Train AI to Think and Write Exactly Like You (Even If You’ve Never Documented a Thing)

Two documents go into every chat I open: one that holds how I think, one that holds how I sound. Here’s how to build both in about a month, and the one job you should never hand over.

Two years ago my team exploded in size, and I started getting asked the same questions over and over by different people.

Things I’d already answered. Things I’d written whole books about. Things I had videos on. The answers existed, but the timing never matched. Nobody wants to read a whole book to get one answer. And the person asking usually had context around their question that would change what I’d tell them.

So I asked myself a crazy question. Could I build an AI that answers those questions as well as I would? Maybe better?

Turns out you can. I built a system called Dan AI, and that one tool bought me back tens of hours a week. It only worked because I’d been documenting everything for years.

Here’s what I learned building it. Getting AI to answer like you is the easy part. Getting it to sound like you is where almost everybody fails. We all know what AI writing sounds like. Words like “delve.” Everything happening “quietly.” That rhythm where every sentence sets up a thing, knocks it down, and swaps in another thing. You can spot it from across the room.

I’ve written a best-selling book and I’m writing my next one right now. AI doesn’t write it for me. But for almost everything else, I’ve got it thinking like me, writing like me, and sounding like me. Here’s exactly how.

Why sounding like AI costs you now

One consumer survey asked people two years running whether heavy AI use made them trust a brand less. In 2025, 20% said yes. In 2026, it was 40%. It doubled in a single year. And Gen Z, the generation that grew up with this stuff, punishes it the hardest, at 54%.

People buy from people they know, like, and trust. Take the person out and they don’t know who they’re buying from anymore.

Nobody’s mad at you for using AI. They’re mad because it sounds like AI. They want to know a real human was involved in what they’re reading.

So we fix that in order: build the brain, teach it how you think, then teach it how you sound.

Build the brain first

Before AI can sound like you, it needs something to work with. A container. I call it the brain.

I started creating content 15 years ago with no idea what I was actually building. Fifteen years of my knowledge, my ideas, my stories, my preferences, all sitting there, ready to be mined. If I had to pull that out of my head from scratch today, it would take years of talking.

That’s the part nobody tells you. The value is in years of stacked-up, documented, cataloged thinking.

If you’re starting from zero, go find everything you’ve already made. Old playbooks. Docs. Emails. Files on old hard drives. The more original material you can point to, the better the AI gets.

Then start capturing. Here’s how I do it:

1. Record everything. Every meeting. Every call. Every decision. All of it becomes context for how I think, the frameworks I use, and where my life is at right now.
2. Save your AI chats too. I keep every conversation I have across all my different AI tools. Those chats are you thinking out loud.
3. Pick a capture tool and stop overthinking it. I use Granola for most of my meetings. A lot of people on my team carry a Pocket, a little physical recorder. Others use Wispr Flow and do a voice dump at the end of each day on where they’re at. The tool doesn’t matter. The capturing does.
4. Give it one home base. I use Obsidian. You can use Notion or Google Drive. What matters is that everything gets dropped in one place, because that’s what the AI will reference for everything that follows.

A recent example. My wife and I were driving to a retreat, and I told her to pull out the voice memo app on her iPhone and just record us. We asked ourselves one question: five years from now, what will be true? We talked it through for 27 minutes.

Then we handed the recording to the brain. It separated our two voices, built her task list and mine, and we started executing that week.

The rule is simple. The second you have a thought or make a decision, get it out of your head and into the brain. Your head is a horrible place to store things.

Do this for 30 days and you’ll have more raw material than most people capture in a year.

Teach it how you think

A pile of transcripts is kind of useless on its own. Without enough of you in there, the AI fills the gaps by guessing, and that’s where the hallucinated answers come from.

Your beliefs are already written down, just scattered. They’re embedded in your emails, your Slack messages, and every doc you’ve ever made. So are your unique points of view and the stories behind them. That matters, because giving someone an answer is one thing. Giving them the answer with the story, the principle, and a visual model shows them how you think.

To get at all of that, plug your AI into the tools you already use. These are called connectors. Gmail, Google Drive, Notion, Slack, Granola, as many as you can. Each one takes about two minutes to set up.

If you feel behind here, you’re not. I recently had 40 top CEOs in a room, people running eight-figure companies, and I asked them to analyze their inbox so we could build a tone document. A third of the room had never connected their email to their AI.

I barely search my inbox with the inbox anymore. I ask the AI. It finds things faster and more accurately than the search bar ever did.

If you create content on Instagram, TikTok, or YouTube, point it at those too. That’s more of your thinking on the record.

