Time to talk about the power of AI!
Chris, George B. Thomas, and Nico discuss organizing AI-agent teams, with Chris sharing his 42-agent structure using one-word role names like Scribe and Marshall under a virtual COO. They explain how command-line interfaces for Google Workspace, Supabase, and GitHub let their AI system Echo handle more tasks while cutting token consumption and improving coordination between agents.
Chapters
Good morning, LinkedIn friends. Value First Nation, welcome to another episode of Value First AI Daily, your collaborative intelligence report. It is Tuesday, March 10th, 2026. How we doing, George B.?
Yeah, doing good. How are you doing, my friend?
I am doing, uh, wonderful as well.
Nice.
Uh, Nico, how we doing this Tuesday?
Doing well, doing very well.
Very nice. Rocking out the flannel, loving that.
And some new glasses, maybe some new glasses.
Finally, uh, we can see after 3 long years. Um, you know, we got, we got the old man's going now, so we—
there you go. Welcome to the club, my friend. Welcome to the club.
Yep, yep.
Oh man, speaking of old man, Like, I had my first, like, slip and fall, uh, over the weekend.
Oh man.
Um, uh, yeah, it makes me want to stay inside and work with AI where it's safe.
Roll, roll up, roll up in bubble wrap every time you go out.
Yeah, just going to stay inside from now on. It's too dangerous out there.
Oh man, I was out for a walk And it had been raining, so it's like puddles form in the normal walking path. So I take some— take a little detour, and then I find some sprinklers, and I'm like, all right, I gotta figure out how to get through this. And it's just, uh, going down a hill, and as soon as my feet got together, I was like, whoop, bye-bye, man. Yeah, I've never hit the ground so hard.
That rain is dangerous.
It is. Rain plus sprinklers, because, you know, there's no reason Like we got to make sure it's watered even though it's raining.
Yeah.
Probably on a timer.
Yeah.
You know, the weird things humans do.
Yeah, exactly. Like, can we get AI in that system or something?
Right? Is raining, does not compute, shut off. Yeah.
Oh man. I was still able to get a ton done on Sunday after that incident. Um, uh, yeah, I, um, I started— I, I think the first, uh, request was that I wanted to, um, kind of finish out the process of building out this organizational infrastructure. And, uh, let me tell you what, like, I've been wanting I've been suggesting that a company that wants to take digital transformation and doing all of the new things seriously, like since 2020, when I joined my first big company, and I'm supposed to lead marketing and digital transformation and customer experience. And guess what? There's nobody around, like, to help me do that. So I got to build a team of, of, you know, 5 more people.
Yeah.
But then you get approvals for things like, all right, we want to do an online store, we want to do an online academy, we want to do, you know, all of this stuff. And my point was always, well, we need somebody that can focus on, on doing that stuff then. And that somebody cannot be the service manager that's just gonna somehow take on this extra hat of something he's never done before and then expect him to do a great job. So I've been kind of org, org building, org designing, like, since then. And so this is scratching that itch. Wonderfully. And I'm just going to share my screen here. It's been super interesting to build this out and definitely some inspiration from you guys as we talk about agent teams and agents and subagents and skills and how to navigate. And actually build it out, um, getting to the point of like yesterday finally, um, I did an all-hands, uh, oh wow, with 42 agents and their 3 leaders and me. And the first one, um, uh, like I said, the purpose was, uh, was to like so that everybody could meet each other Right? Everybody could know everybody else exists because that's like the hard part. It was the hard part so far is to like, okay, build all these agents and skills and like, if you just tell Claude to go do something, he's not going to use any of that if there's no specific instructions or registry or anything to like tell him exactly what to use and when. Right, capable of doing it all himself, of course. Um, and but then when we get done with it, uh, I was like, so how'd it go, V? And he was like, oh, it was cool, uh, but it was really just me writing on behalf of everybody. Like, no meeting people, Chris. Like, they're just agents. Like, what do they— you just call them in when they need to do stuff. There's nobody like actually introducing. And I was like, okay, well, how do we do that? Because Nico and I have talked about an office and a break room and a way that if we're going to build all these identities and roles and responsibilities, it makes sense to me as a human, right? Not necessarily that you have as to why I might want to enable that, right? And, and so here's an example, right? Like client delivery. Another change that was made is they all got one-word names, right? So Scribe, transform session recordings into compounding knowledge. Uh, Marshall, keep SubSpot project and task data honest. Uh, Ledger, maintain documents index per client, right? And these are all under V, the COO, who is super lopsided, like trying to figure out how to, how to balance the org a little bit. We've got content and media, platform, and then self-improvement. So one of the things added was corrective action reports and capability reports. So like when I keep seeing him trying to hit the API in the old way that we're not supposed to be using anymore, and just red text after red text after red text, and he gets it. It's like, oh, I'm not supposed to be using that MCP, or the token's over here and not over there, right? Trying to document those in a way that tries to clean up a lot of that process. So that's where Q, for quality, comes into play. We've got Sage, the chief customer officer, but some relationship intelligence. And just a pop quiz, what do you think Harold does. I mean, any guesses?
