Welcome To Business Unscripted

Dave 0:02

Welcome to the Business Enscripted Podcast. We're here to share real life insights, practical strategies, and the honest lessons that we have learned from our own estate because frankly we have been in your shoes. So whether you need help with operations, accountability, financial knowledge, details, or maybe just getting your mindset right, we are here and it is the right spot for no fluff conversations and tools to help work. So grab your favorite cup of Joe. Let's jump into the show. Grab your favorite cup of Joe. I got mine right here. Let's jump into the show. It is another episode of the Business Unscripted podcast. Uh, we're here for episode 77. It's another Friday morning, and welcome bright and early or bright and late. Like Darren, which is why Dwarne doesn't have his favorite cup of Joe, because he has that already, or otherwise, he'd be up all night.

Duarne 0:48

I have plenty of today. So I'm on water right now. You're on water. Earlier, though, I think there was a single malt that I might have had.

Dave 0:55

So love it.

Duarne 0:56

Love it.

Dave 0:57

So if you're a business owner, aspiring business owner, and you are looking for some place to get some you know good community, learn some things, struggles, we share those here. We share things that we're going through in the past in our business, things we're facing now in our business, and also things that we've helped our clients with. You know, we probably have over 50 years of business experience between the two of us, uh, both in corporate and helping small business owners. So you're in the right place, and we look forward to bringing you here some information that will be helpful to you and your business and operations. If you have any questions at any time, feel free to drop them down below, and we will answer those live on the show as well. So, Darren, thanks for joining me, brother. Glad you you're well rested from last week and and you're uh hopefully you know, give us a rundown. What's uh what's a win or something that you had this week that you're proud of?

Give Your AI Team Clear Roles

Duarne 1:47

Well, it's been a very, very, very busy couple of weeks, but I'm not dwelling on it because there's lots of good things coming from it. The big thing for me is having some realization moments again, it's always fun. So you and I have been talking about AI agents, Hermes, putting together like full teams, etc., these last few months, and it's hard to get your head around that. When somebody says to you, you can hire every person you possibly want in your team to solve every problem you have, go ahead and do it. Your brain kind of goes a bit numb. And for me, it was a moment of oh bugger. Okay, what do I do? What do I do? And it happened to me recently because I built a fantastic new app development team, which is a real person team. I think I mentioned a couple of times. There's 11 people in the team, it's a great team. Now, I don't have an AI agent team that's that sophisticated, and I was asked myself, why is that? And decided to solve that problem this week. So that was my big accomplishment this week. And ironically, it involved me going back to a book that I read on the train on the way to work over a week about 20 years ago.

Dave 3:07

Like 20 years ago, yeah.

Duarne 3:09

I mean, it is nuts, right? It's a it's a book by the by the guys who created Basecamp. So if everyone's out there used Base Camp, great piece of software, it's a couple of decades old now. The guys built the uh platform development platform Ruby on Rails, which was what they built that in, and they built some really cool other you know systems as well over the years. And those guys released a book called Getting Real. And it was about how to take a product, an idea, a concept, and develop it into a product or a web app and then go to market with it. And back then, believe it or not, an e-commerce website was considered an app. It was it wasn't something you get to go and build quickly, it was a it was a bit of a task, it was a bit of a process. You didn't have WooCommerce back then, it wasn't like that. You had OpenCart, you had Big Commerce. Big Commerce back then was Interspire, and you'd pay like $1,500 to get a license to get the software downloaded to build it on top of that yourself. That's the sort of world I was living in back then. So that's when I first read it, and I had some really great revolutions, revelations back then when I did it. When I read it this time, I decided it was time to go ahead and create myself a framework around the book and build out an entire framework on the book. Now, once I had the framework, I was able to build out my team, I was able to build out all the positions I needed, the mentality, the way I want my team to work, the methodologies, all that really cool, complicated stuff that I hadn't really been able to lock down at this point for my virtual team because it's a little overwhelming when you start thinking about I can have as many team members as I want, and the cost doesn't really go up to do it, it's just a case of what do I need and how do I use them. And I managed to do that, and I was really happy. And then we had our meeting on Wednesday, and I talked to you about it, and you're like, that's cool. And I said, now I just need to figure out what I do for my admin team. And we had that revelation on the call where why not take the same concept and roll that over into an administration team? So I built that yesterday, and I'm in the process of implementing that, but that's a huge weight off my shoulders to know that I've managed to put something like that together. And for those listening, I mean, it's a little bit more complex than just a framework. It's like 50 documents to support this, which are being generated and created in order to make this happen. And lots of reading, lots of back and

Keep Knowledge Current With Notebooks

Duarne 6:00

forth.

Dave 6:00

But I'm pretty excited to see how this uh well, I think the the beauty of of that part, like the next layer, right? So the easy part is telling telling Hermes or telling whatever Claude to go, hey, I want you to go create these profiles or create these initial documents. But as we've said before, I think the ongoing biggest support that's going to be is then what's your process after that to then where it continues to learn and evolve. And there's many different ways that you can do it. One of the ways that I have found that has been beneficial. Obviously, we've talked about notebook LM. If you're not familiar with notebook from Google, which I believe now it's just notebook instead of notebook LM. So it's I think it's notebook.google.com. But you can create a virtual notebook of different resources, and it could be YouTube links, it could be videos, documents, whatever it is. You know, and so for me, what we're building on our back end of our team, right, is like each chief then has a notebook behind it that has reference materials. So go search, say it's an HR person, go find influencers in on the HR channel or HR blogs, anything where people are irregularly continuing to post, you put those into your notebook, and now your persona or your you know your chief level or whatever it is that you you know for your your org now uses that notebook to continually evolve and update and see what's relevant. And so now instead of it just being a static, like you'd mentioned like the initial document, instead of it being a static document, now it becomes an evolution and a and a learning individual as well. So that's the beauty of it. It's a living document, right?

Duarne 7:37

And and you're right. I mean, a living document with decision libraries included, just start changing the game. Right. These are things that I hadn't even considered post two weeks ago. And now that I'm playing with these, I'm starting to realize that these teams that I'm generating can be replicated, they can be used. And I know you and I have spoken about this a couple of times, is some of these agents are really heavy on resources, so running them on your own computer is certainly not recommended if you're gonna run your entire team. Maybe run an agent on there like Hermes, but just have it doing your own personal tasks and stuff that you want to reach out and ask it to do while you're doing other things. And I use mine to communicate with my other Hermes agents are running on different machines, so I've set up a little ecosystem. But I've been building out separate devices to run my clustered and departments. It's almost like running your own department and saying, that department, there's your device, you use that. And that's a different approach that I've taken. Time will tell if it's the right approach or if there's a better approach, but I know in our conversations, some resources that these teams can use are pretty significant. They will just chew up the memory and you'll sit there and realize other things installed and not happening properly because there's no memory on the device still.

