I Set Up Claude Code for 20+ Businesses. Here's What Happened.

Over the past couple of months, my team and I have set up cloud code environments inside of over 20 companies consisting of law firms, agencies, property management companies, healthcare clinics, home services.

They're all different industries, all different sizes, totally different levels of technical ability. And

honestly, some of them we messed up early on, but after doing this so many times, certain patterns we found kept showing up that I think anyone running a business right now needs to hear. So, in

this video, I'm going to be walking you through what we actually found inside of these companies, what we actually built, what worked, and some stuff that I wish somebody had told me [music] before we even got started. Now, this gap between using AI and actually running your

business on it, it is way bigger than most people think. So, let's get into it. Now, before I actually get into

it. Now, before I actually get into these systems or any results, let me just paint a quick picture. Now, every

company that we walked into, whether it was a law firm doing $5 million a year or 12 person marketing agency or even an $8 million property management firm, they all looked basically the same when

it just came to AI. So, somebody on the team, they had tried CHBT. Usually, it

was the founder, maybe they'd used it for emails or brainstorming, and they liked it, but nothing had actually changed in how the business was running day-to-day. Now, overall, the team, they

day-to-day. Now, overall, the team, they were still doing the same work the same way. [music] So, the client onboarding,

way. [music] So, the client onboarding, it still took 3 days, and it had six handoffs. the reports, they were still

handoffs. the reports, they were still being compiled by hand every single week. One property management company,

week. One property management company, they had three people spending about 15 hours [music] a week just triaging maintenance requests from their email, logging them into their PM software manually, and calling vendors to [music]

just dispatch. They were handing quite

just dispatch. They were handing quite literally every request by hand. And

that 15 hours a week number, it isn't that unusual either, where property managers, they typically lose about 20 to 30 minutes per work order when they're actually doing it manually. A

midsize portfolio racks it up pretty fast. And for agencies, the industry

fast. And for agencies, the industry benchmark for any billable utilization, [music] it sits around 70 to 60%. Which

means that 20 to 30% of everyone's time, it is going to admin and internal staff that doesn't make money. Now on a 12 person team, [music] that is effectively like having three or four people worth

of hours just gone every week. And

everyone that we talked to, they knew that AI mattered. They really knew that it was the future, but nobody actually knew where to start. And the ones who had tried, they tried the wrong way. So

we had to figure out why that was. Why

does every founder play with Chetch for a few weeks and then just go back to doing everything manually. Now after

about 20 of these the answer was very clear. So let me give you the framework

clear. So let me give you the framework that we use now because this is going to apply whether you work with someone or you try it yourself. So what we've done is we've broken this down into map,

foundation, build three, scale up and compound. So this is five steps. I would

compound. So this is five steps. I would

screenshot this if I were you because this is the playbook that we use for every implementation. So step one, this

every implementation. So step one, this is map. So before you touch any AI

is map. So before you touch any AI tooling, you just need to figure out what to actually automate. Now, most

people, they get this completely wrong because they're simply just picking on what sounds cool instead of what actually hurts the most. So here's how to actually do it. First, just grab a spreadsheet and write down every single

workflow that your team does regularly.

So whether that is weekly reports, client onboarding, invoice processing, or even data entry, follow-up sequences, whatever that may be, you just dump everything. [music] And then from there

everything. [music] And then from there you were scoring each one on three things. How many hours per week does it

things. How many hours per week does it actually eat? How directly does it

actually eat? How directly does it impact revenue? And how feasible is it

impact revenue? And how feasible is it to actually automate right now? You

score each one of them from 1 to five.

[music] Then from there and you add them up and you rank them. So your top three, these are what you're going to be building first. Now we call this an

building first. Now we call this an automation priority matrix and this takes about an hour if that. But of

course it depends on size of the team.

And this also just prevents the most common mistake that we see across every company, which is building some fancy automation that saves 20 minutes a month while the team is still bleeding 15 hours a week on something that should

have actually been first and prioritized. Seriously, like, pause the

prioritized. Seriously, like, pause the video right now and go do [music] this.

It is the most useful thing that you can do before any AI setup. Now, after doing this with this specific offer with over 20 companies, the [music] top three, they almost always fall into the same

buckets that consists of intake and onboarding, which means that whatever happens whenever a new client or lead comes in. Then there's the reporting and

comes in. Then there's the reporting and data, meaning all the stuff that your team is manually compiling on a schedule. And then there's

schedule. And then there's communication, meaning that follow-ups or status updates or all the notifications, you know, the stuff that isn't hard, but it just needs to happen consistently. and it just eats time

consistently. and it just eats time across the whole team and organization.

