Leveraging AI

319 | Grok Bot is the most effective out-of-the-box tool ever with Isar Meitis

• Isar Meitis • Season 1 • Episode 319

Share episode

Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.

0:00 | 22:47

What if you could build a working business app—and an entire team of AI agents—in minutes, just by explaining what you want in plain English?

That’s the promise of GrokBot. And after putting it to work on two very different real-world problems, the combination of simplicity, autonomy, and always-on AI agents is hard to ignore.

In this episode of Leveraging AI, Isar Meitis puts GrokBot through its paces. He builds a personalized travel-management app that checks his calendar and email for upcoming trips, identifies missing bookings, and helps research travel options. Then he goes considerably further: creating a multi-agent AI team to research, develop, and execute a marketing campaign.

No elaborate prompt engineering. No dedicated Mac Mini. No complicated agent orchestration setup.

But there are important catches—including cost, security questions, and the quality of the initial outputs.

In this session, you'll discover:

  • Why always-on AI agents could become an important tool for business leaders
  • How GrokBot compares with the experience of running tools such as OpenClaw
  • How I created a personalized travel-management application in less than five minutes
  • How multiple AI agents can collaborate on research, strategy, content creation, publishing, and performance
  • How an autonomous AI marketing team identified target audiences and developed campaign assets
  • Why creating custom apps through plain English could challenge the traditional App Store model
  • Why GrokBot’s initial output quality still leaves room for human guidance and refinement

If you’re a business leader wondering what comes after AI chatbots, this episode offers a practical glimpse at a world where you don’t just chat with AI - you give it a goal and let a team of agents get to work

About Leveraging AI

If you’ve enjoyed or benefited from some of the insights of this episode, leave us a five-star review on your favorite podcast platform, and let us know what you learned, found helpful, or liked most about this show!

