If you haven't already checked out what

Grockbot could actually do, that's okay.

No big deal. I got you covered. We're

going to dig a little deeper this time.

How you can run a oneperson operation or

even a small team. In fact, not just one

person shops. Every person, in my

opinion, should more or less have their

own Crockbot fleet. We stood up a whole

fleet, chief of AI staff at the top,

specialists underneath. So much so that

even my own employees have their own

fleet of agents that do the work for

them as well. We're talking the same

computer in the cloud, same always on

mentality, and different jobs. So, in

this video, I'm going to break down how

to run it as a true business operation

and as an agency handling multiple

clients. Let's go and get into it. If

you go to x.ai/bot,

and then you can even see the news uh on

everything that they're doing, right?

They are on fire when it comes to

shipping things. Uh just big time. I

mean, even since as of this recording,

they're just constantly shipping. In

fact, there's rumors that you can

actually call your Grockbot now and

actually have a live conversation as if

you're calling an employee, right? So,

interactively, you can obviously click

on here. You can see some of the

different roles on how you can actually

create different agents, but we will

save that for another time. We're

actually going to go ahead and just dive

in right away. We have here the recent

version of Grockbot, and the product has

definitely evolved quite a bit, but at

the bare bones, you basically start with

chief of AI staff. Just say to your

chief of AI staff, hey, this is my

business. These are the different

functions that we do. This is how I want

to operate. Or better yet, how would you

operate this if I'm trying to reach a

certain goal? It could be a particular

ARR in terms of annual recurring

revenue, could be MR, maybe you have a

new idea, maybe you're trying to reach a

particular goal in terms of growth,

right? On the flip side, if you're using

it for personal business or personal

stuff, you may want to set up a CFO. You

may want to do multiple things in order

to delegate or offload specific tasks,

right? So, one of the first things that

we do for Raz, which is my agent here,

is we do a daily agent ops backup. So,

this is a private repository on GitHub.

And the reason I think this is important

is because everything is contained

inside of Grockbot. So, if you if the

house were to burn down or if Grockbot

just didn't become a fit anymore, I

mean, Grockbot is pretty amazing, but

then even when Hermes and Open Claw came

out, usually you're just like, "Okay,

this is the greatest thing ever. Like,

how could it possibly get better?" Well,

I don't know. But Grockpot, you wonder

the same thing. How could it possibly

get better than this? Right? Well, I'll

tell you, you're building a real asset

here. And what I mean by asset is you're

building actual resources to do work for

you with connectors and contexts and all

these things. And so I definitely think

it's worthwhile to set up an agent ops

private repository and then have your

agent such as your chief of AI staff

like Eve here actually daily back this

up on all the context that's happening.

That way if you were to ever jump ship

or have anything where you need to just

start from scratch, you don't have to do

that, right? But you want to start to

recognize as a business what are some of

the common workflows or daily routines

if you will on what you actually want to

do right and that's where you would

actually orchestrate from there so the

first thing is get your chief of AI

staff stand up an agent and then ask

them hey these are the operations this

is what I need done how would you deploy

more agents to be specialists and we can

delegate from here on out right and so

just to go over some of the agents at a

high level we have Now Talia who is the

uh CFO right the chief uh financial

officer agent. Then we have Edna which

is director of uh human and agent

resources. Okay. And then we have Forge

which is our product engineering lead.

So Forge has context even though all of

them can code it has delegated context

on specific ways that we want things

done. Then we also have different

channels. So you can now create a

channel. And you know the thing that's

unique about this too is channels really

just gives you visibility. But you also

want to look into something like this.

Like if I were to click on Edna, you can

clearly see that Edna has been having

conversations with Eve. So these agents

transparently talk to each other. So to

actually have a channel, it's really

about your visibility or if you wanted

to actually post something in the

channel for every agent to actually post

in there so that way you can see it. But

other than that, these agents, you know,

pretty much DM each other, but you can

see the conversation. So, I would say if

you create channels, at least in my

opinion, have it be a particular topic

that is going to be productive because

again, any type of inputs that you give

is going to burn tokens. Therefore, your

usage is going to be burned as well. And

the biggest thing that I have found, at

least in using this for the last couple

weeks now, is uh token burn is real. It

definitely is real. In fact, last week,

uh, it's been floating around all over X

on how could we possibly hit usage caps,

even if we're on the max plans like

Cursor Ultra and Super Grog Heavy, which

I have both, and how am I hitting, you

know, usage limits and burning a hole in

my wallet for all these things. Well,

I'll tell you, a lot of routines, a lot

of token burn, all for good reason. uh

but it made me like re- snapshot of okay

what routines do we actually have here

and does it actually make sense to do

that or do I just start delegating

accordingly right the other thing too is

in this type of operation you know you

always don't have to be the first domino

that triggers each and every workflow

this is the beautiful thing about

Grockbot is you can actually standardize

routines and have it ask you based on

that before you actually fire off the

workflow so that's Another way to

actually save on tokens as well is

instead of just doing the thing, it

could actually wake and then notify you

like, hey, this is what I'm looking at.

