Stop Overcomplicating Hermes Agent... Do This Instead!

Today we're going to talk about how to use Hermes Agent and structure your AI agent so you can actually automate anything that you want, but you don't overcomplicate it. And this is based on a question I had recently inside the AI Profit Boardroom, so we're going to get straight into answering that. So here's the question that we had. It was a question about Hermes and setting it up for a business. And basically, the question from the AI Profit Boardroom, which is a community where I answer questions like this every day, says, I want one AI employee that is a manager for one department of my business and will be overseeing the work of other AI employees in that same department. And this is for a property management business. So they're going to have this structure repeated for different departments once they have one department dialed in. Now, they've created the profiles and given them their sole MD, user MD, memory MD, and set up Honjo for memory and context. And they have ChatGPT codecs as the model, right? So basically, ChatGPT is a brain and the personality, which is the sole MD of Hermes has been set up, user MD, memory MD, and everything like this. And this is interesting to note because there's some details here, which I'll come on to in a second, which already will be super helpful for anyone using Hermes Agent. So I see this as a common problem inside the group. So after setting up, they gave the first task to a manager that includes to do some research, to figure out some stuff in order to make a decision on a tenant. In just that, it came back with an error that said something like the provider doesn't have enough tokens, essentially. Right? So it sounds like basically what's happened here is they've set up Hermes Agent to do this. This is where the problem is. They've set up Hermes Agent to do some research, but they've had some issues with, you know, the setup of the tenant and also the model and everything else. And they said, you know, at this point, I'm pretty scared. You know, I've spent about three nights, 15 hours selling it all up through trial and error. Really wanted to get it set up. Instead of this putting this in a list of apps that were promised but didn't work. This is something I see all the time. By the way, peeps, if you're watching this and you've probably had some ideas but never implemented it, this is super common. So, you know, don't worry about that at all. Now, you can see here that one of the biggest problem here is that they've spent, you know, a lot of time over the last couple of years just to set up systems like this. They've used ChatGPT, Agent Mode, ChatGPT, Claude, Claude Cowork, Coe High Level AI Studio, and now Hermes. And the problem is they all take you down a rabbit hole where it feels like something is happening. But after spending a lot of time, nights, hours, what you get is the same thing you're able to build with ChatGPT 1.0. And hopefully that's not the case after you watch this and you see this tutorial. So let's get straight into this and figure out what the problem is here. Right. So here's one thing that I will say, and I'll put some notes down here just to help you as much as I can. Right. And this is applicable for anyone watching. So the first thing that I see in this whole system is that it may be a little bit overcomplicated. So I know, for example, you feel like you've gone round and round in circles and you've gone down a bit of a rabbit hole, spending a lot of time trying to build stuff. I would say at this point, it's probably too complicated than it needs to be. So I would try and simplify as much as you can, if possible. And the reason I say that is you probably don't need like a full department with an AI agent manager over top. You can build it. We've built something similar with, for example, Paperclip. So it is possible to build. But do you need it? That is the question. And I would say, depending on what you automate, what you want to build, I would say it's probably not relevant. So what I see a lot of people do with AI in general is they're at point A and they want to get to point B. But what most entrepreneurs do, most business owners do, is they go like this and they try and make things like too complicated and too complex than it needs to be. And they end up over here. And so they do get there eventually, but it becomes way more complicated. What I'm going to show you in this system, and this is how I basically build and automate anything with Hermes, is you just simplify as much as you can and you go from point A to point B in the fastest way that I know that is possible. So how would I approach this whole system? Here's the three-step system that I would do. So the way that I would approach this is that you don't try and build anything too complex and you just look at, okay, what do you need to automate? What is the specific task that you need to automate? And then you're going to go into your AI agents and you're going to automate that. And this should not take you more than one hour. So, for example, if I want to, let's say, you know, and I know you've mentioned, for example, you wanted to do this and then that and that. What I would actually do is I would go into, for example, your AI agent. And it seems like you just need an AI agent researcher, right? That seems like all you need to do. So what I would do is I would go into ChatGPT and for whatever I try and automate, I just type what I want to automate inside ChatGPT. You can do the same inside Hermes if you've got it set up properly, but it doesn't sound like you've got it configured properly because the model actually said they're tokens. Or you could do this inside Claude as well, right? For you, honestly, I would recommend, I wouldn't even recommend using Hermes at this point because it sounds like the main brain of the model has not been configured properly. And so if you're struggling with that, you're probably not going to build everything else that you want that's even more complex than that. Like just adding a model to Hermes should be pretty simple. But if you can't do it, which is okay, then I would simplify it even more and just go straight to Codex or Claude. So let's say, for example, I want to automate something in ChatGPT. I just describe what I want, then I build it out. So let me show an example of that. And this is as simple as I can possibly make it. So, for example, when I was trying to build out a web design tool, which you can see over here, I just describe what I want. So that's stage number one. Stage number one, you want to go into Codex and just describe exactly what you want. Make it very clear and make it very tangible. So don't say, for example, you need a manager and you need a team and you need a