Welcome to The Atlas Executive - the newsletter that brings founders and executives the proven frameworks and playbooks to create more leverage through EA and AI-enabled workflows, so they can go from constant firefighting to operating in their genius.

"Colin, my life has changed."

That's what a friend told me two weeks after I spent two half-days teaching him how to build with AI.

He's in the healthcare space. Runs a couple senior care facilities. When we were chatting, he was telling me about how much they spend on software subscriptions and tools… $150K a year. Just in software and tools.

I'm sure you can relate. Our tech stacks cost a lot. There are operational expenses we've all just accepted as normal.

So I asked him a simple question.

How many of those tools can you actually just build yourself?

He has a little bit of a technical background - dropped out of engineering. But he had enough there that I was like, you know what, I think you can probably solve a lot of these yourself.

So I said: hey, let me show you something. We spent two half-days together. I taught him Claude Cowork and Claude Code. Then I left him with it.

I wasn't sure how much he was actually going to apply.

Two weeks. That's all it took.

He built skills in Claude Cowork that created healthcare policies and documentation - the kind required for accreditation.

The same volume of policies his team had taken two years to produce? He did it in four days.

Then he used all of that as the context needed to start replacing his facility's software tools. He built a full production-ready accreditation and investigation software - and to be honest, I don't even know what that means - but in one day, he took an idea, used the context from those policies, and shipped something production-ready for his management team.

One day.

He's mapped out five more tools with PRDs created. He thinks he can cut his software expenses by 90%.

So I encouraged him - go sell this to your competitors.

He called another facility owner that same day. Asked what they were spending on software. $100K. Offered to build something custom for half the price. The answer was immediate - yes, absolutely.

He pre-sold a tech stack he hasn't even built yet. His competitors are still paying six figures for tools that barely fit. He's building ones that actually work - because he has the context they don't.

Think back to how it used to work. You have an idea. You're clear on the problem. And then you need a dev team, a product team, a marketing team - the full operational buildout to go to market.

The barrier was never the idea or the expertise. It was the infrastructure required to execute.

Most executives I talk to still carry that mental model. They see the tools. They hear the hype. But underneath, they're still thinking: I'd need to hire developers for that.

You don't. Not anymore. And my friend's story isn't an outlier - it's the new pattern. The question is just: how do you actually do this?

But tools alone aren't enough.

There's a lot of people out there right now teaching AI. "Just start vibe coding, just start building agents." Not exactly.

The same way you'd set up a new hire for success - there's architecture needed first. You wouldn't bring someone in on day one and expect great output without onboarding, documentation, and clear expectations. The same thing applies here.

I've mapped the core components that make any AI tool actually effective - and every one mirrors managing a human team.

When you bring someone onto your team, they need handbooks, documentation, and knowledge about how the business works. An AI tool needs the exact same thing - a clear system where all of that context lives.

But here's the part most people miss.

It also needs a memory system. The same way a team member goes from B minus to A minus to A plus through corrections and feedback - the AI needs a way to log what it learns and compound on it.

Without that, you're starting from scratch every session.

A great employee has clear playbooks for their responsibilities and a defined routine. Your AI needs the same.

I map out every skill - inbox triage, reporting, content workflows - and define exactly when each one runs. Monday mornings. Twice a day. End of week.

That routine is what turns a tool into a reliable operator.

Everything needs a home. One single source of truth - not files scattered across five platforms.

And just like you'd scope permissions for a new hire - NDAs, access levels, what they can and can't touch - you define the same boundaries for the AI. What it can access. What it can say. Where it reports.

Without clear boundaries, it gets messy fast.

For your business:

For you:

Block two hours this week. Pick one workflow that frustrates you. Open Claude Cowork and say: "I want to build a tool that does X - walk me through it."

Start there.

Next issue, I'll go deeper into what maker days look like at Atlas - one day a week where the entire management team builds together.

Your Pal,

P.S. We're building an AI builders community. If you could join a hands-on training based on what I shared today - building tools, setting up agents, learning the workflow - what would you want to learn first? Reply and tell me. That'll shape what we build.