Something like that, right? So, moving

forward, we have ROS, which is really in

our GTM and RevOps. Okay. Uh, and on

that note, we have your different agents

and then we have different sections. And

if you just rightclick on the agent, you

can actually create different sections

as well, which for me is be able to

delegate and kind of organize uh based

on their actual business function. We

have packed. Uh packed is a closer and

partnerships agent. So any transcript uh

that it pulls such as from meetings and

things like that uh it'll actually

automatically bring it into context and

then it'll maximize the opportunity

whether it went well what are some

things that I need to improve on what

are some things in order for me to you

know maximize the opportunity and really

just to make a an ultimate decision on

how I want to move forward or not. Okay.

Then we have Baymax which is top cow

specialist. Uh that's another SAS

product that we have in order for us to

just do outbound just per usual. Then we

have Marlo which is our chief marketing

marketing chief marketing officer.

Tongue twister there uh for both paid

and performance. Then we have pulse.

Pulse is really just keeping a good

pulse on the business, right? So we have

a revops scoreboard. Now you may ask

yourself what does that actually mean?

Well, it's terms of pipeline. What's

coming through? Uh who do we need to

nurture with? what are some

conversations that I've had uh in the

past and maybe I need to follow up with.

Then we have signals. So we have scout

which the name speaks for itself. It's

just really making sure it's scouting

out and making sure that we are evolving

our offer and providing what we do uh

for top of funnel. Then we have radar

which is having a good pulse on what

latest LLMs are coming out or especially

in the AI space uh so we can stay sharp.

We have Optimus which is around social

signals and I originally had this as a

Hermes agent uh monitoring X but uh it

does a great job at monitoring not only

X but LinkedIn and Facebook and things

like that. Then we have operations,

right? Pretty simple, pretty standard.

But we have postmaster, which is

actually connected to multiple uh

inboxes. And when you set all this up,

you're going to set up, you know, based

on connectors and CLI tools, which we'll

go over here in a second. But Postmaster

is able to really just monitor all of my

inboxes, at least the highle ones that I

want actually monitored. And then it'll

actually talk to other agents like you

know like Talia my CFO and making sure

that things are kosher and running

smoothly. Then we have yard. Yard is for

fleet ops and spend governor. So, one of

the things that Eve recommended and I

thought this was actually pretty good is

yard will actually monitor, you know,

for example, if we get a Stripe

notification for a charge or let's say

we have a particular tool that it has

installed and let's say it's not using

it or let's say over the past 3 months

if the charge went through and you know

we didn't use it or let's say that

there's another tool that's you know a

little bit better. it'll actually flag

me to say, "Hey, just letting you know

like you haven't used this tool. It's

costing you x amount of dollars or

whatever." It actually just governs it

and make sure that you're actually using

the tools that we're paying for. And you

know, with with the GTMs and Outbound

and all these things, sometimes you want

to, you know, pivot and test new tools

and things like that. So, this actually

has been super helpful to be able to

like, hey, you know, are you using this

tool or not? Just put on the chopping

block or consider uh moving resources

there. Then we have max which is I don't

know why it said send identity but this

is really just to monitor uh

infrastructure such as inboxes and

domains uh just to make sure that

deliverability is good and if we need to

trigger a placement test or have one of

my employees uh consider uh migrating to

a different infrastructure we can do

that. We have the 820 roundup which I

think was pretty awesome because Eve

came up with that. So you can see right

here, weekday fleet roll up to Brandon.

And it's really cool because at 8:20,

you know, that's kind of an odd time.

And Eve said 8:20. This is good because

8 a.m., you know, usually like you get

to work or you have something like at

8:00 a.m. 8:30 could be like a meeting.

So 8:20 is like that off hour where it's

like, "Oh, hey." And Eve insists like

8:20 roundup. Let's keep that as is. All

right. And then we have Amara, our

client delivery account manager. One of

the things uh that Mar actually does is

monitor clients. So in such a way that

just for performance can do daily checks

across different clients within

instantly and hey reach and things like

that. So with Amara you have your

account manager just like any person

right moving into content we have real

which real is a little interesting

because real uh can actually generate

videos. Now coming from film school for

myself it's just not quite there but

it's actually not bad either. So, you

wonder like, okay, keep trying it. You

know, it's close, but maybe it's all in

the prompt, right? So, real um was a was

an agent. I'm like, you know what? Let's

give it a shot. I mean, we can even just

look into this and you could see right

here, create a new video on your

capabilities. And then if I even so much

as do this, I'm like, well, I mean,

that's some good Ken Burns or zoom in

like, eh, it's all right. You know what

I mean? But, uh, I digress. So with

real, you just never know what you want

to do and how much things are actually

going to involve. So on Luma, we have a

content producer. Now Luma is actually

pretty legit. Luma understands my voice.