GMT20260817-211731_Recording_avo_1280x720

Hello, and welcome to the Leveraging AI podcast, the podcast that shares practical, ethical ways to leverage AI to improve efficiency, grow your business, and advance your career. This is Isar Meitis, your host, and I have an awesome episode for you today. Today, I'm going to show you and teach you how to use GrokBot. And why the hell does this matter? That's the first thing we're going to talk about, so let's get started. Before I talk about GroqBot, I wanna give you a little bit of background, and the background is why I think GroqBot is a big deal. So when I use Claude Cowork and Claude Code every single day. I use ChatGPT Work and ChatGPT Codex almost every single day for my clients. And when OpenClaw came out earlier this year, and everybody was excited about it, I was one of those people. I was really fascinated by an always-on agent that can do multiple cool things and connect to anything. The first thing I did in order to run OpenClau instead of going and buying a Mac Mini, which is what a lot of people did, I was trying to think about how do I make this the most effective from my perspective, lowering the risk, maximizing my availability, and not spending money on a new Mac Mini. And my solution was to spin up a virtual machine in Google Cloud. So that sounds really complicated, but it's actually not that hard. So what the hell is a virtual machine? A virtual machine is a computer just like your own computer, only instead of having a physical thing in your house, like a laptop or a desktop or a Mac Mini, it lives inside a cloud. It has its own compute, it has a browser, it has whatever you wanna install on it, just like a regular computer, only without a physical body. It lives in a server somewhere where you don't really care. Why did I want to do this instead of running it on my own machine? First of all, from my own machine, I was terrified. It's a brand new kind of technology. It's an always-on agent. I don't want it having access to everything that I have access to on the machine I run all of my businesses through. Second is access, right? If I want to operate it, add stuff, install stuff, remove things from its capabilities, I need access to my computer, versus if it's a virtual machine, I can access it from anywhere in the world. So from a flexibility perspective, it is fantastic. And the third reason was I didn't have to spend money on a Mac Mini or some kind of other hardware. So you pay monthly for a virtual machine, depending on how strong the machine you wanna buy from a compute, from a storage perspective, but it was a few dollars a month versus buying a Mac Mini. So I could have paid for my virtual environment for a very long time, while still enjoying more benefits than actually having a Mac Mini or any other kind of hardware. Now, the problem was, it was a pain in the ass to set everything up. It wasn't an easy process to set up the virtual machine. There was issues with the browser and how it connected, and connecting everything took me a while. And even once I had it running OpenClaw itself requires you to be really technical and figure things out and how to connect them and so on. And I'm a technical person, and there's a lot of YouTube and AI to help you through it, but it was still not a straightforward thing for me to use. The outcome of that was I played with my OpenClaw agent for a while, and then I deserted it and went back to using Claude Cowork and all the environment that I built through there. Still, the concept of having a chat interface like Telegram, which is the way I controlled my OpenClaw environment, so having like a simple messaging interface, but having access to really powerful always-on agents is something that I was looking for to do. So if you want, my short list of the ultimate tool would have been always-on bot that I can just give it a message or will run on a schedule or will do things based on whatever triggers that knows how to run on its own while communicating with multiple other agents without me having to figure out how to set them up, a virtual machine, so I don't have to have a physical machine to run this on, and the biggest thing, not having a lot of technical setup. This is exactly what GroqBot is. It is magical from all of these perspectives. So you go from knowing nothing about it to having it installed and doing really cool things for you in a few seconds once the installation is over, and a couple of minutes if you're actually waiting for the installation So what I wanna show you is two examples that will prove to you or will demonstrate to you why I'm so excited, why going from zero setup to immediately helpful while being always on and connected and being able to communicate through a messaging interface is so, so powerful. And I'm gonna show you two very different use cases that do very different things. One of them is a single bot that helps me manage my travel through an application that it has created, and the whole thing took about four prompts. So going from nothing to a full working solution in four sentences without any fancy prompt engineering, without downloading any existing plugins, without any of that. And then the other one is going to be a complete redesign of my marketing campaign for the multi-agent orchestration course, and including who is the ultimate target audience, what messaging will resonate with them, creating the actual emails and social media posts, et cetera, et cetera, all with just a few simple sentences in English without nothing other than that. So what I'm going to do, I'm going to share my screen, but those of you who are not watching the screen and just listening to the podcast, don't worry about this because there's very little to see. It's just a simple messaging interface. It looks like Telegram or WhatsApp or any other chat platform you've ever been to so on the left side, you have all the chats that you can have, and they have names, and you can rename them in whatever names you want. But otherwise, will give them names that make sense. You can see Creator, Campaign Lead, Travel research, Performance, Publisher, and so on. Again, you can rename them however you wish. And you-- if you click