Now, step two, this is the foundation.

So, this is the step that most people actually skip and it's why their AI is giving them generic outputs. Now, the

biggest mistake that we saw was founders jumping straight into building automations without giving any context to AI about their business. [music] So,

if you actually think about that, it's literally just like hiring someone and never onboarding them. So, of course, like the output is going to be mid. Like

there's three pieces to the foundation.

First, that's your claude.md file. Now,

this right here, it is just going to live in the root of your Claude code project. And this tells Claude

project. And this tells Claude effectively everything about the company. [music] But here's a practical

company. [music] But here's a practical tip with this is you do not just put your company description in there. Like

that is what everybody does and it barely helps. It's actually detrimental

barely helps. It's actually detrimental at times. What you do is put your actual

at times. What you do is put your actual conversations. So how you actually name

conversations. So how you actually name files, what your text stack is, how you want things formatted, how your team is communicating with clients, your brand voice, you know, things like claude should never actually do in your

workspace. The more specific [music] and

workspace. The more specific [music] and actually opinionated you are in the file, the less that you actually have to spend correcting [music] the output later. Second, this is the memory. So

later. Second, this is the memory. So

quad code and open claw they can persist memory across all other sessions which means that decisions [music] that you had made even last week last month you know files that you [music] worked on

and preferences you've set all of that is going to carry forward. So if you're starting from scratch every single time that you open a session you really do not have [music] a system you actually

just have a simple chatbot if that with extra steps. And third this is

extra steps. And third this is connecting your actual tools. So you can think of your CRM, project management, email, [music] analytics, whatever and wherever your team lives in. So MCP

integrations, [music] they are essentially how you actually do this.

Now with the introduction of CLI, [music] very new concept and approach.

MCP, this is for the most part how you actually go about this. And this sounds technical, but it really just means that you're plugging things in. Now, most of this, it takes under 10 minutes. [music]

So once the foundation is actually in, that's when the AI actually stops being generic and starts truly understanding your specific business. [music] Now, one of our clients, they run a clinic and we set it up so he can actually just query

a medical model with his images from his practice and the answers they save right back into his workspace. Another client,

he had his automation platform just wired in through MCP. So he can spin up workflows and he just pushes them into production quite literally in seconds.

[music] and that kind of thing. It only

works when the foundation is really rock solid. Step three, this is what we call

solid. Step three, this is what we call build three. Now, this is where you just

build three. Now, this is where you just take the top three things from your priority matrix and you're building automations for them. You're not doing 10, you're just doing three. And the

temptation here is it's going to keep you going. Like you finish the first one

you going. Like you finish the first one and you're feeling good and you want to do five more because you are open to the world of possibilities. You don't. You

just get three solid ones. You get the team to use them. you make sure that like everything actually works in production and then you can expand from there. Like we've had clients try to

there. Like we've had clients try to rush past this and it always just creates more problems than it actually solves because nobody on the team has the time to actually learn the first

ones before new ones keep getting added.

Now the selection criteria this matters a whole lot too. So this is where you just pick things where the team actually feels a difference within the first week. It is not the first month it's

week. It is not the first month it's just the first week. So if the first automation, let's say hypothetically it saves your operations person like 5 hours and everyone on the team pays

attention like that creates momentum for everything that is going to come after.

So hypothetically if it takes a month for anyone to notice like you probably picked the wrong thing and that's where you want to go back to the matrix. Now

I'm going to show you two real examples of what these actually look like for our actual clients in just a second. But

before we do that let's just talk about step four which is what we call the skillup. Now, this is just turning

skillup. Now, this is just turning one-time projects into something that actually sticks. [music] So, once your

actually sticks. [music] So, once your first automations are running, you just start writing skills for pretty much everything that your team is repeating.

So, client proposals [music] every single week, that is a skill. Weekly

reporting, you turn that into a skill.

So, the specific way that you onboard clients, that is a skill. Each one takes 15 to 20 minutes to write and it saves you that time every time it runs after.