It has looked in emails. It knows my

LinkedIn. It knows X. It knows all those

things in order for us uh for me to be

able to actually generate content that's

of interest where I can just literally

brain dump what I'm thinking about. it

can formulate in a draft in such a way

that makes sense. Now, a lot of times

I'll have to refine it and make sure

it's actually human, you know what I

mean? Because they're just not there.

And frankly, I like to make sure I, you

know, put some thought into actual

posts, uh, versus just, you know, AI

slop or anything like that. So, uh, Luma

is that producer. And then we have a

content channel, so I can tell the whole

content team. And again, Eve is, uh, the

robot in charge. Then we have another

app that I'm actually developing uh

which is Goat Ffit. And this is really

just to continue to dog food and um you

know basically improve it from there. So

uh some things I want to call out too uh

is plugins. So you'll notice these are

actually all MCP servers uh which are

also known as connectors and

[clears throat] your account from Cursor

if you started with Cursor will actually

automatically transfer over to Grockbot.

So, it's going to understand full

context of all of the things that you

have like your Gmail, your GitHub, uh

your notion, all that kind of stuff. So,

definitely encourage you to uh consider

connecting what you need. And then with

that, I actually have it connected to

several inboxes that I like to have

monitored for Gmail, for example, and it

will monitor the inbox and literally

keep it clean. The other thing that's

really cool too is it'll actually

communicate with, for example, Telia,

which is my CFO, and be able to, you

know, manage and just kind of know like,

hey, what's coming in, what's going out,

that kind of thing. And it's it allows

you to be able to really just manage

things a whole lot more effectively. And

on the connector front, uh, if your

connector is not actually listed, let's

say you're using GTM tools, well, I got

you covered because GTM as you know, it

is hot and there's a lot of things. So,

if you go to topfunnel.com

and go to resources, uh you'll actually

see a bunch of agent tools that we

actually developed, uh that I, you know,

open sourced and everything and things

like LinkedIn or hey reach uh we have

ClickUp, we have instantly, we have Lean

Magic, Ocean, all that kind of stuff

where you can literally just copy it and

then put it into the agent and say,

"Okay, install this." On top of that, I

actually also made an open source repo

called Grockbot for GTM. And this is

basically a rolodex of all the things in

the CLI tools that are from that site

and to actually give your agents like,

hey, these are these are tools to

consider. Uh, do you have a subscription

to them? Maybe it's something you can

actually install. So, if you feed your

agent to this, be like, hey, I want to

set up outbound. These are some of the

things. Help me install it. And kind of

get you started from there. So, as I

mentioned, you want to start with at

least a chief of AI staff and then

orchestrate accordingly into different

functions every business if you're

managing humans and agents especially.

This is the same organization chart.

Nothing changes. So, I'm going to talk

through some outcomes as we go and when

we dive into it and then I'm going to

pull back the curtain on how we actually

run this at top of funnel which is our

agency. With that, let's go ahead and

start with the live fleet. So, you want

to definitely um get your agency all set

up in terms of different sections. Uh

so, be able to, you know, have different

niche uh areas of the business like

operations, content, that sort of thing.

So, I think a lot of this stuff will

fall into place. And again, all you need

to do is just go ahead and rightclick

and then you can do sections uh which if

this expands, you'll be able to see a

lot of that stuff. So, if we hop over to

RAZ, which is one of our outbound tools.

One of the cool things that I like on

setting this up as as an actual agency

to handle multiple things is, you know,

I noticed you um it says you have

instantly in CLI installed. Is that

correct? Said, "Yeah, we have it

installed." And then what I did was I

took the instantly CLI um and I went

ahead and set it up for agency use,

meaning that we can assign flags to a

particular thing. So let's say you have

client A uh and then you assign an API

key to that particular client. What

it'll actually do now is instead of

having one workspace, it could actually

flag through the profile with that API

key to load it and then it will load the

workspace and verify the actual ID of

that. So that way we're not cominging uh

you know agencies and workspaces and

things like that or clients rather. And

then from here, uh, the actual cool

thing, too, is we're just going to say I

need to add a another client to the

instantly CLI. So, pull up, uh, the API

key where I can enter it. And we'll just

see what that looks like. And the

interesting thing that I really like

about it is you can actually securely

just paste the API key right in the chat

and not have to actually, you know,

worry about inference like on Hermes or

OpenClaw or have to go inside the

terminal or the setup of the

configuration. It just becomes kind of a

hassle. Uh so this is what I really like

about Grockbot on being able to actually

load and you can see it right here. So

we're doing profile demo and then we can

literally just paste an API key and then

we can save that securely. So being able

to configure MCP and CLI tools to

actually set up for agency use is huge

and we've been using that already. Uh so

that way they don't get these actually

confused. So so when you're setting it

up for agency use you're literally you

can assign an agent which can understand

full context around the actual business.

