is an agent?

[0:05] Alec: Thanks, Justin. I'll contrast it with how most people use AI, which is conversational AI: you're in your browser using ChatGPT or Claude, asking it for answers. The difference with an agent is instead of asking for answers, you're asking for outcomes. You're asking it to take action on your behalf. It's hooked up to your tools, your emails, sometimes iPhones, all your systems, with specific roles and actions it takes for you. That's the contrast: conversational AI, which everyone's using, versus an agent, where you ask it to take action on your behalf.

[1:20] Justin: Cool, so who are we? I'm Justin. I sold my first company out of my college dorm room, and I've spent the past decade in growth, go-to-market, building agents and automations, helping a lot of companies grow. Everything I'll show you, I build and run inside AI4NTP and for our clients. Ian has been building software for decades, runs a global business, worked with public companies, and has a fleet of agents that run his businesses every day. Alec was inside a really big company, quit, and they hired him back to build all their agents, and he tripled or quadrupled his income within 30 days of going off on his own. That's the three of us. We do this for a living. E, do you want to go first?

[2:35] Ian: Yes. Before I show you the tool, some backstory. On my Mac Studio, I have an OpenClaw named Atlas. I call him the king, because he controls and manages all my other OpenClaws. A few months ago, Justin and I did a call with an SEO expert who's getting paid tens of thousands a month, and he showed us a tool for AEO/GEO, basically how to be found by AI. After the call, I went to Atlas and said, can you build this? He said, hell yeah. It took about a week of back and forth, and we built a tool called Pulse Check, where I could plug in all my sites and see if they're mentioned by LLMs. Justin said other people would find this useful, so I turned it into echocheck.app.

[4:05] Alec: I think it's funny for anyone in the audience, that's the difference between agents and conversational AI: you literally hear Ian referring to Atlas as if it's a person, because it can take action as if it is one.

[4:20] Ian: That's true, I like treating it like a person. So this is Echo Check. Right on the front page, you can plug in your website, AI4NTP.com, and run a free check. This one's doing well, 70. Good job, Justin. It tells you a robust output: page speed on mobile needs to be better, AI crawl access is really good, we need better FAQs, there's a lot of video content AIs can't find well, doesn't have that schema in there. When I started out, I was at an 8. I started implementing what it told me and bumped my score up. You can see how LLMs see your site: they don't see images or JavaScript, they see schema markup and text. It's a Matrix-style view of your site.

[6:50] Ian: As I started implementing this, plus programmatic SEO, another suggestion from Echo Check, this is our Google Search Console. Before May we had about 26 indexed pages, very little. I started implementing these things, and we went to 2,000, 7,000, now all the way to 14,530 pages that Google knows about us. And this is our traffic. Today, Sunday, 108 visitors already, twice as many as our big days before. So this proves the AEO, GEO, SEO stuff works. AEO is Answer Engine Optimization: you put FAQs on your site and AIs know what to answer. GEO is Generative Engine Optimization: the AI summary at the top of a Google search.

[8:40] Ian: A few months ago we were getting zero demos on BrandSauce from ChatGPT or Claude or Perplexity. About a month and a half ago, we had five demos booked in a week. In fact, it was all the demos that week. They all found us through ChatGPT, and they all became clients. I was shocked at how fast and how well this worked, and it was all thanks to a dare to my Atlas King. So check it out, review your site, let me know what you find. Back to you.

[10:35] Justin: I'll jump in next. The big idea here is how I build a month's worth of content in one day of work, thanks to many agents. Anytime you see this emoji, that's an agent at work. This is what a month of content looks like: our Instagram, our YouTube shorts and long-form, our newsletter, all on brand. Content is very time-consuming, and if you outsource it, it's expensive. To get this output I'd need three or four junior marketers at 60 grand a year each, north of $200,000 in headcount. And the moment you hand content to an agency or someone new, you risk losing your brand voice.