Dave 9:37

Well, and and I and I think the for most people that are watching this, still they're like, I don't I don't want to have that. Yeah, I don't want I I want to have it. I don't want to know, I don't want to know, like, you know, I don't want to handle like when it breaks down, I don't want to handle how I continue to retrain it, I don't want to handle all that. And so I guess like the main concern, I think, what if you're just a general business person listening and you're like that all sounds great, but I don't want to do it. Like, I think that's the beauty of people like Jorn and I you can listen to us or maybe tinker around. It's funny because

Hardware And Connected Assistants

Dave 10:13

like I've had some conversations recently with people that just started tinkering and then they tinker more, and now they tinker more. And and it's you know, I would say don't don't shoot the messenger before you give it a shot, you know, like don't just be the negative Nancy. I think because we say it all the time, I say it all the time. Like nothing that I'm that nothing that we say or advice that we give or things that we do are not things that you couldn't do on your own. You can totally go out, you could do all this yourself, and basically what we have been providing is is we save time, right, anxiety, and stress. That's it. Like if you don't want to worry about it, you want to manage your clients, you want to, you know, kind of continue to focus on what you're doing and building, like that's perfect. Let somebody like us support you. Find somebody, you know, it doesn't have to be us. Like find somebody that you trust that you have a relationship with that's that's doing more than you and and can help you build it, and then build it. But also, like you don't have to jump to a full 85-person team either. You know, I think the biggest thing is like, you know, just figure out. We've talked about the two lists time and time again, so I won't say it here again. If you need to listen to your two lists, go back to a couple episodes. And I think I've said it the last four episodes in a row, find the two lists, or you can see it on my social media. But pick one or two things that are gonna save you an hour or two a week. Like those hours that hour to a week is 50 to 100 hours a year. You know, now you're getting back, you know, two to almost three weeks of time just by saving an hour or two or adding an extra hour right into your day.

Duarne 11:38

But you're 100% right. I mean, you and I have spoken about this a bunch of times, and this is how I've always approached having a virtual assistant because I would do a lot of outsource staffing solutions for clients abroad with staff here in the Philippines. And the first thing I would always say is like work out what you're doing every day that you don't really enjoy doing, and write them down and then allocate those to your virtual assistant, create a process and create and hand it off. You could do the same thing with your AI agent, create a couple of tasks. Now, we just went through Paige and I and had a conversation about all the things that we've been doing and we don't want to do and the things we want to do, and that's what I built my administration team around, and that's a 12-agent team. Now, the idea is not to have a 20-30-person team that includes a personal assistant for me, a personal assistant for her, and an orchestrator at the top who runs that team

Start Small And Buy Back Time

Duarne 12:33

and all the other little components that are needed. Now, that's really helpful because it means that that team is now shared between a couple of people who are doing running the same business or the same organization, but are doing very different tasks, but have overlapping interests.

unknown 12:51

Right.

Duarne 12:52

Because now you have a single source of truth with a single source of a team to support them and back them up.

Dave 12:58

But here's here's the cool part right that I that I think about that like when I'm like is that the agent isn't like like if you had one Nancy working in HR, right? Nancy can only talk to one person at a time, work out one project at a time. But you guys are trained up, right? A Nancy virtually, that now potentially can work on 17 different people's work at the same exact time because it's not like just one agent, and it's like, okay, that agent can only do this, and I can't send out anything else. It's like, no, every time that process starts, it's a brand new session and it's a brand new Nancy's memory, right? In terms of like everything she knows. And the other beauty of it is ultimately as we start thinking bigger, right? And I'm starting to think of like these bigger organizations and and and nonprofits are a good opportunity for this, uh, is you can have uh much more resources, right? Or different computers that have the personas loaded, and then you have like an obsidian or some sort of vault memory system that everything's routed back into. So then it's always pulling from the same information as the memory and the files and the you know the policies and procedures, whatever it is on the HR side. Like that's the beauty when you build it right, that it's not one person now, like you know, that one persona can now do 17 activities if it needs to at the same time. You're not limited to and in what hour that person that agent is working, and that's the beauty of the virtual side.

Duarne 14:26

Absolutely. And like when you start talking about Obsidian, there's open viking that just came out, which is a three-layer system designed specifically as an open source platform you can host locally. And the cool thing about it is it helps reduce your token usage because it tries to give the smallest amount of context initially on every request. And if it doesn't have enough context, it'll go back and ask for the next layer, and that'll work its way back from you know one line to a paragraph to an entire documentation folder that's set up with that information as needed.

Dave 14:58

And and I've and and then as you get bigger too, then there's like you know, vectorized and then write the graph, you know, graph routing to help agents learn more what does what, so it kind of helps navigate easier. So like but as a solo printer, you probably don't need that.

Duarne 15:14

So I mean, like I've got agents there, one of them's called Atlas on one of my platforms, and his job is to govern who gets access and who can be authorized to do what when it comes to secret information or company resources. So he'll make the decision based on the authorization requirements for that individual agent underneath him to decide whether it's suitable for him to do it. It's no different to giving permission. So, you know, if Nancy at the front, don't we doing reception, isn't allowed to look in your accounting package to see, you know, bank bank records or something, you can set up your agents with a similar sort of permission levels. They can't or they can't see certain bits of information. It's the same thing, and it's all about once you start chatting and consulting with the your platform, you'll start to realize what you can build out and how you can build it out. But it is seriously overwhelming, and it's not something you should rush into because my first two, I'm probably gonna end up rebuilding them within 30 days, I think, because my second two are so much better with the way I've structured them and built them out. I've learned so much, and it's not a waste because in my instance I use Hermie, so I can say, hey, can you go and create me a soul and a handoff document for each one of my agents? And then I can load that up into another system and say, hey, I've got these agents. How do I make them work and do something in this ecosystem? Where's their benefit? How do I utilize them? But I mean, you're right, it's super overwhelming for tinkerers and big thinkers like us. It's fun, it's overwhelming, it's a rabbit hole. For the average person, they just want one system they can log into. You know, something like your command center that you're building, right? Where it's got a bunch of agents built into it, it's got a bunch of centralized storage and it solves a bunch of problems, and you can just chat to it and set tasks off and run them. It's a bit that's a bit more digestible for a lot of people. And if you find yourself outgrowing something like that at some point and needing to add more on, that's when you start tinkering and you start testing. But you might have a lot more time to do that if you've already got something that's pre-configured ready to go for you, and you've already picked up that extra couple of hours a day in savings before you get to that point.

Dave 17:38

Yeah, and and and I think like so when when you're when you're listening to this, and honey, I love you, I love your text every time I'm on scroll. She always watches, and she like at the right moment sends me a text just to like get me off the track. So I love it. But what I would say is is again we are talking much more in depth uh and a lot of times because Warren and I love this shit like a lot, and we're into it. But to kind of simplify it, I think again, just think what are some things that you're doing on a regular basis, and those are the things to start with. But don't be afraid to ask questions. Like e every model has a learn mode. Like you can even say, like, hey, teach me how to do this, not just like go and do it, like you can teach me. I highly suggest as well, and and and Jordan, you can get into this a little bit, but because we talked about a pre-show, but there's always gonna be a human involved, right? Because of the fact that unless you create the perfect like model on the back ends that these teams have, you as a small business owner are always gonna have some sort of person involved, even even medium businesses are gonna want to have some person involved, you know. Because of you, like you just what how did you describe it in in the pre-show? I think you said you can tell it to go cut a sandwich, but there's a lot of different ways to cut a sandwich, right?