on any of them, then you can communicate with that particular bot. The first one I created was the travel planner. Now, a real problem that I really have is that I travel a lot, probably way too much, which means I consistently am terrified of missing something that I'm supposed to book and I didn't book, either an entire trip or a component of a trip. So I book the flight, I didn't book the hotel, and so on, because it happens so frequently. I started working on a project like this with Claude Cowork a while back, and I never finished it, without any really good reason other than the fact I had more urgent things to do. But I thought this would be a great thing to try right now because I have two months of really tight schedule of travel. So all I did is I installed GroqBot out of the box, and I told it what I want to do And what I told it was a very simple thing. I told it basically what I told you. I said, "Listen, I travel a lot. There's a lot of emails and calendar things on my calendar that shows when I'm supposed to travel. Some of them is already defined and set up, meaning a client has sent me a signed agreement and have made a payment, so I actually need to go there. Some of them are still in discussion and not approved, but I need them to be on my calendar, be ready for those. I want you to look through my calendar, and I want you to look through all my emails, and I want you to look at the meeting summaries from Fathom, which is the transcript tool that I have. And I want you to find all of those, and I want you to help me make sense in them. Show me in each and every one of them what was booked and what wasn't booked already." And that's basically it. And it said, "Hey, Isar, I'll pull your travel out of the email and calendar, then show each trip with what's booked and what's still missing, flights, hotel, ground, starting with Google Workspace since that's where signals live, checking what's already connected." And then it popped up a pop-up for me to approve connecting my Google Calendar and my Gmail, and it looks just like any time you connect to Google Calendar, so you give it whatever permissions you want on the Google side. Since I didn't want it to have access to everything, I said, "You just have read access." It couldn't change anything. It couldn't manipulate anything. And again, this is coming from the Google side, so I don't think there's a workaround for that, so I'm safe, right? It's just gonna read the information. And then it went ahead and did all the different steps, and in about a few minutes, I don't remember exactly how, but very few minutes, it came up with a list written in simple English saying the destination and the time and the dates, whether it's approved or not, and what things did I do or didn't do as far as booking travel. In a few seconds of my effort, just by explaining what I want and by allowing it to connect from a read-only perspective, I get a list of the things I need to pay attention to as far as the travel I need. Then I ask it for some additional information on one of the trips. It added that additional information. It was missing another trip, so I ask it to check that as well. It found that other trip, and then I had all the different components that I needed to have. But that was in text. That is not very user-friendly, even though it achieves the purpose. So I asked in simple English, "Are you able to create graphics interfaces or just text?" Meaning, can you create an application that will show this in a nicer way? And then it said, "I can. This chat is text, but I can build visual aboard a small web app and keep it updated from mail and calendar, making it right now with your current trips." So I didn't define how I want it to look like. It didn't ask me anything. It just spun up this application and gave me an index HTML file because it's just an HTML page that I had to download in order to see. And I ask it, "Is there a way for you to allow me to engage with it, uh, or do I need to deploy it through something like Vercel?" Because this is just a static webpage on my computer. And then it said, "You don't need Vercel to click around. The file, uh, is going to work. There's two easy ways to do this. One is open the HTML, the other is, uh, use something like Vercel to deploy it." But then it told me something more interesting, which is, as I mentioned, it has its own virtual machine. On its virtual machine, on its browser, it's always live and connected. I don't need to download, I don't need to do anything, and it works as if I would have deployed it to a Vercel or any other deployment platform. So it saved me all these steps. All I have to do Is open its own virtual machine, which is on their top right corner all the time. And once I click Open, it opens that, and those of you who are watching the screen can see an application that shows me each and every one of my trips. And for each and every one of my trips, it shows me what is booked and what is not booked, and you can see it over here in color coding that looks exactly just like my website. How does it know to make it look just like my website? Because I told it. It created a first version that was pretty ugly and with bad color choice, and all I told it is said, "Visit multiply.ai, learn the brand guidelines and apply to the mini application you created." And that's it. I had a working application that looks like my brand, that shows me all my trips for each and every one, whether it's approved or not approved, whether travel is booked for flights, hotel, and ground. It shows me what's missing. And if I click on anything that is missing, and that's I added afterwards, I literally ask it, "Okay, if I'm missing flights, I wanna be able to click on that, and you're going to do your own research and show me what flights I should book. Show me the top few options and explain the pros and cons, so I can go ahead and book the things that I'm missing." So if I'm clicking on Flights right now on one of those, you can see that it is showing, uh, the best option, and it explains exactly why. It is the cheapest one for nonstop to the destination. You've flown with them in the past, blah, blah, blah, blah, blah. Gives me the, the