[12:40] Justin: Here's the one thing to take away: this is not one big agent. It's a chain of specialized agents, each built for a single step. It kicks off with a phone call between me, Ian, and Alec about what our next show should be. We record the transcript, and an agent turns that transcript into a landing page. That creates an Eventbrite, a Zoom page, a newsletter, an Instagram post, and it's distributed across the internet to get you in the room. We keep a human in the loop for quality control, because I don't want to send slop. Then this hour becomes an hour of content: straight to YouTube, with an agent optimizing the headline, image, and description.

[15:00] Justin: Then we take the transcript and post it to the website, re-optimized for ranking in ChatGPT, Perplexity, Claude, and Google. And an agent turns the 60-minute session into a bunch of shorts. If I did this without AI, it would take days just to make a handful of shorts. To win in this day and age, you have to be everywhere all at once. That's how you collapse time. Every tool we talk about goes to a tools directory page, all AEO/GEO/SEO optimized. So page creation, distribution, YouTube, shorts, landing pages, tools, Instagram, each one has its own agent. It's a team of specialists I've built over time, each doing one job well. One day of work runs all month, compounding while we sleep.

[18:00] Ian: The important part about what you showed, Justin, is that it started with a phone call between humans about what to add and how to add value. Then the agents take it. The agents aren't creating stuff out of thin air, they're taking what we find valuable and amplifying it, which is the prime use for AI.

[19:35] Alec: I'll jump into agent number three. My Miro board doesn't look as good as Justin's Figma, but we'll roll with it. Quick extra value: five minutes before we jumped on, I went to Higgsfield, dropped a prompt, and it built this graphic for me. That applies to everybody making a presentation. So, AI agent and iMessage. When you get a green text, your perception is it's scammy or automated. A blue iPhone text indicates human. So one element here is that a blue text reads as a real person. And the agent piece is it takes action on your behalf. I have a business where we get paid per appointment, so we need to get leads booked and showing up.

[21:05] Alec: What happens is people opt in but don't book. So we have an AI appointment setter using iMessage. Once we get the trigger that someone hasn't booked, we send automated texts, with tools attached to send emails and links. Say you're an auto tinting company running ads: instead of your team taking calls all day, an iMessage agent handles it. Once someone opts in, they get two texts asking the questions that factor into pricing, like what vehicle you have. We add a delay to make it look human. We can use these agents for following up on unsigned contracts, reengaging cold leads from a CRM, and handling responses. With our model, we get paid per appointment, so this saves money (no person doing it) and makes more (more appointments booked).

[25:50] Justin: I've seen you build these for industries beyond automotive. Anyone who requires instant lead follow-up and nurture benefits, because speed to lead is very important. If someone submits a form, we want a response immediately. If we don't get that response, we'll search for another vendor. So this increases operational efficiency (you don't need an assistant following up all day) and it's a key revenue driver, capturing every lead and converting them right away, hands-off.

[27:15] Ian: Let me share my next one. My wife and I started working on a school a year and a half ago, called Saints Classical, a classical tutorial in Middle Tennessee. When it came time to launch, she said we need parents and students to log in, and teachers to enter classes. It kept growing, and I realized I was building an LMS, a Learning Management System. I'd never seen one, because I graduated before LMSs were a thing. So I brainstormed with Claude, then Claude Code, and built the first iteration. We launched with 95 students, when I thought 60 would be a win. So I installed an OpenClaw on the server to build the next iteration, brainstorming with my bot named Saint.

[29:50] Ian: You can view it as a teacher or a parent, see all your kids' classes, and we have houses, like Harry Potter. It's a tutorial: they meet Monday, home Tuesday, back Wednesday with high-quality teachers, home Thursday where parents lead with the lesson plan. It did such a good job I've been showing clients, who want to use it for training their employees. Beyond that, the OpenClaw runs a blog, posting 2 to 3 times a day, attributed to C.S. Lewis, all relevant to the school. It's over 350 blog posts, which gets us past Google's EEAT mark. Largely because of programmatic SEO and this AI automation, we're launching with over 230 students this year, a two and a half times increase in enrollment.