Duarne 19:06

Absolutely. Cut a sandwich in half, like and look at how many ways. We've all seen those videos where you know you give them a like a subway sub

Clear Instructions And Human Review

Duarne 19:13

and then cut it straight up the middle, like in subway. But what if they were to cut it lengthwise or across the middle? I mean, it's all we and then that's a joke related to prompts, but instructions can be like that. And I had the same instance with one of my team members just this week, where I'm like, I'm giving what I feel are in simple instructions. How much more granular do my instructions need to be to get the right result? Tell me what you feel would be the most appropriate way to give you this task so that now you've done it retrospectively, so that I could have reduced a lot of the friction going back and forth. And what I'm realizing is maybe I need to go back to explaining to AI what my task is and asking, is that clear enough for you? And see what it says.

Dave 20:03

Well, and it's it's it's so funny. Well, well, I guess multiple multiple things that came up to me right there, right? One was you know, we've talked in the past about how students in school are gonna have their own AI tutors. I think like next level, like when you think business owners, like your each employee could have its own directional communication team, right? So that it could train itself and give feedback. So imagine you give an instruction to your your your guy, right, or whatever you call it, your your sec chief of staff, right? To send it to Saturn, whatever, or whoever. Okay. But that routes through Saturn's own model, right, or own agent that has learned how he wants to get instructions to help him. So then it rewrites that command, knowing your knowing your knowledge, knowing your information and everything that is for your business.

Duarne 21:00

That's what I'm telling, saying multiple orchestrators within your organization.

Dave 21:04

That's the key that then gives it to Saturn, and now Saturn's got a direct communication. Like, how many, like I could even incorporate, I could imagine how many there was a lot of miscommunication from the telephone effect. Like this kind of eliminates it, you know what I mean? Absolutely. Um everybody can learn things.

Duarne 21:22

We've just done assessments with our teams, right? To gauge their skill set, current skill sets on certain topics and their roles. Part of it, what we did is we included a disk assessment. So we've got now personality assessment details there too. Part of that is so I can train AI models on how to communicate better with those individual staff members. So as my agent teams start to communicate directly and hand tasks off, I can make sure that they're actually using language which is perceived better by those individual staff members that I, my real staff that I have, like you know, the satins of the world and the like. Because there's going to be situations, and it's really when you start working with AI agents, you tend to find yourself getting very short with your messaging. You find yourself getting very direct because you can, because they're AI agents, right? We don't think of them as real people. When you have real people, you can't do that. You typically would adjust how you deliver work to them. And you would adjust it accordingly, just get the best results out of them. And with a lot of us working remotely now and having teams working remotely. Remotely and not having that face-to-face or the video call as much as we do, it's really helpful to have your agents trained up to help communicate that better. And I know you've got an instance where you were saying that you've got Slack set up with all your agents in it, and all your team jumps in and gets an asks for assistance. You know, I last week built a second brain which is trained on my Clifton strength results and a bunch of other stuff as well, like five years worth of decision library from my emails that it went and pulled together to work out how I communicate and what sort of decisions I make based on all of that, to basically be the first step. If you have a question for me, go and ask that first. Go and ask my chatbot first, you know, and see what it says. If you feel like that's the right direction, then go ahead with it. If you feel like it's something that you still want to raise to me, that's fine too. If the agent feels like it's something that should come to me, it'll raise it directly to me and make the suggestion. These are the sort of things that can help save you a little bit of time, but if you're willing to put that sort of work into the tools to generate them and train them and improve them. For the majority of the people out there, if they were to go and grab themselves Grokbot by you know and start using you know X's Grok bot, they're probably going to find they're going to pick up a huge amount of time back in the day just by using that one tool without having to go down the path of learning all these new tools and creating their own installations of Hermes and using their machine and everything like that. But be careful because what tends to happen is, like you said before, once you start down this path, you go, what else can I use this for? What else can I do? Your brain starts ticking and going, Oh, I've been wanting to do this for a while, but it's a little bit complicated and it's gonna take a bit of time to put together. I wonder if my AI could do this for me. I wonder if I could get my AI to work on this for me. And you'll start finding yourself tinkering and working on well let's I mean let's let's give a real real world example, right?

Dave 24:44

So for me, I have we have our system, and I have I have a team member, right, through

QuickBooks Imports And Client Questions

Dave 24:51

through you, who is part-time, but she helps, she's an accountant, and she helps kind of keep things going on a day-to-day basis. Before, it would be a lot of questions for me, right? She would send questions, hey, we need what what's this, you know. Can you can you request this information? Now, uh, since we've kind of installed and have the the the bot that's been trained and and and understands everybody's kind of financial situation from the notes and things that we've had from our sessions, now her first instance is to go and and and ask the bot and that and the bot and and Hermes, the chief financial officer of our department, helps her by actually going, researching the question is connected to QuickBooks for the clients that have QuickBooks uh through our command center and and can research the information to give it back to her to say, here's here's what you requested. Some of the other things that she has on is maybe sometimes there's a bank disconnect. And instead of her sitting there manually going through and having to type all the entries in for the end of the month so that we can keep things moving with the close, she goes to the Hermes and says, Hey, here's the statement. I need you to create me a QuickBooks import file and then go and import it into QuickBooks for me so that she can then go and reconcile the account. And in like 15 minutes, it's done. So those are kind of the real world situations that you could be doing, and we talk about all the time like how could you save hours? Like I think for her, like she's able to do in 20 hours what some people need 40 hours to do. And it saved me time by me not having the right to actually have those back and forth. And the next level for this, now that I've kind of seen how it's been working, is now instead of her saying, Hey, I need you to reach out to the client and ask what this is for, I now can trust Hermes, where she asks a question in the thread. Hermes will then say, Oh, okay, that needs to go to the X, Y, and Z client. Let me draft an email, drafts an email, and then it sends it, or puts it into my queue for me to review, and then I can hit approve, and then it sends it. However, I want. It's truly like what you feel your security level is there. But those are like real-world examples that can be done in in situations that just you probably before would have taken you know a couple hours back and forth, right? So and it's funny you can say that.

Duarne 27:15

It's we've been talking about this, and there's a big and this is where it gets really intriguing. One of the key points of that is your agents have a central source of truth, right? A source of truth, documentation, files, whatever, system access to be able to access all of what they need. One thing that I identified is because we have clients that send us work in, and we need to then we have to take the work out of their systems and put it into our systems so we can work on it, report on it, allocate time to it, and then copy that data back and put it back in their system for them to work on.

Dave 27:54

Right.

Duarne 27:54

So I asked my team to work out how long are we actually spending just on the moving the data back and forth. And turns out we're spending 20 minutes every ticket moving data back and forth between two people. Oh wow. Which was just insane. And it's about 20 to 30 tickets a week. So pretty quickly that adds up. So I'm like, okay, I built a middleware software package this week. It's in testing right now, with the entire intent to get all of that data into our one ticketing system that we use, which already feeds to our brain, which acts as our single source of truth. So it's actually now part of a searchable platform for my entire team to be able to get access to. Now they don't need access to the client system. The client system costs about $80 per user to get access to. So clearly, they don't want to be spending that for my entire team. They just apply one license for our team to get access. But due to multi-factor authentication, we can only have one person utilizing it. And this tool that we've just created allows us to act as an intermediary to pull that data across and dump it where we need it. And that's huge. So once you start identifying this, you have to realize that yeah, having these agents that can do all these things, can answer questions is great, but you need to be able to show where the information and give it access to where the data is in order to be able to be useful to you as well.