it can understand uh which API key or

demo excuse me uh which profile you

have. So anytime you're doing an actual

CLI call the agent what it does is it'll

actually do a command such as you know

instantly and then it'll have profile

and then we'll say you know client A or

something like that and then it'll

actually assign it. So that's way you

can actually create a campaign like with

copy step one, step two, step three,

that sort of thing and being able to

actually do the whole process. So if we

go over to instantly just like I showed

previously uh in another video, this is

Grockbot campaign. I did not set any of

this up. I just said, hey, create a

campaign out the gate with demo and it

actually, you know, properly formats the

first name, which this is camelc case.

Uh so yeah, it's in double curly braces.

And then we have proper formatting

there. Waiting zero days. That needs to

change. So this is one of those quality

things you want to make sure. And then

step two is also properly formatted.

Again, we would probably need to tell

Raz no m dashes cuz AI. Uh and then last

note. Okay, all good there. So you can

see literally end to end where Grock bot

is literally executing on the things for

you. I suspect in due time with the rate

that's going on right now that we'll be

able to actually do real-time voice

which you can right here. Uh but also

similar to like live chat of being able

like a Jarvis to say hey understand

context around this customer or this

prospect research their Facebook uh or

LinkedIn uh watch their videos on

YouTube which you could actually do that

it'll actually pull it up on the

computer. And then to expand on the

actual computer piece, again, this has

their own virtual machine, which you can

actually interact and do a whole lot of

things here. So, we have Google Chrome.

We have a file explorer uh for

everything that's in there. And then we

have the good old terminal to be able to

execute a lot of things. And if it's

struggling on any sort of thing that

you're maybe logged into, uh you can

actually just click right here and teach

it a task, which is pretty freaking

awesome. So, I would say just keep, you

know, sharp on routines. Uh, so that way

usage is not getting crazy. Um, in terms

of being able to monitor your actual

usage, right? I ran into that actually

quite a while ago uh where I racked up

uh quite the bill on the Grockot usage.

One of the biggest things that you can

actually do, and I'm surprised it

doesn't exist now, but there are several

ways to actually track your usage, which

obviously you can click on this, but the

agents are actually not aware of their

actual usage. You can use another tool

called Codex Bar, which I'm personally a

huge fan of, uh, cuz you can see usage

right there of Crockbot. But one key

thing that I actually built in order to

give agents to make them self-aware of

usage is an open-source CLI tool that I

just made called Grockbot usage CLI. And

what this does is it actually has you

log in uh using Grockbot and then it

will take your cookie and then it will

actually allow it to execute commands.

So we just say what is my current usage

for Grockbot right now. And the good

thing here is that it will actually know

usage that it has uh to be self-aware.

So that way you can actually use the

agents to be like, hey, you know, we're

already at like 50% usage and we're like

less than halfway through our week. So

what it could actually do is be like,

all right, we have a set budget and if

we're exceeding the budget ahead of

schedule, we need to take a look at

pausing some routines. We need to see

what we need to do in order to stretch

this out maybe a little bit or anything

like that. So you can see right here. So

Grockbox weekly is 13% used which is

accurate. Yep. Perfect. And then we have

the reset time of September 1. That is

up 10% from 154. Still the loose meter.

Ultra was tight one earlier. Okay. Now

we're self-aware of knowing like, hey,

if it's skyrockets to like 80% and we

know the reset time is, you know, not

for another 6 days, we need to take a

look at like what's going on and go real

lean, right? So, think of it like giving

a child like an allowance. They want to

know how much they're spending so that

way they're more self-aware. So, this is

open source and free to use. So, link

will be in the description. And if you

run any issues, please let me know. Uh

we can get that resolved. But, uh it's

been working great for me. And I think

it's just another essential thing for

your fleet to know like what kind of

tokens you're burning. If you want to

see the system behind running a business

on agents, go ahead and check out this

video right here. That's where we break

down how to actually build the agent

system that makes the business easier to

run. If you reached up to this point,

really appreciate you watching.