[33:00] Ian: The parts I failed to mention: it's way beyond an LMS. It takes tuition, allows enrollment and clubs, has its own email system, and I built Slack into it so you can create groups. I'll launch the iOS app in August. Having an OpenClaw on the server, if problems come in, I copy and paste the email my wife sends me and say, fix this, then email them when it's done, from me. And he does. It works like a charm.

[34:20] Justin: Good work, Ian. I'll jump into the next one. It's similar to what Ian built: I replaced 8-plus tools with one system I built in-house to run our operations. We build agents for other companies, so there's email, project management, Asana, time tracking, billing, Google Docs, client reports, all disconnected. This was built to scratch our own itch. When you arrive, you sort by clients. We're looking at Warm Timber, and every request lands here in one centralized view, whether it came over phone, email, or Slack. I left an internal note to Danny. And this right here is emailing the client, trained on my voice from tens of thousands of emails I've written.

[37:00] Justin: I click draft AI reply and it takes all the context from the thread: due date, client, internal comms. "Matt, quick update, the reporting agent is built and ready to walk through." I can respond right there. There are also status updates sent to the client over email. We track time, important for billing and paying our team. And if we click Reports, we create a new report, select the client, this month, generate. This used to take me hours every week. It does it in under a minute: the summary of every thread, its status, touchpoints, ticket number, time spent, what's completed, total time. I send it to the client with their invoice in one click.

[39:10] Justin: The big idea: I'm challenging you to be the architect. Whether it's a Figma board, a Miro board, Ian's LMS, this is about taking your workflow and custom-tailoring it to you. Most software is a house someone else designed. You move in and live with whatever you have, maybe no washer-dryer, maybe no deck. This is building your own, or being the architect, to design the exact house you want to live in. There are probably 20 to 50 agents inside Mission Control that save us time and make it look more professional to clients, all running in one unified spot.

[40:30] Ian: That's amazing, Justin. 20 to 50 agents sounds daunting, but you didn't start there. You went iteratively. You asked, how can I do this, and just started tasking the agents. That's the core thing: I'm having a hard time doing this, you ask your agent, and it builds that thing. So it can look daunting, but just start with square one.

[41:45] Alec: A hundred percent. The key word is consolidation. In the past this would've been 10 tools that weren't customized for you. We all share the pain of clicking between fifty tools every day. When you have one place to work out of, it eliminates that friction.

[42:25] Alec: I'll give us something more basic. This is not 25 agents, more like one. I pulled this screenshot from right before we jumped on: the last 7 days with a client, all cold emails. We started the sequence for about 1,700 people, probably 3,500 emails a week, and we've gotten 112 opportunities. This is Instantly, the cold email software. These reply and opportunity rates without AI are impossible. Typically 3% reply rate is good, and half of those might be telling you to stop. What the workflow does: our client has a database of prospects, we scrape it, pull in all the info, and AI writes hyper-personalized emails showing we did our research.

[44:50] Alec: A very practical way to get into building agents: one tool we've talked about on AI4NTP is Apify. Data is often the bottleneck of AI, and Apify has off-the-shelf scrapers, a whole store. You can scrape Facebook pages, TikTok accounts, Google search results. One of my businesses sells to attorneys, so I found a lawyer and attorney lead scraper. You give the agent the scraper ID and an API key and it pulls in data. I'll pull up Codex. The agent enriches the data, goes on the website and pulls stats, like "I saw you've been around since 1997," an extra layer of personalization. Then it pushes into Instantly. It's a simple agent: pull in data, write personalized emails, push into a cold email sequence, get meetings on autopilot. It just runs.

[48:30] Ian: Kudos to everyone who stuck around, because this is the big daddy, the one I'm most excited about. BrandSauce, my company for about 8 years, grew about 5X in the last 6 months, from about 5 employees to around 30. It's completely remote, from Ecuador to Ukraine to the Philippines and all over the US. I call my bot Saucy, SaucyBot. My employees all said they wanted a Saucy too. So I built an interface that lets me create a new bot for each employee and walks them through the process. I'm the Sauce Boss. It initializes each bot, sends it to them, and walks them through: download Telegram, find the BotFather, get a token. Then they get a pairing token, it's all secure, and their bot initializes.