Dave 30:02

Correct. And and ultimately, you know, we talked about the you want to have a memory, whatever works for you. You want to create a memory for yourself. The other thing that I want to kind of go back to because I we kind of moved past it because we do that a lot,

Prompts, Context Limits And Handoffs

Dave 30:17

but I think it's important is you know, you don't necessarily have to be a prompt engineer. Like I know like a couple years ago, it was the idea of how do you build the perfect prompt to get your results. But I do think you have to be mindful of the prompt details, right? As you mentioned, you can cut a loaf of bread in half kind of different ways, right? And AI, every time, unless there's a skill behind it or specific instructions, will pick different ways every single time. That's why, that's why there's the skills, that's why there's the memory of of the actual process. You want to give it that and set it up. The other thing just to be mindful of, and and plenty of times I've gotten frustrated at my AI, you know, probably use a couple curse words here and there, you know, time and time and time again. But uh what it also comes down to is uh when I analyze what I'm doing, there's a couple things that I've noticed. One, if I look back, the prompt is probably too vague and not specific without examples. So, like screenshot things is perfect. So if you find something, screenshot it, send it back, tell it exactly what you're looking for in the image, because it can read, it can look at the image, read it, look at it, figure out what's going on, and then go from there. Number two, I've also realized that the context window has drastically eroded. And so a lot of times, like in in in if you're just using like the web or the desktop version, they don't necessarily have a good way for you to see in the chat what your context is. So what you have to do is you have to build that into your your back end routing of saying, hey, when you get to like 60% of the context window, I want you to kind of like begin to warn me so that I know as I go further, it's probably going to erode. Now, I use a lot of the CLI, so and the below the message bar, I actually have a little bar graph that kind of tells me green, yellow, red. But what I find is is when things are getting off the rails, that was a big thunder. So if I get cut off, that's right, because there's a big storm outside right now. But what I found is is that when the results begin to waver, right? And I'm like, man, what the f is going on? Like, I like, aren't you? Are you not an expert in this that I told you to act like? But and I look down at the bar and I realized the bar is either yellow or red, and I'm like, oh, like that makes sense. Okay, and then I I create a handoff document, and the handoff document tells the next session like what it's been doing. I find that's better. I mean, you could just compact the conversation and you know, start to continue. I found handoffs are better because then it still has all the context and it creates all the document instead of compacting it because it may take something out. You know, so if you compact the conversation, it's it's basically it's basically figuring out what it thinks is the best parts of the conversation and that it needs to keep before it then deletes everything else and then starts new. Whereas a handoff document, at least it has all the information that you've said and it'll put most of that into the next document and then go from there. And then I start a new session and the results can typically turn around.

Duarne 33:22

It's definitely helpful. I mean, anyone who's worked with AI agents knows that these handoff documents can save you so much time. Yeah, a lot of people don't realize that you need to do it. And this is why a lot of these automated systems they build this sort of functionality in. They don't rely on your own.

Dave 33:45

Yeah, Hermes is that way, right? It's like I like and and Hermes, when if you are wanting to play around with Hermes, like you can actually set the number of iterations, which is basically the number of times that it keeps resetting or will run for a process, whether it's like 50, 100, 150, whatever it is. But that's like the compaction number of times that it'll compact before like it's like, hey, I'm I'm wiped out on this, like start brand new, brand new. But the beauty of it, yeah, like you said, it's built in. So you don't have to think about it. Like you run a Hermes task, it's doing it itself at 60%. And you set it. You could set it, you could just say what what you want based on what model you're using in that particular point. But also, like one of the first things that you gotta do is you gotta you gotta have it run an audit on its context files at the start of every session. Like, what's bloating the session? And you may find, especially if you have never done this, you probably found that like all your tools are loading, all your MCPs are loading at the same time. And the beauty of something like a Hermes is that you can say, Hey, I want you to put a top layer where none of that loads, right? And this is what I've done like none of nothing loads unless when it gets to that particular chief. Now that chief has the ability to pull in information and knows where to find it from. So that the beginning context doesn't automatically get bloated because everything in is is bloated and and being brought in. And then you constantly have to like audit that. So like every week have an audit that runs and says, hey, how have things been going? Like what needs to be fixed? And the beauty of it is it'll find it.

Duarne 35:19

And a great way, you know, for people to visualize that is for those who are a little bit older like us, we used to have these things called GPS's we'd put in our car. And before that, you'd have your Refedex or your street directories and stuff you

Keep Information Current

Duarne 35:34

carry around on a big book and you'd go flick through it. But then what you'd do is you get these GPSs you'd mount on your window and you plug them into your cigarette lighter, and you'd be able to get navigated around, right? Like your garments and all those other brands out there. When you got it, you had to make the decision how many maps you would install before you start using it. And you'd have to install those and preload those onto the device. And there was limited space on the devices, so it's not like you could go and do the entire you know, US, for example. You'd have to limit it to a couple of states or something like that in most cases. And it's pretty much the same. Imagine the more stuff the more you put on, the slower it was to load. If you just loaded your city, it would be pretty accurate and it'd be okay. But over time, if you don't keep that updated, what happens? Streets change, roads get blocked. It's not like using a cell phone with waves on it or Google Maps now where it warns you that there's some road work coming up, there's some traffic coming up. These things are live alerts coming through. It wasn't how it was, it was static information back then. It was like having a digital book sitting.

Dave 36:47

Yeah, you had to you had to like download like the yearly update or something, like, or like get the new. I remember I think I had an old Honda Accord that that had like you know, the the it and it it was the coupe and it had the screen in it with the navigation, but it said, you know, hey, this is as of the date, right? So it said if you want the newest version, you got to go to the dealership and buy the updated, you know, DVD to upload. And it's just like, I don't give a shit about it.

Duarne 37:13

Like most of those, you know, but yeah, it's crazy how like some of them were like a thousand bucks to upgrade for the dealership. Yeah, it's crazy.

Dave 37:20

Like they were getting they were getting booked money for that. But now you think about it like on your phone, and literally you have, you know, if they start construction tomorrow on a street that you just drove down, it's already in the system, it's already there, you know, and it's real-time user data because you've probably you've probably seen it, and it's getting feedback from them every time. This is where we talk about how do you build into your own system feedback where it's like, hey, there's construction here, is it still there? Or hey, there's a cop here, is it still there? You know, are they still there? So it can update its system. Yep, your agents are the same way. You you build these checks in, and again, this is the part that's like not integrated yet for all the stuff because it's so custom that it wouldn't make sense for them to build it in, but you would have to build it in to your system. Like, how do I constantly check? How do I how do I update? How do I get feedback in? How do I provide inspiration or updates, anything like that? So that's the biggest piece that you can do, and that's why I always say, like, there's there's always gonna be a human involved. You know, I I really just think it's gonna be you're gonna have when we think of the US, and I guess this is my prediction, you're gonna have a lot more small businesses that can do more and make a bigger impact. And these people that are losing their jobs or whether that they feel like are gonna lose their jobs, I think they're just gonna end up being more productive. Hopefully, they learn a new skill and it can be more productive for themselves, or they go work for another smaller business that you know needs help. Like, because again, and and I'll and and I know you were about to say something, so I apologize, but I I think you still only have 24 hours in a day. So no matter like as you begin to support more people and you begin to you know have more outputs and more meetings and more advisory and more, you know, kind of impact on other people, you are gonna need a little bit more people, you know, time, hours from people to over to do oversight to do dev, right? Like I'm finding that right now. That's why you know we just started with a with a part-time, and I'm sure eventually it's gonna be a full-time, but help on the dev side to keep things moving because you can't do it all yourself. You're always gonna have meetings, and then you're gonna get pulled away, you're gonna have conversations. And so if you have people to help you move things forward, that's the beauty of it. And that's why I always think like I don't I don't see it as eliminating jobs, I think it just makes people more efficient and more productive.