Of course, it'll [music] depend how long it takes you right now. But the real unlock here is actually getting this team adoption. And we learned [music]

team adoption. And we learned [music] this really the hard way where you don't train everyone at once. You just pick one person on the team. You know, the one who is the most curious [music]

or most frustrated with the manual work.

You can call this your AI champion, but you're just getting them set up. You

help them build their first skill for something that they personally do every day. And then you let them show the rest

day. And then you let them show the rest of the team. Now we had one company where the founder he tried to force everyone on it on day one. What we

actually found from this I mean it was just slow adoption lots of resistance.

Now in comparison with another company we worked with one junior person like she got into it and she started showing her co-workers what she had built. She

literally spread through the team in a few weeks on its own. Like it's

completely different outcome. Now step

five this is compound. Now this is the one that most people never reach because they just quit too early or they're just not doing the right things. So the first couple weeks it feels like setup. Like

you're loading the context, you're writing [music] the skills, you're connecting the tools and it really doesn't really seem like much is happening. [music] And then around week

happening. [music] And then around week three, sometimes four, something actually starts shifting. And this is where you stop explaining things to Claude and it really just starts anticipating what you need. [music] So

from here, you're not prompting anymore.

You're just working actually. And here's

what makes this really different from any [music] other AI tool is every document that you load, every skill that you are writing, every workflow that you're automating, it all is just stacking on top of each other. So what

you're working on day 100, it looks nothing like day one. And it's not because the model just got upgraded.

It's just because your layers got deeper and deeper. And when the model does get

and deeper. And when the model does get upgraded, which happens pretty regularly, everything that you've already built, it gets better automatically. And with that, like you

automatically. And with that, like you don't rebuild anything. It just improves underneath you. And most people, they

underneath you. And most people, they just treat AI like something that they use. This is something that you are

use. This is something that you are building on top of. And the longer that you're building, the wider the gap is going to get and the more that is actually going to be compounding for you and your team. Now, real quick, if you guys want to see all five steps, I'm

doing a free workshop March 31st where I'm going to be showing breaking down all of this in much, much more depth.

So, if you guys want to upscale your team, learn how to actually become AI native, AI first, whatever you want to frame it as, link will be down below in the description. We're limiting it to

the description. We're limiting it to 100 seats. But that being said, let me

100 seats. But that being said, let me show you guys the results. First one I'm going to go over is with a property management firm. So, they had about $8

management firm. So, they had about $8 million in revenue. They had a few hundred units where they had three ops people spending about 15 plus hours a week on maintenance requests. So, it

goes a tenant emails in a request.

Somebody's going to read it and then categorizes the urgency. Then it's going to get logged into a PM software and then you just have to look up like which vendor handles that issue for that property. Then you call the vendor and

property. Then you call the vendor and then you create the work order and follows up. So every request every

follows up. So every request every single time what we had to do is we just connected their email inbox to Claude code through an MCP. Now when a request is coming in Claude simply reads the

email. It pulls the tenant, pulls the

email. It pulls the tenant, pulls the property info from their PM software and then categorizes the urgency based on the context. So if something had a water

the context. So if something had a water leak, this goes to a high priority. If

there was a cabinet hinge loose, it goes to normal queue. So you get the idea.

Like it creates the work order, it finds the right vendor, sends the dispatch, all before anyone on the team even touches it. So what this actually went

touches it. So what this actually went to was from 20 to 30 minute manual work orders to under 3 minutes with human review only on the urgent stuff. Now, if

you want to build something like this, the stack is simply just having email to cloud connection via an MCP and then a skill that defines your urgency categories and your routing logic and

API access to your PM software. Now,

once the foundation from step two is actually in place, like this took us less than a day to be building. Like it

is nothing crazy. As [music] a result of this, they moved 12 of those three ops people into tenant retention. Now,

people went from actually just doing data entry to work that actually grows the business. and their ops manager.

the business. and their ops manager.

They called this the highest RARI project that they had done in about 5 years. And it was just one automation.

years. And it was just one automation.

It was one workflow [music] directly off of the priority matrix, the 12 person marketing agency. So a founder, they

marketing agency. So a founder, they just told us upfront he did not want to depend on us long term. And that's

completely understandable. He wanted his team to learn it. So this was just a step for play, also called the skill up.