I mean, is the name— are these names literal or are they just given?
They're literal in most cases.
Um, does Harold bring everybody together and/or like bring forward ideas?
Well, so I— listen, when I saw that, the first thing that came to mind was Hark the Herald Angels Sing. Glory to the newborn king. But I have no idea if that's the direction that you took this, uh, agent or not.
Uh, prepares Chris for every client session so nobody walks in cold. Synthesizes relationship intelligence, HubSpot status, uh, team coaching content. So lets everybody know what's, what's on the horizon, right?
Um, that's great.
And So as you click in here, like key value indicators for humans, for AI. And this has been the challenge and why I ended up going here because when I first said I wanted job descriptions for everybody, Vee was like, okay, well, since it's just for me, like I'm the one calling these agents in, I only need this. Skills, receives, feeds, not an actual job description. And I was like, yeah, man, but I'm trying to create an understanding so that I can help everybody basically not see that red text over and over. When we did it right yesterday, why can't we do it right today? And all of those things. And for me to be able to impact that, I have to understand like what the role is and why they were called in and what the expected outcome is. And so that's where there's a, you know, everybody's got a job description that has for humans and for AI in here. And it's just been really enlightening. And one of the inspirations was listening to something about agent teams where it's like one agent, one role is an effective way to manage and to create agents at scale or manage agents at scale, orchestrate them, I guess, is the right way to put it. This has been a fun endeavor and then last night, Finally, I was able to, you know, just give a prompt without instructing specific named agents to be used. And then those named agents got used. And I was like, there we go. Crossed out. Happy. Right.
So that's like crazy though. Like when you first see So, so like orchestration, like the whole, um, you know, even with the Google CLI, so like that real quick, like what's Google CLI and why is it so crazy now? Um, Cloud Code doesn't natively have like external connections, not like, um, Cloud Desktop does, because Cloud Desktop can leverage your connectors like Cowork and Chat can leverage your connectors. Cloud Code doesn't. It's kind of disconnected from that. So the Google CLI at least allows Cloud Code to have this like, um, native connection to your Google accounts. So Gmail, Calendar, Drive, all that, which is really awesome because then you don't have to start, uh, keep storing stuff on your desktop in order to have code, understand the context of it. Um, but all of this stuff is not like— like OpenClaw and all that, it's really great, but it's like really great singular action and like really great if you want to like stay there and work on the one thing, I guess. Um, but like I'm sure you guys are the same way that I am, where it's like, hey, I've already, you know, I start the day with like 3 windows open, right? And like, project 1, okay, boom, Aries, you gotta go work on this thing, see ya. Project 2, hey Aries, you know, we're working on this thing, see ya. All right, project 3, hey Aries, we're working on this one, right? Like, and then in those they're reaching for their other agents. And maybe in one it's grabbing a PRD architect and then another architect and then grabs a security guard. And then the other one, it's grabbing a review agent and then it's grabbing another agent for checking out the planned architecture and whatever the updates are. And then it goes, checks with the custodian. There's all these check-ins and stuff that happen. Before any, like, tasking even gets done, right? And yeah, it's— I know there's, you know, some people talk about, like, token waste and stuff. I came across a nice tool yesterday that actually, like, cuts that down tremendously. So, like, now, man, I was running, like, real hard all day yesterday and didn't, didn't get close to— I mean, I got close, it was like 75, 80%, but being able to orchestrate multiple agents and have conversations and we're We're talking about like there's a shared message board where they all talk to each other, like they all leave notes for each other. I've gotten to the point now, you guys, where I have them send tasks to each other. So like mid-working on stuff, and they could be— it doesn't matter, it could be separate IDEs. I might come up with an idea while I'm working on