[50:40] Ian: Frank, our front-end coordinator, built a big spreadsheet of all our clients and onboarding tasks. Anna Belin and Gabby do finance and have been running data analysis on our finances. Each employee has their own use case. Everyone has read-only access to the company but can only build in their own folder, so they can't mess up everything else but can build like crazy. Viviana, a store builder, uses it to set up new product catalogs. As each employee uses their bot, I put it all into the Sauce Boss, which grabs all the learnings, memories, and challenges and puts them into this big brain.

[52:40] Ian: So new employees can come and say, I ran into this problem, how do I do this? It pulls from this hive mind: all our learnings, transcripts, catalog, everything our employees have done. If there's no answer, they ask a partner, and it gets added to the brain. Daniel is getting better English but isn't fluent, so he built a translator: on one page, as I talk it types it in Spanish, and when he talks it types it in English. I've been inspiring them to think what's possible. I always try to find the boundaries with agentic AI, and I've not found many.

[54:15] Alec: Ian, this is incredible. This is what real innovation looks like. Everyone's saying we're trying to use AI but don't know how. This is AI deployed in a real corporate setting: real people, now AI-enabled, 10x-ing their output. Important to note, Jensen Huang from NVIDIA said on stage that every CEO needs an OpenClaw strategy, and pretty much no CEOs do. Many don't even know what it is. Ian has fully built one out, deployed and operational inside an organization. OpenClaw is the opposite of conversational AI: Ian's OpenClaw is building out and setting up other OpenClaws and software for them.

[56:00] Ian: Thank you. There's a lot of fear around OpenClaw, that it'll wreck everything and delete your database. But if you set it up right, with someone who knows OpenClaw, they can't mess it up. It's impossible. And I always tell them, when they ask, can it make me a PDF of all these products, I say, ask Saucy. And sure enough, it built a beautiful PDF for the client.

[57:45] Justin: Very well done. We collaborate on BrandSauce, so you've onboarded me to your Saucy bot, one of the most streamlined processes I've gone through to set up an OpenClaw. It can be technical and scary with security concerns, so I felt like I was in good hands. I hope you enjoyed the seven agents. This is what we do for a living, building custom business solutions. We have slots opening up. I'll drop my calendar link, and you'll work directly with me, Ian, and Alec to build custom agents like the ones you saw, tailored to a challenge you're experiencing. We can do a 15-minute call to see if we're a good fit. Bring your best challenges.

[58:45] Justin: We've gone 10 minutes over, but we've got a lot of you still here, so drop your question in the chat, raise your hand, or use the Q&A tab. Ian, Alec, stick around for a few minutes for questions.

[59:00] Ronnie Lubwama: Sorry, folks, too much background noise, but I had my hand up. Not at all, thanks guys. I appreciate everything you're rolling out. I appreciate it big time.

[1:00:00] Alec: I'll ask one question of anyone still in here. What agent resonated with you the most, or was most applicable to what you're doing?

[1:00:20] Justin: That's a tough one to answer, they're all good, and that LMS is so cool, Ian. What I loved about Alec's cold-email agent is it's so specific. It knows what it is and swims in that lane. Sometimes you see robust AI builds that don't have an identity, they do everything okay but nothing great. I love how specific you built that solution.

[1:01:45] Justin: If someone wants an OpenClaw for their organization, what would that look like?

[1:02:00] Ian: It would really be talking to whoever runs the company, setting them up as manager, a similar setup to SaucyBot, where they create the bots and walk their employees through it and some training. The hardest part is showing what's possible. That's why I do weekly calls with all my employees: what's possible, how do I use this, what's it good for. It's like Cesar Millan training dogs, he actually ended up training the human, because the dog was fine, it was how the human responded. That's how I see this: you've got a bot, and I need to train you on how to use it. I'm not going to train the bot, I'm going to train you on how to work with it.

[1:03:05] Justin: Well, guys, that's a wrap. I'll give you a call later, we'll debrief and plan the next one. Thanks, everybody.

This is a lightly condensed version for skimming. Watch the replay for everything, verbatim.