Even with all of that connected, there will be gaps. So flip it around and have the AI interview you. Set aside 20 minutes, answer in your own words, and let it save everything to the brain. Here’s the prompt I use:

I want you to have all the context you need to think like me. Use all the tools I’ve connected to learn as much as you can about how I think, what I believe, and what I’ve done. Then interview me to fill the gaps: my points of view, my contrarian takes, the stories behind my decisions. Ask at least four rounds of questions. When you’re done, output a document called [Your Name] Brain System that explains all of this simply.

Save that doc. Every AI chat you open from here on starts with it.

Teach it how you sound

Now it thinks like you. It still sounds like a robot. This is my favorite party trick, and once you see it you’ll wonder why you ever did it the hard way.

Feed it your real writing, by medium. You email completely differently than you post in Slack. You text your friends differently again, and you probably talk on camera differently from all three. So give it separate buckets.

Have it find the patterns. Sentence length. Punctuation. Favorite words. Capitalization. How you open, how you close, how you sign off. Read your last 20 emails out loud and you’d hear a pattern that’s yours and nobody else’s. The AI is great at finding it.

Hand it a banned list. Write down the words and punctuation you’d never use and give it to the AI. Delve. Leverage. Robust. Quietly. These show up in every model because humanity used them a lot, and you may never have written “quietly” in a single email in your life.

Here’s the prompt:

I want you to write in my voice so well that nobody can tell AI was involved. Pull my 20 most recent sent emails, my 20 most recent Slack messages, and my 20 most recent documents. Break down each one separately: the length of my sentences, the words I use, the words I never use, how I open, how I close, how I sign off. Keep each medium separate so my email voice stays different from my Slack voice. Don’t blend them together. If something isn’t in the samples, ask me instead of guessing. Then output a document called Voice Style Guide that I can paste into any future chat.

Then keep adding sections. I screenshotted my text messages and had it build a section just for texting. I gave it outlines and transcripts from my videos so that when I’m writing a script, the bullets come out the way I actually talk.

It turns out I’m a different writer in every place I show up. When I text, I don’t punctuate. My emails are a little more formal. On Slack I have to be crystal clear, because the whole team is going to stop and read that post. And I have never once put an emoji in a video script, though I’d drop one in a text reply without thinking.

Now you’ve got two documents: your Brain System and your Voice Style Guide. Both go into every chat.

And you still review everything. I use AI to generate ideas, draft, research, and give me three ways I might say something based on what I’ve written before. It gets me to a draft way faster. But nothing goes out under my name that I haven’t read and that doesn’t make sense to me.

Where the human stays

Here’s where it gets dangerous. The better this gets, the easier it is to make a huge mistake.

I feel it every day. People know me as the AI guy, so when I’m in my DMs replying with my real thumbs, people assume it’s a bot. I post a video and the comments say “AI, AI, AI.” Guys, I wake surf. I mountain bike. I run companies and I’m a dad. I didn’t make that up.

I did that to myself, because I teach people to use AI to buy back their time. So the most human thing I can do is show up as the person. Real camera, real markers, a real pen and paper. That’s the connection, and it’s the best use of my time. AI is the tool that makes me fast enough to shoot videos every week.

The real mistake is to stop thinking. MIT put 54 students on essay-writing tasks and watched their brain activity. The AI group showed the weakest brain activity of any group. And 83% of them couldn’t quote their own essay minutes after writing it. They never thought it. They just watched it appear.

It’s the same reason most people can’t remember a phone number anymore.

Would you watch F1 if every driver was a computer? Of course not. The human is the part you love.

So my number one rule: use AI to pull from you, not push onto you. You should be at the inception of every idea, and AI’s job is to get it into the world as fast as possible. Don’t let it tell you what to publish or how to run your business. You don’t ask a hammer how to build a house. You decide what house you want, and the hammer helps.

Here’s how that looks when I need a newsletter:

I need to write a newsletter about [topic]. Use the tools I’ve connected to pull the most context you can around this topic. Then interview me to get my thoughts on it. Use my Brain System and my Voice Style Guide to make it sound exactly like me.

The Brain System and the style guide are the recipe book. You still bring the ingredients.

Everybody has the same tools now. Same models, same prompts, same videos, same speed. What decides who wins is what you think, and that comes from investing in you: your reading, your skills, your habits, the muscle of learning and teaching and applying.

Build that, and AI multiplies it. Defer everything to AI, and one day you’ll wake up wondering how everybody ran circles around you.

Pick the one part of your work you’ll always keep for yourself, whether it’s the sales calls, the networking, or the writing that matters most. Then let AI handle the rest.

-DM