department to do all this sort of stuff. Just describe the actual workflow that you need to automate. So, for example, that could be research for tenants like you've talked about. You might say, oh, but I've got loads of workflows that I need to automate and anything else. I would just focus on one. Like don't complicate it anymore than that. Just focus on one workflow at a time. So, for example, if I wanted to build a web design tool, which I did yesterday, you can see that I gave it the prompt here. Now, I actually built that into my agent OS, but you could have a standalone tool. You could have just a website that does that. You could have, for example, a conversation in chat, Chippity, that does that. And then you can see that we got the web design OS built over here. Now, stage number two is that you need to test it. Test it. So step one is build the MVP. Step number two is you test it. So you test whether the automation you built in stage one actually works. So, for example, when I built the memory galaxy inside Obsidian, it was pretty messy. It didn't really work the way that I want. And the design wasn't that nice. And so I just give feedback back to Codex or whichever AI agent I'm using, and I tell it to fix that specific problem. And so, for example, here you can say, like with the web design tool that I gave an example of earlier, I said, can you actually make sure the design OS works? And put some examples inside the design library. I tested it out. It doesn't work. And then it tested out. And then I gave it some more feedback. And then I gave it more feedback. Went back and forth in the space of probably about, I would say, 30 minutes until it eventually worked. So just keep testing it, testing it, testing it, right? That is step number two. Iterate and improve. Now, you've built the workflow for whatever you want to automate. That is step number one. Step number two is you test it, as we talked about before. And the final step is to iterate and improve. So over the space of the next seven days, after you've built this workflow, you're going to see things that you want to change. You're going to see things that you want to improve. You're going to see new opportunities to make this tool even better. You keep building those out into Codex, and you keep making it better and better. And once you've done that after seven days, once you're happy with it, then you move on to the next workflow. But you don't do anything until that point. And that's basically it. It's a very simple process. And that simplifies things way more than trying to build like a department with a manager, with a Hermes agent, and then, you know, going round and round in circles until you don't really get anywhere. It's better to go from point A to point B because that will take you. I mean, this system here is a lot of stress, and it might take you like 15 hours, right? This system over here from point A to point B, that probably takes you about one hour. Very simple automation. And then if you do one of those per week, well, that's 50 automations per year, and you absolutely fly them. How do I know that works? Because this is what I've built inside our agent operating system, right? We have the ideas pipeline, and then we built out, for example, Hermes inside there. And then we added custom automations into Hermes itself. Now, some point is to avoid frustration with AI, because I think this is very important, particularly when I look at things from your perspective. So, first of all, what you don't want to do is try and build multiple automations at a time. That will just be distracting and confusing. Also, only focus on the simplest version of whatever you're trying to build. So, again, don't try and build like a full department of AI agents with one employee and then another employee and then honcho for the memory and context and, you know, messing around with SolMD and UserMD and MemoryMD files. Like, that's going to get very complicated very quickly. Instead, focus on what is the simplest version of whatever you're trying to build. And then also don't switch tools again. So, for example, if you use Codex to build out this whole system that I've talked about today, then all you need to do is stick to that. Don't switch to Claude and then Hermes and then OpenClaw and then the Aging OS. Just focus on one thing at a time. Keep it simple. So, those are some of the things I'd recommend when it comes to this whole system. And the reason that I say this is important for anyone watching is like I see this happening over and over again with people where they just overcomplicate things. And there's nothing wrong with it. Like, it's just it's human nature. It's how we all operate. But this is why I built out a three-step framework where you can make it as simple as possible and you can build out whatever you want. And that's basically where I started with this whole tool. You know, if you look at the agent operating system, it's beautiful. It actually works. I use it every single day. I don't feel frustrated or stressed about building it. There's loads of cool things inside there. And I've just built it like one step at a time, one stage at a time. I tested everything, iterated everything, and then shared it with a community inside the AI Profit Boardroom. So, that's the way that I would look at it. And, you know, if you want me to answer your questions, this was a fantastic question. That's why I created a tutorial on it. But if you want me to answer your questions like I've done today, feel free to check out the AR Profit Boardroom. Link in the comment subscription or go to the AI Profit Boardroom.com. This is where I answer tutorials like this and create custom tutorials for people to help people as much as I can. You can see all the awesome reviews right here and loads of people winning with this stuff. Inside the community, you can post questions. You can also direct message me if you need any help too. And then also inside the classroom, you can get access to all my best training. So, if you want that agent OS system, you can grab that over here. If you want my new daily updates, you can get these over here, right? If, for example, you want to learn how to go from like complete beginner to expert with AI automation and simplify as much as you can, you can get that over here. The whole goal for everyone inside the community is to get more leads, to get more customers, to grow their business and to save time. Also inside the calendar, you can jump a weekly coaching course, ask questions, get help and support in real time. And inside the map, you can meet people in your local area who are building out with AI agents like you. So, feel free to get this link in the comment subscription or just go to the AI Profit Boardroom.com. Thanks for watching.