Duarne 39:43

I think so. And this is one of the things that you got it. Well, you've got this there's this argument that I keep hearing that you know, people are gonna lose jobs. Well, some people will lose jobs, I guess, if they don't choose to upskill. If they upskill, then they can

Help Your Team Build New Skills

Duarne 39:58

become multipliers. The AI can multiply their output, reduce their stress, and help them achieve more with less. Now, there's a flip side of this too, because I'm so sick of seeing the publicized, hyped-up videos out there where you know, oh, one man team with an AI of 150 agents is out there making 20 million dollars a year now. Well, that's great, but how are you helping people? Like, you're not employing anybody, you're not helping anyone put food on their table either. You know, as a business owner, that's one thing I'm pretty proud of. The fact that, you know, we have 35, 40 staff now, I think. I've lost count, but I remember every payroll because there's a number that comes up, but yeah, you know, it's somewhere between there. But at the end of the day, I mean, we're talking about that's 35 people that get to feed their family, they get to go out and enjoy and contribute to society and spend money in the community as well. That helps everything grow and develop. Yes, I have AI agents, I have AI helping my team do their job. We are probably looking to upskill some of the positions we have now and backfill with AI agents to do the more mundane stuff, but I mean I've also got agents that just do something simple like go and check timesheets and then point out variation problems. I've got other ones that you know go and check emails and make sure that the emails are actually, you know, going and getting filed in the correct manner. Those sort of things, like, yeah, I could have a person do that sort of stuff, but wouldn't it be better to train a person how to do something more interesting?

Dave 42:30

Well, here's and I think back to you know, you kind of mentioned that, and I think back to a conversation I just had recently, and it's obviously people that aren't using AI. And we were wrapping up a project for a nonprofit, and I was trying to get some information because there's some convoluted, very detailed, sort of like billing stuff like that. And you're like, oh yeah, we just can't do this because there's just so much information, and you know, it's just hard to do. And in my I make the suggestion because we were wrapping up, anyways, from another, but and we've suggested it before, so I don't want to like hit a dead horse. But I those are the situations where I'm just like, like, if you just trained an AI agent to like get the export, break out the information you need after you've trained it and told the director what to look for, and then reformat it in the in the way that you need, and then either upload it or give it to you to review, and then you approve it, like that you're talking five minutes. You know what I mean? Like there's nothing that is hard anymore. It may be time consuming, but there is nothing that is hard anymore.

Duarne 43:36

Everything everything can be done. I mean, you you're in numbers, right? You how many spreadsheets have you had to deal with in your life? How many times did you have a task given to your staff with two spreadsheets and say, I need you to go through and check this against this spreadsheet for this particular concern that we've got? Now, that used to take days in some instance when you've got thousands and thousands and thousands of lines and multiple taps on these spreadsheets, and they would still miss things because there's, you know, you get fatigue doing that. Whereas an AI can go through and do that, cross-check it, reference it, recheck it, 10 minutes to do something that you know would have taken somebody days to do. One prime example that I've got is we had a list of Facebook groups provided to us, and we went through another list that we already had collated, and I needed to make a single list without any duplications. That was a five-minute job for the AI. That would have been a couple of days' work for somebody individually to do it, and I guarantee they still would have made mistakes.

Dave 44:51

Well, I think I I don't know if I gave this example of, you know, I was at a poker tournament. I was sitting next to this nice old lady, and I might have been talking about it when I was just on a solo show. But I was playing a poker tournament, nice old lady, you know, somehow the conversation turned into AI because I think somebody asked about football drafts and you know how they were using it to help them, you know, do some research on the football drafts. But she's in HR and she had mentioned, like, yeah, we just got it, you know, starting to use it. And she's just so excited because her her raw data spreadsheet has you know X number of blank columns. But what she needs when she needs to upload it, she can't upload it with blank columns. And so she used to manually you know have to go through and delete all the blank columns every time she needed to upload it. Well, now with her AI, she just said, Hey, I needed you to remove all the blank columns. She said it was done in like three minutes. She's like, Oh my god, it was such a time saver. You know, and so it's it's it's it's a humbling experience to to remember that there are a lot of people out there that are either at that stage or Or haven't even touched that stage. And that we understand that we are probably and and I wouldn't even say like I'm like top of top, you know, in terms of like percentage of like users. I would say, you know, more top tier, yeah. But I'm more advanced than a lot of people. And I yeah, but but but you're always learning exactly. Like I'm still learning on things, you know.

Duarne 46:13

Every day I learn something new. And you're right. I mean, if you look at it right, one of the things I love doing is I've noticed how much I've been able to teach my wife since we've been sharing an office while we waited for her office to get built out. Yeah. Um, just with how to use AI and some of the tools and how to integrate it into other bits of the business. And it's helped me understand how I can introduce it to other team members as well. And when I look at team members and I ask them to tell me how did you do that? And they tell me, and I go, Oh, did you consider doing it like this? And then they get that, oh, I didn't know I could do that.

unknown 46:51

Right.

Duarne 46:52

It's like, yeah, try that.

Dave 46:53

Well, and I I think the thing you also have to when you talk about team members, and I guess learning this, is you have to train them up that they can do more at the same time. Like, it's not necessarily now about, oh, I'm gonna tell the agent to do this, I'm now gonna sit here and watch it work. Like, which I feel like some people that are starting to do it, they do it. Because it yeah, they're like, they're like, oh, we started, we we did this, we were working with Hermes on this. It's like, okay, but what else did you do then? Oh no, that was it. It's like like no, like, like that's not how you use this effectively or productively. How you use this is hey, I'm we need we have I have this seven things that I need to get done. I'm gonna kick off every one of them, right? Yeah, in with a session. Okay, yeah, and then you come back and you check, and then you say, Okay, where's it at? All right, boom. Now you're now you're again, this is where we talk about productivity. Now you're able to do seven things at the same time because those are all working, and they're all right, and they're all gonna finish at different times. You give feedback, you can check it, and they're all you know being done in in different, you know, shorter sections or departments, however you want to, you know, kind of organize it. But I think that's the training of it.