one. And tell Aries, like, hey, you're already parallel working on this. Can you please send a task over to the Architect Ninja? Tell him to sketch this in place on the other project while we're working on this. Boom, task made, written on the board. Architect, I'll go, you know, I don't even have to do anything in the other one. I'll just see the yellow name pop up. And here goes Architect Ninja, just read a task off the board, blah, blah, blah. It's like, gosh, it's getting nice, it's getting real nice. Like, all I need is just that, that separated— like, wish I could just pull up my phone and just be like, hey Aries, what's going on this morning? Without even having to like pull up an app or anything like that, right? Like, it's that, that shortcut button on my phone goes straight to Aries, right? Like, that's It's coming, it's coming. Like 3, 3— I know I keep saying it, but like 2 and a half months now, maybe like 2 months, I don't know, maybe a month. Whatever the next iteration is, uh, is gonna plug that gap for me. Yeah, and it's only because I'm too lazy to look into developing mobile apps right now, so I don't have time, right?
What do you got, George? I know you've been up to something blown up over there.
I can see it's blown up.
I mean, well, I'm, I'm listening to you guys and it's funny because at first I was like— which, by the way, I didn't even notice, um, Nico's hat at first— but when I saw that thing, uh, that you showed of like the different agents and you were kind of going through that, and I was like, And this dude is building like Claude SimCity where it's like he's got all these like fake humans, these agents instead of humans in like a SimCity scenario. And, and that was interesting to me because I really hadn't thought about it at that deep of level. Now I have used agent teams, but I haven't like made those agents so that they're like, we recall per se the same agent every time. Like, well, we'll recall a writer. Is it the same writer? I don't know, doubt it. But is it a writer? You know what I mean? And so I hadn't gone to the lengths of like naming them and like an organizational chart. And I think that's partly because my brain doesn't work that way normally. But I find it very interesting. Especially when Nico talks about like this board idea and being able to pass an idea over to another one. Because I too, uh, I mean, listen, since morning is no different than many mornings, I can look over here and I have two terminals running in parallel working on stuff, right? And that started at, I don't know, uh, butt crack dawn this morning when I walked into my office to just get things going. And so it's something interesting that I, I think I'll think about of like, should there be more structure around Echo? And if so, what does that look like for me? I've been hammering down. You mentioned the Google CLI, Nico. I've been hammering down in two directions. One, we're updating or upgrading Echo with The Google Workspace CLI, Supabase CLI, GitHub CLI, um, there's another one but it's, um, I'm forgetting what it is, but, but a bunch of CLIs and kind of basically, uh, rocking off of MCPs and going into CLI directions. Again, hashtag token usage. Um, oh, Playwright CLI, that came out, which is like way lighter than Playwright MCP. You should definitely check that out if you haven't. So we're rocking in that way where we're just giving like Echo more powers with less token usage. The other thing I was telling Chris in the green room that we're rocking out is allowing AI to kind of take its journey even when it scares you. And what I mean by that is I asked the question, what are we missing? What should we build next? And this was specifically for the spiritual side of leadership site. Um, and all of a sudden it just came up with a project plan of like, um, accountability partners and, uh, membership— not membership, community. And, um, it even threw out this idea of a, um, community prayer board and like all these things that I really was not like, that's gonna be on the roadmap for this project. But I sat and looked at it and I was like, huh, it knows at a deep level what we've been building. It knows what we've got. I asked the question of what is the gap and this is what it's showing me. I don't know the destination, but enter. Let's build it. Let's build it and let's see what happens. And so it's just interesting to to almost jump off of a cliff in a very exciting way, um, knowing that I'll be working hand in hand to learn what we're building as we're building it. So that's kind of like two main things over here.