So, we showed them how to actually set up Claude code, write their claw.md file

for their agency, build their skills for their client workflows, and then we actually connected it to their reporting tools, all their analytic platforms, their content management system. Now,

the first skill that we actually built together was just a simple client report generation. So, what actually happened

generation. So, what actually happened is it just pulls data from Google Analytics and their ad platforms and then it just formats it into the brand template. It writes the summary. And

template. It writes the summary. And

this used to take a strategist 45 minutes per client per week. Now it's

about 3 minutes of review and within 6 weeks like their junior strategist they were building things on their own. One

of them even made their own content repurposing system and that takes a long form blog post and breaks it into two social posts for four different platforms and then adjust the tone for each individual one and it cues them all

up. Like we didn't even plan that. She

up. Like we didn't even plan that. She

just understood how skills actually worked and they just ran with it. Even

billable utilization it went from 60 to over 85% and industry average it's about 70 to 80. So, they went from below average to well above it. Now, on a 12person team, that's roughly three

full-time people worth of hours recovered every single week without even having to hire anybody. And then, of course, there's us. We'd be hypocrites if we were not running this ourselves, right? So, I ran three companies.

right? So, I ran three companies.

Personally, when I first set up Cloud Code for myself, honestly, like it was fine. I saved some time here and there.

fine. I saved some time here and there.

It was nice to have, but nothing that I would necessarily call life-changing.

And then a few weeks in, I hit step five and something just completely shifted.

It was like this paradigm shift where it's hard to describe, but I'll try. So

I just stopped being the person explaining things to Claude and I started just working alongside it if that makes any sense. Now it already knew my business, all my preferences, my

team, my standards, the decisions that I make. And with that, like the friction

make. And with that, like the friction little by little just started disappearing. So, I went from actually,

disappearing. So, I went from actually, you know, spending 4 hours a day on operational work to around 11 minutes.

Now, I know that sounds insane. It's not

a typo and it didn't happen because I just had some master prompt engineer. It

literally only happened because I stopped prompting and I just started building layers just [music] over weeks and at some point those layers, they just became deep enough that Claude

could just do the work for me. Now, what

that actually looks like, every single day, for example, I get a morning briefing right before I'm out of bed, you know, I get revenue across all three businesses, team updates, anything that looks unusual, you know, priority

actions for the day. Like, that used to be an hour of me jumping between dashboards, looking through Slack channels, having to actually respond to all my emails that are just menial things that I had to respond to. Now,

it's just a document. All those emails go out. It's all waiting on my phone

go out. It's all waiting on my phone when my alarm goes off. Now, we use the same system to plan and execute a webinar as well. This did multiple six figures in sales just last week. Now,

I'm talking offer analysis, content strategy, the full email sequence, even the funnel build. Like, that's a quarter of work for most teams. Like, we did this in a fraction of the time because the system already had all of our

business context. Now, I can't even say

business context. Now, I can't even say whether or not like we would even fathom being able to build that entire webinar end to end if we did not have AI, you know, this particular AI system set up.

And this also wasn't starting from scratch. Like this was pulling from

scratch. Like this was pulling from everything that we had already loaded into it. So it knew our business. And

into it. So it knew our business. And

this video that you're watching right now, the research, the outline, the supporting data, all of that was assisted by the system pulling from our knowledge base in real time. Like I

don't even think about it as AI anymore.

Honestly, it is just how the business actually just works now. And that is what step five is. Now here's a few lessons from doing this over 20 times.

This is all learned the hard way. all

from actual real implementations within all of our clients. Number one, safety is not optional and almost nobody is talking about it as much as they should be. Now, claude code like it runs on

be. Now, claude code like it runs on your machine with your permissions, right? So, with this, [music] it can

right? So, with this, [music] it can read your files, write files, run commands, it can hit your APIs. That is

what makes it powerful, [music] but it's also what makes it dangerous if you are not setting boundaries. So you have to clearly define exactly what it can what it cannot do before you actually let it

run anything. So which files it can

run anything. So which files it can access, which commands it is allowed to execute, what it should never touch, what it should touch, and just start with everything locked down. And then

from there, you can slowly loosen as you actually build trust with it. And you

also want to audit this weekly so you can just check like what [music] it's been doing. We even had a couple

been doing. We even had a couple situations early on where it took actions that we didn't anticipate because we hadn't [music] even been specific enough about all the guard rails. Very dangerous, but nothing

rails. Very dangerous, but nothing catastrophic likely happened, but enough to actually just make us take this a lot more seriously. Number two is team