Duarne 48:11

It's and it's teaching people how to how they can use those tools. It's it's like when you hit print on a big job. Do you go and stand by the printer and wait for your job to print out, or do you just let it go and then just jump up and go and grab the printer, go and go to buy the printer and pick it up later? Right? It's you've it's all about how you choose to think about your job. If you are excuse me, if you're a a builder and you're building something, you've got a plan. You know how you're gonna build it, you've got to build it in stages. But typically you can go and build in different areas on the same job while something's drying or something's getting delivered, or you're short on materials, there's always something to move on to. And this is where I guess as a business owner, we have that mentality that we're always working on lots and lots of projects at the same time, and you've got to make sure your team has a similar sort of mentality when they're using these tools, because if they don't, you won't see any form of growth from them. We've got like we have one staff member in the design team, and they rendered their resignation about a month ago, and they had their final interview with HR this week, and they're asked a question like, I mean, hey, when you're doing design, how do you get your influence? Like, you know, how do you come up with new ideas and stuff? Oh, you know, I just stand up and walk around and do other stuff, and just it comes to me eventually. It's like, oh, okay, that's your process. Yeah, yeah, yeah. I just do a little bit here and then I go and make a coffee and do something, and I'll come back. And even if I'm stuck, I'll go up and you know, read a book or something, and it's like, oh, okay. Now, from a creative point of view, I see why that's the case, but then I start thinking about well, we've got like 15 projects on the go at the same time. If you're stuck on one, maybe you could look at another one, get some inspiration on that one, and then come back to the one you're working on. And that's how I work. Like I've always got three to five things running at the same time, and it gets overwhelming to the point sometimes. I have to go stop, focus on one, full focus on one. And the teams I find you've just you've got to introduce some discipline and training to them so they can utilize that. It's no different to oh, I asked Bob, and Bob said that you know he had to check with Mary and he'd get back to me. That's why I didn't finish that task, right? Or did you follow him up? Bob said he would get back to me.

Dave 50:55

Yeah, yeah, I was just waiting for it. Yeah, yeah.

Duarne 50:58

Well, so what'd you do while you're waiting for it? No, nothing. So there was nothing else you could work on at that time?

SPEAKER_02 51:05

No.

Duarne 51:06

And that's where I think it comes down to your training technique. And you and sometimes like one of the things we're going to be introducing is I want to an agent that's basically a motivational prompting agent designed that just comes along and checks in occasionally on people. It says, Hey, what are you stuck on? What's going on? You know, checking in and oh yeah, have you thought about this? Tell me something you're stuck on right now, and you know, maybe I can give you, I can, you know, help you out with that and you know get some assistance on that. And it's really no different to a person doing that. But from a small business point of view, you also have to be careful of the psychology behind the users and your staff using these technologies because some of them think they're training something to replace them, right? And you have to be you have to be wary and say, Well, you know, do you actually enjoy doing that part of your job? Well, no, if you don't, why don't we get the AI doing that and tell me more about the things you do enjoy doing? Or let let's try and get you doing more of that and let the AI deal with that.

Dave 52:11

And so and if you're a business owner, like same way, you know, that's that's how you think of it. It's like, what part of your business do you not enjoy uh working on? And that's where you start because then that frees you up to do more of the stuff in your business that you and do enjoy. So maybe that's having the client conversations, or maybe that's you know, thinking through the analysis or whatever it is. Like the idea is that we're not telling you to uh, you know, uh replace your entire business model with AI. That's not the goal. The goal is like, how do you become more efficient to free yourself up to do the things that you want to do uh in your business? Because once you're happy in your business and you're doing the things that you love in your business, that's when your business is actually gonna start thriving. Because you're not wasting away doing the things that take up a lot of your time that don't give you the freedom to actually build your business and grow it and make bigger impacts. That's the beauty of where you're at right now. You have you and you're on like the cost of this is like the I mean, I guess that you know there's always the worry of like, oh my god, this is gonna, you know, kind of fall flat, it's gonna, you know, kind of be a bubble, right? I don't think you as a small business owner have to worry about the bubble, right? You know, like you're not building and you're not trying to grow an AI model and try to sell it. Like if you build systems and you build processes, the the anthropics, the claws, right? All you know, the the open AIs, the Googles, they're not going away with their AI. It may change, I'm sure, eventually, once they figure out, hey, we can't just keep burning hot millions of dollars you know a month, you know, on all this stuff. But it may change and it may look a little bit different. But in the end, the people that are going to still be thriving and successful are gonna be the people that have the system built that yeah, they can still run it, whatever that looks like. And that's where you're at the opportunity right now to do that.

Duarne 53:59

And that's a good point. I mean, if you look at it, like people who say, look, I don't really want to spend $20 a month or something to help me save a bunch of hours of my time. They're gonna have a real struggle to spend a hundred or two hundred dollars on the bigger, more successful plans, or get two plans to two different software packages that can do the

Budgeting For AI Tools

Duarne 54:20

same, that can do similar things, but better, some do better than the other, right? And play them off against each other. So there's gonna be those people, but there's gonna be a lot of people out there like who are gonna go, you know what? $200 a month, I'm able to achieve. I think that's what in if you give it about a year or two, I think that's what the average business is gonna have to be spending for their AI, at least $200 a month for the subscription, is gonna get them like a good use case with a lot of functionality and still gonna have and not gonna stall out halfway through generating applications, use cases, etc., all the time. So I mean, if you if you do let's say it's $200 a month, you're gonna have to commit to a platform to help you save hours and hours and hours. Or as a small business, not have to outsource your design work because you can do design in-house, not have to outsource your you know, content writing for your website, because you can do it in-house, not have to, you know, always go and outsource, you know, to get extra help in the uh busy seasons for you know when you have to take more phone calls because or get more influx of emails that you have to handle because these systems can help you handle that sort of thing. I mean, there's gonna be there's a lot of use cases where you're gonna save so much more money than it's it gonna cost you. Yes, it's a heavily subsidized right now, and most people could probably get away with the $20 a month plan, but I reckon in about 18 months we're probably gonna be working on an average of $200 minimum plans to be able to have anything of substantial use so that these AI companies continue to actually operate and pay, you know, make sure that we're gonna do that.

Dave 56:10

Yeah, I I don't think I don't, yeah, I think the $20 plans will go away, or at least the 20, the $20 plans, I think, will only be like chat versions, right? It won't be like, hey, you could use co-worker, and they're even starting to do some of that, you know, like the high-level new things are only coming out at like the 200 max plans, pro plans. But I think that's gonna be like the basis for you to use any sort of like CLI, anything like that. I think they're gonna completely have to remodel their their tokens. I think 200 is gonna be the min. And for people that are on 200 right now, I think we're gonna be paying somewhere in the north of five, six hundred dollars a month for that plan. But again, when you look at what it's saving and what it what the ability it gets you to do now, if you're using it in a way that's generating revenue or producing outcomes, there's some people that are just like building just a build because they have the idea and they have no way to monetize it. There's a difference. So if you're using it because you're you're helping build something for a client or you're helping build something that you're gonna be able to monetize, then great, do it. But be mindful. You're not gonna I I think gone is the day of these ginormous conglomerate SaaS companies. I think you have to keep in mind how do we build something that supports people that still has the ability to be customized by a business owner because uh everybody wants something a little bit different, and that's what's gonna set you apart. So if you if you have some sort of solution that is a general box, but that box

Software Costs And Integrations

Dave 57:39

can be customized or used as the general, like because some people may want and ultimately what I I think of it like that's what we're building the command center around. Like we have we're gonna have general functionalities that people can, you know, A, D or C pro plan. Here's the usage you get, etc. But then we also have the custom version of it that allows people to set it up how they want, and we'll be able to kind of go into theirs and and customize their their ability to do things in there. And so that's where I think the future of SaaS is. I don't think we're gonna have these big huge SaaS because people can just do whatever they want then if they if they really want it, then you can do it yourself.