George?
Yeah?
Would you ever preach to a digital congregation?
Oh man, okay, so now you're getting— you're getting— so I don't know if I would preach to it.
You open the door.
Yeah, I know. I don't know if I would preach to a digital congregation per se, but it's interesting that you ask me this question.
Do you think that it would be beneficial to them?
Well, so here's what I've always wondered, because by the way, you've got it, you've got to think about this. Is for— let's see, 200. Actually, I can tell you, I can tell you. And by the way, the things that are about to come out of my mouth are, um, I am not, um, I am not, uh, I'm, I'm not— I don't— how do I say this? Like, what I'm about to say, I'm not, um, bragging because it hasn't always, um— I need an eye in studio. Uh, hang on one second. I'm not bragging, but— no, not even but— I'm not bragging. Okay, so I'm just trying to see. Okay, for— so I know I got some in the queue still. So we're, we're rounding close, Nico and Chris, to 300 days straight that I've woken up, I've grabbed my Bible app, I've taken that daily scripture, I've taken that daily devotion, and I've put it through a content system where I dig deeper and learn. And the output is the article that then is focused on this topic for faith-based leaders. So like, so I've— through this journey, I've often wondered, like, what am I teaching the AI, or what am I putting into the system that maybe is like new synthetic understanding based on the direction that we're— because by the way, we're filling a gap. There really is no, um, person out there. There's pastors that are preaching like life, and there's business leaders out there teaching leadership, but there's not really a pastor leader teaching faith-based leadership, at least at like the direction or level that we're— I'm sure there's another human out there somewhere, I just don't know them, right? And, and, but it's not like mainstream. And so I've often wondered, like, what am I doing to impact the actual model based on the thing that we're creating. And I don't know, um, but I wonder if there are enough humans, if they were using AI for their spiritual walk, if it does something to the system. So again, would I ever preach to an agent congregation? Have I been? Have I been for 300 and some days? I don't know. I, I don't know. It's weird to talk about, to be honest with you. But, but like, but it's no different than— this is gonna sound weird— it's no different than who I am in the real world. Like, I've been spreading kind of spiritual seeds along the way instead of 2x4s, and I'm kind of doing that in the digital realm as well now. Like, anyway, interesting question. Interesting question.
Yeah. Oh, that's the part where it's like, since you've been doing this for a while.
Yeah.
This being like just content and sharing knowledge and just, you know, putting yourself out there. People come in, they start talking about AEO and it's like, When we understand how AI works and that we're having conversations with it every day and making content, like, it's like, yeah, it's gonna, it's gonna take care of itself.
Yeah.
I can't tell you exactly how.
Yeah.
But that's what I'm super interested in when we talk about like training data and overcoming like how people view HubSpot, right? I'm like, what? That's what's super interesting and kind of a challenge that I want to like see when it flips, right? Like when does the training data get overcome and how much content is actually needed, what kind of depth, like all that stuff, right? And there's so many areas for that, um, as far as the way things have been done versus the way things will need to be done now that we have this new intelligence layer of of civilization. Yeah. And yeah. And that's where we'll leave it for today. When we come back tomorrow, Nico can tell us more about what he's been working on. And if you're out there listening and you want to share stuff that you've been building on the show or just in the comments, we'd love to hear from you. It's the best way to teach each other, folks. So I appreciate you joining us, and until tomorrow, have a great day.