solution, but it shows me other options as well and their pricing and where they are and so on, so I don't have to pick the one it gave me. And there's also a Refresh button that I can click and see it currently because if I ran this yesterday, then probably the flights have changed. All of this going from having nothing to having an application that shows me all the trips that I'm supposed to do, which ones are finalized, which ones are not, which travel I booked and which I didn't book, and showing me the best option to book the flights or hotels or ground transportation I need in less than five minutes by explaining in simple English what I wanted to build Again, I use Codex almost every day. I use Claude Code and Claude Cowork every single day, and it would have been a much bigger effort to do it over there. Totally doable, not a big deal, but the simplicity of this is absolutely mind-blowing. Now, those of you who have been listening to this podcast for a while, a long, long, long time ago on this podcast, I said that I think the concept of the App Store is going to go away. And the reason I said that is because I assumed something exactly like this will happen. Why will I ever download an application from the web if in just a few minutes, in a few simple sentences in English, I can create the application I need, exactly the features I need, in exactly the display that I want that will work for me in the most effective way? Now, I'm not saying more complex apps and whatever are not worthwhile, but I'm saying this is where it's going. If you can spin up exactly what you need without any extra features, without wasting memory, without doing all these things, in literally explaining in simple English what you want, the concept of the App Store doesn't make any sense. So this is example number one. And then I said, okay, let's try something more complex. Let's try figuring out how multi-agents run together. So all I did is the same exact thing. I explained to it what I want to do I told it my company, and then I gave it the link to the website, sells AI training courses and has been in the past three-plus years. We are currently focus- focusing on the multi-agent orchestration course, to the public and workshops to private companies. Would like to improve the marketing and drive more people to the public courses. I would like you to come up with suggestions for the different agents we need in order to, research, analyze, create deployment, monitoring, and overall execution of a broad marketing campaign, both organic across multiple channels that can grow in a highly profitable way the courses. I'm open to any suggestions. I think we need to start with research, but I want you to suggest what kind of group of agents we would need, blah, blah, blah, blah, blah, right? Very, very simple, but I'm voice typing the whole thing, so there's a lot of not very clear sentences, uh, and some things it didn't pick up correctly, and yet it understood exactly what I wanted, and it suggested creating the following agents, a research agent, a campaign lead, a creator, a publisher, and a performance agent. With the campaign lead is the one that's gonna run all the other agents. And all it asked me is that sounds... If it makes sense. I said yes, and it created all the agents. Those of you who are watching the screen can see it on the left side. Now I have the campaign lead, I have the research, I have the creator, and so on. And then when I went to the campaign lead to ask it what I need to do, I started the way I assumed I need to start. I said, "Hey, what I need to do now to get started?" It said, "No, no, I'm already working on it. Uh, I'm on it." And I'm like, "Okay, this is really cool." So it created all the agents, and the agents immediately started working. They didn't wait for me to tell them anything to do. So the campaign lead started with the research agent. It gave it multiple tasks. The research agent came back, and then the campaign lead used that to trigger the creator to create specific emails and so on. So I only need to communicate with the campaign lead. And you can see I'm asking it questions, and I'm giving it dates, and it actually found conflicts on my calendar because it's already connected to the calendar because of the other projects, uh, that I did. And we moved the dates, and it immediately gave me all the different components that I need based on the research, and then it started creating the actual campaign components and content that I need. So again, I went from having nothing to having four agents. One of them is the lead that manages everything that is autonomously developing, designing, and creating the ideas. Now, the way it did this, it actually went and researched who signed up for my courses so far. How does that align with existing communities I have access to? It found that there's a big alignment with two different communities I'm already working with from the past. It said, "These are the most likely people to buy from you because a lot of people from these communities bought from you before." So some of you are yours, people who are listening to the podcast. Some of it are people from my community that meets every Friday. By the way, you're more than welcome to join that. We call it the Friday AI Hangouts. There's an amazing, amazing group of people that meets every Friday at 1:00 PM Eastern, and we just talk about AI, and people are sharing what they build, people are sharing issues they're struggling with, and other people are helping them solve it. Really amazing, fantastic group of people, uh, and a community that, again, you're all welcome to join. So makes sense that these two communities who know me for a long time will be the people who are most likely to take the courses, and it's built a campaign for that, plus a campaign for people who took previous courses that I was teaching in the past three years. It figured out the perfect target audience. It figured out the messaging. It figured out the emails, the newsletters, the post on social media, all of that on its own, and it did all of that without me having to do anything other than explain to it what is it that I want to achieve as a final goal. This is mind-blowing. Now again, can I do all of this with other tools? 