Duarne 58:17

I think yeah, I think you're getting to that point. So I mean, you look at something like Salesforce, HubSpot, these platforms are charging north of $50 to $100 a user per month. You've got an organization that's 50, 60 users, it adds up really quickly. Hundreds of users adds up really quickly. It's got to the point now where you can actually put a lot of a lot of these companies have been stuck in positions of using these different applications that have high costs associated with them because they are the best for their industry. Not because they're the best software, but because it's the best option for their industry, or their data lives over here, and this software communicates with it natively. Well, we're getting to the point now where you can build integrations with AI. You know, I was playing with something earlier today which said, hey, there is no API for this platform. But if you would like to do these multiple steps here and give us some access, we might be able to build some sort of interface to interact directly with it if you could if we could pull this data out of the system. Would you like to connect it and try? And the app I was using didn't do it, but I gave it to another app that I have, and it was able to run those tests, and it was able to prove it, and it was able to build me a prototype API, which works. And that's the sort of world we're living in right now. It's a little scary, but there is possibility. Now, if you're sitting there going, I don't know what you're talking about with APIs, I don't want to develop anything. That's cool. You don't have to. There's people out there like us who can do that, and there's people out like us out there willing to do that, and it doesn't cost hundreds of thousands of dollars to do it anymore. It costs a lot, lot less, and it doesn't take 12 months, 18 months, two years to do it anymore. It's done in a matter of weeks and months, right? So, this is the difference we're talking about. This is where these big SaaS companies, I think you're right, I think their days are numbered. I think they've gotten away with being able to charge a lot for a long time. They've got big teams, and if you look at a lot of these big corporations out there who are SaaS companies, for example, they have lost a lot of staff, they have made redundancies, they have reduced the amount of staffing in their companies because they know that AI is able to do a lot of the stuff that their teams are doing right now, and they're adapting it. But what we're not seeing is them passing those savings across to the front of their products. We're not seeing that yet. We're not seeing massive reductions in the subscription costs, we're not seeing massive product feature releases that we don't have to pay extra for. So there's you know, people are gonna finally catch on to this. They're gonna go, well, what are we getting for this money? I can go and sign up to this and get this. And there's got to be a benefit there for it. And if they're not seeing the benefit, then they're gonna really struggle to keep it.

Dave 1:01:35

Well, and and and you gotta, yeah, I mean you gotta think about it. Like 80% of people are still only gonna use the the basic version of it. So I I I don't think I guess I don't think if it I don't want this to deter anybody from saying, like, well, I don't want uh I was gonna build a platform or some sort of service that solved the problem. But I also feel like when you're building it, just build it the right way. Like just know on the back end that you may have people that love your software, but maybe it's missing one or two things that they want to see. So great, let's go ahead and and you know, we have a custom version that we could help you customize it for yourself, right?

Duarne 1:02:11

Or you know what there's options there too, like you know, Zapia, everyone's familiar with Zapier, right? The app that links everything together through WebWalks. Right. Zapia went out and released an MCP server. They've also got an SDK pack for installers for developers, and they give them away to be able to be used in app development now, so that when you build an app, you can basically say, Hey, from day one, I'm gonna allow you to connect it with anything you want. Just go and sign up for Zapia and you can link to our app. I mean, that's a huge statement. There's very, very little development cost or time and effort to do that. And then suddenly you're giving people the ability to get data in and out of your system that you've built, your platform, into anything else they've got.

Dave 1:03:03

Well, but I and it I also like Xavier is great, I think, for a lot of things. I don't really use it at all. I think the negative side to that is now the client now has to pay extra for everything that they want to connect, right? In terms of zaps, every every time it like sends one piece of information, that's a zap. And so, like, if you're talking, you got a lot of data, you got a lot of stuff for them. So it really depends, I guess, depends on one parts to it. Yeah, there's there's cross and guns. I would say, like, too. Like, if you are developing, if you are a developer listening to this, like find the APIs, like that's the best way is to connect through APIs.

Duarne 1:03:40

Um absolutely, but I mean, you if you're an A, if you can do it where you need to just get something out there and start using it, do Zapia to start with, and then as people come to you and say, hey, look, I love that I can connect my app to you, I would love to do it a bit cheaper. Do you have a roadmap that you're gonna be integrating with this platform in the future? You know what? They got a public API. I could do that. Let's put it on the roadmap, let's get that done for you, you know, and build it out accordingly. But you've got to you could use it to start with and then grow and develop it rather than you know. You know how many times I've spoken to people in development world and they're like, Oh, we need to be able to link with this, this, this, this, this, this, this, this, this, this on day one.

Dave 1:04:23

So everybody wants to get to everything on day one, yeah.

Duarne 1:04:26

It's like, do you need to? So, all right, well, there's an option to do it if you want, but how about we do a native integration to this and then we put this in, and anyone else who wants to integrate on day one can just pay them to integrate with us and do it that way. If they need more than that, let's take a vote as to what goes next. And if those people who want to use it are going, you know, a substantial volume of customer base, then let's do it. But yeah, there's definitely pros and cons with all of these things, but it really comes down to again, what are you trying to achieve with this? And like we keep talking about development because we love the doing the development. But for the majority of our listeners, I I believe they're probably going to be more in line to get something like Grokbot and do a little bit of tinkering, Claude Cowork, do a little bit of tinkering there, maybe experiment with a little bit of Hermes and sit there and get a little bit scared from the initial load up. It's a lot easier with a Windows desktop version now than what it was, but it's still effectively just a chat interface. So until we actually get some sort of level of beautiful, easy to interact with dashboard with can-band drop and drag functionality of tasks and you know, complete C, you know, clarity on what our agents are working on. I think it's going to be kind of like work from home. You got a bunch of agents working from home, you don't really know what they're doing unless you ask them. And then they tell you, oh, I've been sitting here twiddling my thumb for 45 minutes because I had this problem that was blocking me.

Dave 1:06:01

Right. Or or or the gateway broke down and when it came back up, it didn't get restarted, all that stuff. So it's like, and and that's half and and and so these are the these are the everyday frustrations I think that we have, right? As developers and people that are using it on a regular basis. So it's not like it's not all you know, unicorns and and you know, rainbows and unicorn kisses, right? It's it's always like, and I just remember that my buddy who's a tattoo artist told me that one time, and I say it all the time. He's like, it's not because I asked, you know, it was one of my first tattoos. He's like, All right, you ready? It's not gonna be rainbows and unicorn kisses. I was like, Yeah, let's go. But you know, yeah. Pain. But yeah, I mean, the these are the everyday sort of pains and struggles that you know everybody's going through. Like, these aren't things that are just like just you and I, right? These are gonna be things that you're going to experience. And I think what you do is you just kind of work your way through it. But you know, you mentioned the the dashboard and all that kind of stuff, and and and I just realized we're at an hour and ten minutes, so gotta jump to a 10 a.m. that I typically have. But I think that's maybe a good conversation next week is start, you know, maybe we even um you know, it's a week's a pretty long time. I think we're both in agreement. We've been talking about this, like, you know, kind of dashboard able to like view it. I think that's something we can talk about. Yeah, I I think yeah, because like in in command center, I'm I'm that's what I'm kind of doing. So I think maybe we we definitely work together. We can like bring that up and talk about how we do it next week, right? In terms of yeah, what it looks like and why it's important. Because I think that is important. Like

See What Your AI Team Is Doing

Dave 1:07:32

you can't just let it all go off. And part of the struggle is how do I, you know, Hermes in Slack, right? Or Hermes in Command Center, which is something that we're building, right? Where you can have it in command center and just talk to it in the same way you do a Slack. There's eight million different messages that you could be having going at one time, you know what I mean? So it's like which ones are open, which ones aren't. Like I think we talk about it next week at a better way to manage that.