more seriously. Number two is team adoption is a people problem, not a tech problem. Now, the technology works over

problem. Now, the technology works over half of cloud code usage at companies like Epic, even the healthcare [music] tech company, like this comes from non-developer roles. So people who are

non-developer roles. So people who are not inherently engineers, they use this daily in organizations that have adopted it properly. But with that being said,

it properly. But with that being said, like getting your specific team to actually use it, [music] it really requires a champion. Like I mentioned earlier on the inside, not just a

mandate from the top. So one person, one quick win. Just let the enthusiasm

quick win. Just let the enthusiasm actually just spread naturally. So, I

already talked about this in step four, so I won't repeat it. But just know that if you roll out all of that, if your roll out strategy is just everyone start using this on Monday, it is going to fail. You will not build well. We've

fail. You will not build well. We've

watched this happen time and time again.

Number three is that the compounding is real. But most people, they just quit

real. But most people, they just quit before they actually feel the results.

Now, I keep coming back to this because it is truly the number one reason that most people abandon the whole thing. So,

the first couple of weeks, they are a lot of setup work for what feels like a pretty small return at the beginning. It

genuinely does feel like you're doing a bunch of work for not much, but it's like compound interest, boring and slow, and then just one day you just wake up and you can't believe the difference.

So, in our experience, a threshold is about 3 to four weeks of consistent building. It totally depends on your

building. It totally depends on your team size, how your organization is structured. But if you can get through

structured. But if you can get through that, whether it's two weeks, three weeks, or a few months, if you can get through that, you will feel it. Most

people, they do not make it that far and they blame the tool and then they totally are just going to get beat out by their competitors who are going to figure this out before them. So don't be that guy. Now, number four, to be

that guy. Now, number four, to be honest, the windows closing. Deote's

latest report, they say about 82% of companies haven't even started an AI training strategy for their [music] employees. like 82% right now almost

employees. like 82% right now almost nobody is doing this at scale for mid-market businesses. Now this is going

mid-market businesses. Now this is going to change in the next six to 12 months undoubtedly and as awareness is going to catch up and these tools get even more accessible. It's hard to argue that now

accessible. It's hard to argue that now the founders who actually just get their teams running on this while the field is empty they are going to have months and months of compounding layers built up in

the time that their competitors are just starting. That is not an easy gap to

starting. That is not an easy gap to close. Now, the companies that we set

close. Now, the companies that we set this up for, they just look completely different after. And I don't mean that

different after. And I don't mean that in a very vague and general way. Like

the teams, they genuinely work differently. Like the founders time

differently. Like the founders time looks different. Work that used to eat

looks different. Work that used to eat everyone's day. It just runs. And we're

everyone's day. It just runs. And we're

still early. Like models, they keep getting better. Tools keep getting more

getting better. Tools keep getting more and more capable. [music] And everything that you build right now, it just gets better and better with each individual upgrade, which is coming at, you know, every single hour it feels like, which I

think it genuinely is. every single hour they just get a crazy new update. But

with that being said, here's what I would do if I were you. Step one, just map things. Go do the priority matrix.

map things. Go do the priority matrix.

Take an hour. It will change how you think about where AI actually fits in your business. Now, you can do that

your business. Now, you can do that right now. Grab a pen, grab a piece of

right now. Grab a pen, grab a piece of paper, and if you want to see all five steps and how they actually connect, again, we're doing a free workshop on March 31st. You'll get real client

March 31st. You'll get real client workspaces on screen, live automations and building that stuff, the whole framework end to end in a Q&A at the end or you can actually bring your specific business and I'll tell you what I would

do first. So the link to register, it'll

do first. So the link to register, it'll be down below. Again, only having 100 seats, make sure to grab one. And if

you're past the workshop stage and you want my team to actually come in and do this for you, build the automations, train your people, there's also a link below to book in a free discovery call.

We'll look at your business and tell you what we would build and what we would not build. And if you're more on the

not build. And if you're more on the side of just wanting to learn AI and build a business around this, I've got a free training on that, too. Link will be down below in the description. But

that's all I wanted to share with you guys. Thank you guys for watching. If

guys. Thank you guys for watching. If

you found any sort of value, please drop a like, drop a comment. If you're a business owner looking to upscale your team and actually become AI native this next decade, then make sure to sign up for our webinar next week. You will not

want to miss it. But I'll see you guys in the next