The #1 AI rule

I sent out more than five hundred personalized speaking outreach emails the previous week. It took me about twenty minutes of active time.

I uploaded my target list and my project documents into Claude Cowork. It drafted every email in my voice, personalized to the event and the organizer. I spent my time lightly reviewing them in Gmail, not writing them.

That's what infrastructure looks like when it's actually built. Not talking about leverage. Using it.

The file, the Project, and the method

There's a framework I use for everything I build. I call it the Connector Method™. It's four steps: fast action, evidence, confidence, results. You take the action before you feel ready. You look at what happened. The evidence builds confidence. The confidence produces results. Then you loop.

Most people stall at step one because they want to understand the whole system before they build any of it. That's the trap. The system reveals itself through the building. Not before.

I'm telling you this because what I'm about to ask you to do can feel like a lot for one afternoon. It's one file, one Project, and one hour. And everything you produce with AI after this will be different.

Here's what's happening for most service-based business owners using Claude or ChatGPT. They open a new conversation. They re-explain who they are. They re-describe their audience. They re-paste their tone. They wait for an answer that sounds mostly like them. They copy what's usable. They close the tab. Tomorrow, they do the whole thing again.

Every session is a first date with a tool that has enormous capability but no working understanding of you.

A Project in Claude or ChatGPT gives the work a persistent place to start. You load your voice, your frameworks, your brand rules, your templates, and your context. Every conversation inside that Project can start from a fuller picture of who you are and what the work requires. You stop re-explaining yourself. You start getting work done.

The most important thing that goes inside your Project is your voice-reference document. This is the file that makes the difference between output that sounds like the internet's average voice and output that sounds like you.

Here's what belongs in it:

• Your tone and how you write.
• Your no-list, including the words and phrases you never use.
• Your frameworks and named concepts.
• Your values and worldview.
• Two or three of your best outputs as quality benchmarks.

You don't need to type all of that from a blank page. Record a five-minute voice note. Talk about how you communicate with clients. What you sound like when you're being real. What you refuse to say. What you believe about your industry. The phrases that are yours. Transcribe it and use the transcript with your best writing samples.

Apply the Connector Method™

Fast action. Block an hour. Record the voice note. Transcribe it. Paste it into your AI tool with your best writing samples. Run the prompt. You have a voice-reference draft before lunch.

Evidence. Take the voice reference you just wrote and drop it into a Project. Run a real repeated workflow inside that Project. Look at the output. Compare it with what you were getting before.

Confidence. The output sounds more like you because you gave the system something concrete to work from. That's the evidence that the foundation matters.

Results. Every piece of content, proposal, and client email you produce through the Project has a stronger place to start. You built the foundation once, and now you can keep refining it as the work changes.

That's the sequence. One file, one Project, one afternoon. Then Skills have context. Agents have context. The foundation makes everything else work.

Prompt to build your voice reference

After you transcribe your voice note, paste it into your AI tool along with three to five of your best emails, posts, or newsletter issues. Then run this:

Here's a voice note where I describe how I communicate, and here are five examples of my best writing. Analyze everything and produce a voice-reference document that covers my sentence-structure patterns, words and phrases I use repeatedly, words and phrases I never use, my tone and register, how I open and close pieces, my values and worldview, and any named frameworks or concepts I reference. Format it as a document I can load into an AI Project and refine over time.

Edit it until it's actually you. Then load it into your Project. That's the foundation. Everything after this starts from your context instead of the internet's average.

If this is your kind of conversation, join me for the next Seed & Society® newsletter.