For a long time, I thought I was using AI well.

I'd bring it a business idea. It would tell me it was brilliant. I'd bring it a decision I was second-guessing at 11pm. It would validate every point I'd made. I'd walk away feeling good.

And then I'd lie awake wondering why it still felt like I was making decisions alone.

The problem wasn't a lack of information. I was giving it context. It just didn't do anything with it. Everything was "yes, go for it, it's a great plan." Agreeable. Enthusiastic. Completely useless when I actually needed someone to push back.

What I needed wasn't a smarter prompt. I needed a system that applied what it knew about me - and could tell me when I was drifting from what I'd already decided.

This post is the written companion to my Hometour video, where I walk through my full AI system in real time. If you'd rather watch, start there. If you want the written breakdown to refer back to, keep reading.

---

What does "AI team" actually mean?

Most people use AI like a vending machine. Prompt in, output out. The result is fine. Generic, but fine.

An AI team is different. It's a structured system - files, skills, and a foundation of context - that means every time I open a session, my AI already knows me. My values. My voice. My current priorities. What I decided three weeks ago and why I decided it.

The simplest way I can describe it: imagine the difference between cooking in an Airbnb kitchen and cooking in your own. In the Airbnb kitchen, everything you need is technically there. But every visit, you're finding the knives again. In your own kitchen, nothing needs explaining. You just cook.

That's what I built.

---

The filing system: where everything lives

The first thing I set up was a folder structure. Not glamorous - but it's the skeleton that makes everything else work.

Before this, my content ideas, launch plans, and business decisions lived across a dozen different chats, files, and my own head. Now everything lives in one structured file explorer - organised into folders, just like a regular computer.

Take the Post Office as an example. That's my content folder. Inside it:

- Platform queue files (one per platform - what's going out this week, plus a running Idea Bank for future posts built right into each file)

- The Archive (everything posted, filed and searchable)

My AI can reference any of these files at any time. Nothing disappears. Nothing lives only in my memory. It's a simple thing that changes a lot.

---

The foundation: why context is everything

The "fresh context" problem is the reason most people feel like their AI content sounds like everyone else's.

New chat. Back to square one. Twenty minutes re-explaining who you are before you can do any real work.

I solved this by building foundation documents - files I created once that my AI references every session:

- About me (who I am, how I work, what makes me tick)

- My business context (offers, ideal client, where I am right now)

- My values and non-negotiables

- My Human Design profile (I'm a Splenic Projector - this shapes how I work best)

- My voice (specific words I use, specific words I avoid, the stories I come back to)

Feed this in once. The AI doesn't just remember - it learns. Over time, it starts to feel less like briefing a stranger and more like working with someone who's been around long enough to know what you actually mean without you spelling it out every time.

---

The voice skill: content that actually sounds like you

Generic AI content has a particular tone. You can sense it - enthusiastic in a slightly hollow way. Polished in a way that doesn't quite fit. My audience would spot it immediately. And I spotted it in my own early AI drafts and it made me cringe.

My voice skill changes this. It teaches Claude my specific anchor stories, my platform formats, and my phrasing preferences - down to the level of "children, not kids" and "organised, not organized." When I give it a messy voice note transcript, even a rambling one, it produces content that's 80% ready. Not perfect. But mine.

---

The part I didn't expect to value most: strategic pushback

Standard AI is a yes-machine. You bring it an idea, it gives you a full plan. It doesn't know what you've already agreed to. It doesn't know what you're building toward. It just says yes.

My system is built to push back.

Earlier this year, I was mid-spiral about rewriting a significant part of my course based on some feedback that had rattled me. My AI stopped me. Not harshly - but it flagged that the decision was reactive, not strategic. It referenced what I'd already decided and why. It was right. I didn't rewrite the course.

That's not a chatbot. That's a thinking partner.

---

Do I need to be technical to build this?

No. I genuinely cannot write code. I built this entire system using Claude Code inside VS Code. Nothing that requires a technical background.

The Hometour video shows my exact setup in real time, including how the files are structured and how the skills work.

---

Common questions

What is the difference between Claude chat and Claude Code?

Claude chat is the familiar conversation interface. Claude Code lets you build files, create custom skills, and connect to your own folder structure - so your AI isn't just remembering things, it's working inside a system you've designed. That's what makes the team possible.

Do I need coding skills to use Claude Code?

Not to use it day to day - everything runs through normal conversation. The initial setup requires a few technical steps, but once it's done, no coding knowledge is needed.

How long does it take to build?

The core foundation took me one focused weekend. The system grows from there - every session adds a little more.

Does this work with ChatGPT?

The principles apply broadly - give your AI context, build a foundation, work toward something that persists rather than starting fresh each time. But the tools matter too. The file system, custom skills, and the way everything connects is built in Claude Code, and ChatGPT doesn't replicate that in the same way. The mindset translates. The full system doesn't.

Why Claude Code instead of Claude Cowork?

Cowork is excellent for connecting to tools you already use - Gmail, Notion, Slack. Claude Code is for building a system shaped around how your specific business works. One fits you into existing structures. The other lets you build your own.

---

Watch the full Hometour video here.

And if you want help figuring out where to start with yours - DM me.