100%. I can do this with more or less any AI tool out there. But the simplicity, the ease of use, the going from I have nothing to I have a solid outcome is mind-blowing. No other tool that I've ever used, AI or other, provided so much value to me with such little effort. So quick summary on the good side and then a few potential bad things, and then we're gonna sum it up. First of all, it has an incredible user experience. As I mentioned, you don't need to know anything. You need to install it and then just tell it what you want it to do. And even if you don't know how to set it up, there is no setup. You just tell it what you want it to do, and it will spin up the agents which become chats on the side, and you chat with the main ones you need to chat with, and it will spin up everything else that it needs. If it needs access to something, it will ask your permission, and you can decide whether you wanna give it permission or not. It will work from any machine because it's run on its own virtual cloud, so you don't need a dedicated machine or put a risk of your own hardware getting access to anything and so on. It runs on its own virtual machine that you don't need to care about where it's running. It's somewhere in Groq's, or if you want the company name, somewhere in SpaceX AI's, uh, servers somewhere. Uh, it, as I mentioned, achieves the most with the least amount of effort compared to any other AI tool that I know, and it supports all the standards you know from the AI universe. So you can import skills, you can import plugins, you can use connectors, hooks, all the stuff that you know that works in other AI platforms also work here seamlessly, so you can increase the likelihood of this working the way you want it to work because you can bring the skills you're already using in other places, and then it will apply them wherever it needs to be applied. You can even choose which skills are available to which agents, so you don't have to give all of them access to all your skills or plugins and so on. Really, really cool. Same thing, by the way, with connectors. You can define which agents have access to which connectors, uh, and you can turn it on and off Depending on your needs and your warm and fuzzy on how much you trust all these things to have access to different components. So what are the downsides? Downside number one, the only way to get access to it is with premium plans. You cannot pay $10, $20, $30, $40 a month to get access to this. The smallest, lowest plan you need is Cursor Teams, or to be specific, Cursor Teams Premium, which is $120 a month to be able to use this tool. The next tier is $200, the next tier is $300 which is the only way to get access to it. In my eyes, no-brainer. Just the thing that I did, if one person sign up for the course, one person from this campaign that it created, it is going to pay for the entire year. So it makes perfect sense. Again, I don't know if you have the same kind of ROI, but if you do or if it helps you do anything that's productive or saves you enough hours, then it's worth doing. For me, the fact that I'm not gonna miss any flight, and I didn't so far, but I'm always terrified about it. So the fact I can sleep better at night knowing I'm not gonna miss any flights is already worth the money. Now, the other thing is, going back to the downside, is how secure is that virtual machine? I don't have a clue. Meaning whatever happens on that private computer that GroqBot is running on is running on that machine. You're giving it access to whatever you wanna give it access, or even if it's just what you wanna build, that's already giving it access to information about you and your company. Where is that stored? What security measures, uh, cover that? I don't know. I seriously doubt against-- u-unless that's gonna be very well defined, that any larger company, definitely enterprise, will say, "Oh yeah, we'll let, you know, Groq virtual machine manage everything that we're doing." So I think that will need to be proven. I'm not saying it is not secure. I'm saying I do not know, and I haven't seen any clear directions or, more importantly, third-party tests that prove how secure it is. And then the last thing is the quality of outputs. Uh, the agents themselves that I-- that it span up without additional information from me were okay. They were not great. That being said, I gave it very little instructions and very little cycles of improvement, which I would do and would have done with any AI tool that I use. And as I mentioned, the easiest solution is skills. I can use skills I've built in other places and give it to GroqBot to then be able to do things significantly better aligned with the way I want it to work because it's using the skills that I created for Claude and or ChatGPT. Bottom line, if you're willing to spend $120 for the first month, and you get a free test. By the way, I created the travel planner with the free test, and then I ran out of credits, which were supposed to last for a week and were about enough for about an hour. But that being said, you get a free test that you can try, but if you want to be able to run this with a slightly higher scale, I highly recommend spending the $120. It's a new really good implementation of an existing proven, really powerful concept that we all came-- or I don't know all, but a lot of people, geeks became, came to love with OpenClau, just without all the technical limitations and technical requirements of OpenClau. So it gives you the best of all worlds. In my eyes, worth the $120 once to test it out, see if there's use cases that you wanna use it for, and then you can decide after a month if you wanna continue it or shut it off. That's it for today. I'm personally really excited about this. Again, it's been up for less than a week. It's already providing me, really great value with almost no effort. So go check it out, write to me on LinkedIn, let me know what you think about this. I would love to learn what use cases are you trying, And we'll be back again this weekend with another news episode. And until then, have an amazing rest of your week.

Head Shot

Isar Meitis

Host