Duarne 1:07:57

Absolutely. And look, I've been testing and playing with a few things. I've been playing with Hermes 3D. It looks really fun. It's like a 3D looking map, but it kind of useless in other essences, where you can't really use utilize the chat functionality to the full aspect just yet. The big concern for me was the amount of resources it was pulling. I realized I had two instances running and I was pulling three gig of memory on one. Wow. And I was like, that's just a huge resource draw. So I shut them down. But it was really nice to get some visibility to my team, but I now know what I like and what I don't like. And I don't need a 3D view of my team, but I kind of would like the idea of a smaller 2D style, you know, Zelda PG style rooms where I could see if my agents are going to work together, they wander off and work together. And they can be off working and having a meeting somewhere talking about a topic. And if I click there, it'll tell me what's all about. And I can check some meeting notes about it and things like that. I think there's lots of fun things you could do. But at the end of the day, I mean, do I need all that, or do I just need a big Kanban board that I can see whatever?

Dave 1:09:17

I was I was, I mean, the way I think about it is you have like, here's all the different like for me, you know, we have our chiefs of staff, right? We have chief CFO, CEO, whatever. And then under there you have all your departments. For me, I'm thinking of it as each one almost has its own sort of like, you know, what's been what's what's in work, what's in process, what's stalled, right? What's what's wrapped up or what's ready to review, and then each one of those, and then you you log in. Because when you think of like your room view, I think that's good if you just had a person, but like how is how is it gonna handle when you know Jane is doing five different tasks with five different people? Is it gonna show her talking to all five people? You know, now you're talking about five extra five, like five, five extra, you know, sort of Minecraft carrier or Minecraft, you know, people walking, you know, the rooms. And so now you're talking about resources, now it's it's there's a lot of extra stuff.

Duarne 1:10:14

So it's like, how do you that's a really good point, and that's and that's where like you're right, it really comes down to how utilized are your agents, and do you need to create more of a type of agent because it's over it's got too many reasons, it's doing too much. Right. These are the sort of things that you need to be able to visually see. What I've realized is I'm probably gonna run a big TV up on my wall over here with whatever

Calendar Agents And Traffic Alerts

Duarne 1:10:42

dashboard I'm running, so I can have a visual and just look over and see what's going on at any point in time. But there was one thing I was gonna mention earlier. We were talking about, I know we normally do the what would your takeaway be, but when we were talking earlier, one of the things you were talking about was how to utilize agents and make it work better, etc. And we start talking about maps and things. Right. I had an idea. If you've got an agent whose whole job is to sit there and look after your calendar, or for example, you've got a fleet of people who are out doing meetings and appointments all day, because you're a service-based industry, you may want to have your agent upskill in the future to be able to check live reports of traffic situations. And because they already know where the meeting's going to be, because it's a physical meeting that they have to be traveled to, you could teach your agent to come up and warn your team members with alerts and warnings to say, hey, look, there's a traffic situation on the M4. It's probably best taking the M3 and cutting a crossover to this road here, or potentially saying, Hey, look, you probably need to leave 10 minutes earlier than you probably would normally. So you should leave at this time because it could help you. Right. It could also have the ability to integrate into your system if you're a service-based industry and say, hey, look, just letting you know the traffic's a bit heavy over here, service technicians take a little longer to get there, they'll be there here, be at your place a little bit late. We're estimating about this time of arrival. And all of that could be automated through your agent potentially, which could give you better customer satisfaction scores. And why is that useful? Because for a small business, a lot of small businesses don't have somebody sitting there able to handle that sort of workload on top. This is where once you create the basic foundational structure of what you need, you can come up with these little ideas and go, I wonder if I could do this. How could I make this a little bit better? And it's worth just sitting down and going, What can I improve now? What can I add to this to make this even better? And I think that's the key. When you start down this path, you will find yourself improving these systems you create with your agents, which are just basically this is what this is and adding this is adding on to the systems and the procedures and the processes and the SOPs that we're always talking about. We're talking about the execution level now. Talking about being able to execute it with less people, with less, no additional time resources from yourself, being able to do it with the assistance of some AI agents to help you out.

Dave 1:13:27

Right. And again, it all comes down to yeah, I mean, I love it. I think it all comes down to start small. Like if you haven't done any of this, like don't get overboard. Like you're you it you're easy to get overwhelmed in terms of going down a rabbit hole. So I think really if you're just getting started, one of the first things we talked about, which I think was helpful to you, was uh you know putting some sort of guardrails around yourself. And you can even build that into your into your system where it's like these are my priorities over the next 90 days. This is what I need to solve, this is what I need to do to grow. Anytime I start a new session with you, you need to check right my roadmap. And if it doesn't align, you need to push back on me and say, is this something that's immediate because it's for like a client need or a specific, you know, focus point that maybe changed. But if it's not, roadmap it. And then after you achieve what you want to in your 90 days, then you can start, then you can go back to it and say, okay, I've achieved what I wanted to. Let's look at the roadmap and kind of make our next 90 day or even start 30-day goal and then have it keep you aligned because otherwise you're just gonna you can you can go a million different directions and you get off track very, very quickly. So 100%. I I mean I have it in all of my projects, you know, I I think the limit is five things being worked on at any one time, and then it does push back and say, hey, this isn't on the roadmap. Like, is what are you taking off to put this on, or do we put this as one of the next up things, which that we're not always consistently working on? So we're not saying stop your ideas, but you basically say, Hey, I had this idea, put mark it down for me, put it on the roadmap so we can come back to it later. And then anytime you say, Hey, give me your roadmap update, what's there, and then you could have it help you revisit it, etc.

Duarne 1:15:17

That's but with that, I love it. Scopes, right? You create a scope, and if without a scope, move it to milestone, the next milestone.

Dave 1:15:26

Yep. Love it. Gotta jump. We love you all. Thanks for watching. If there's anything you have questions on or anything like that, always feel free if you're watching the replay, or even if you're listening to this audio version on you know, wherever you're listening to the audio podcast we have out there. Uh, feel free to drop a question. There's an email down below. Feel free to drop uh an email with any of your questions as well. We'll be happy to answer those. And as we always say, do all that fun algorithmic stuff,

Questions And Closing

Dave 1:15:49

like, subscribe, all that fun stuff. And we look forward to seeing you every week. We go live right for the right around 8 30 a.m. uh every Friday morning. So we hope you guys have a wonderful and amazing week. Duarn, thanks for joining me, brother. Go uh get some rest, enjoy your weekend. Tell Paige has said hello. And for everybody else, see you guys in the next one.

Duarne 1:16:06

Thanks, Dad. Take care, everyone.

Dave 1:16:07

Bye all. Peace.