- The AI Brief
- Posts
- I Learned to Document AI the Right Way (It's Not Prompts)
I Learned to Document AI the Right Way (It's Not Prompts)
And the 2-Step Process to Make AI Sound Like You

In Today's 5-Min AI Brief
Your weekly briefing on working smarter with AI to stay focused, creative, and ahead of the curve.
🛠️ Your AI Toolkit: The playbook framework top AI operators use to document their work
đź’ Prompt of the Day: How to continuously refine your AI writing style guide
đź’ˇ Last Week in AI: OpenAI launches Atlas browser & Sora 2
đź“– What I'm Learning: Future of Business in the age of AI with Hormozi and Replit CEO
Hey, here's this week's AI Brief!
I've been heads down in an AI bootcamp recently, and it gave me an important framework for thinking about how we can best transition workflows to AI tools. So, today, I’m sharing that with some BIG AI releases, so stay tuned.
🛠️ Your AI Toolkit
AI Playbooks Are More Powerful Than AI Prompts
Most people are saving prompts when they should be developing AI playbooks instead. A prompt is just one piece of a playbook.
An AI Playbook is a complete document that captures the business problem, the inputs needed, the process step-by-step, the expected outputs, and how to evaluate if it worked. Ask yourself, if I gave to this playbook to a new intern, could they do the job?
So for the next workflow you want to automate with AI, document the following:
What problem does this solve?
What information do I need first?
What's my exact process & what prompts do I use?
What makes a good output?
đź’ Prompt of the Week
Continuously Refine Your AI Writing
For whatever writing project you're working on: newsletter, social posts, client emails, you should have a style guide that trains your AI to write like you. Today’s prompt tip assumes you have a style guide, but if you don’t just watch this video tutorial first.
But here's what most people miss: Your style guide isn't a one-and-done document. You need to refine it every time you notice AI drifting from your voice.
Every time you catch AI writing something that doesn't sound like you, you're identifying a gap in your style guide. This prompt turns that gap into a documented rule, so it gets better every week.
Step 2: The Writing Refinement Prompt:
I just reviewed the content you wrote, and I noticed a discrepancy between the output and my actual writing style.
Here's what you wrote: [Paste the AI-generated section that sounds off]
Here's how I would actually write it: [Paste your edited version]
Please update my style guide in this project to capture this difference. Specifically note:
What pattern you were following that I corrected
What my actual preference is
When to apply this rule going forward
Then show me the updated section of the style guide.
đź’ˇ Last Week In AI
Two Major Launches Last Week
OpenAI Launches Atlas: AI Browser
The trend is clear with AI browsers. We went from 1 AI browser to 4 in the last 3 months because AI companies know that users want to use AI where they're already spending time.
The agentic capabilities of these browsers are really impressive (here’s a tiktok demo). Soon we’ll be able to make AI workflows by talking to our browsers.
Sora 2 Video Model: Impressive Tech, Copyright Uncertainty
OpenAI launched Sora 2, and it's TikTok for deepfakes—hyper-realistic AI videos of anyone, including celebrities and presidents.
Sam Altman shared the world needs to "contend with incredible video models" now, acknowledging the copyright and bad actor concerns directly.
Should you use it in your work? My personal take is use it only for organic & internal materials, but not for paid media or services until its more legally clear.
📖 What I’m Learning
Future of Business in the Age of AI (Alex Hormozi x Replit CEO)
I just watched Alex Hormozi and Replit CEO Amjad Masad break down what's actually working in AI businesses right now. Here are two things you can act on:
1. AI-Powered Lead Magnets. Alex and Amjad discussed building a "funnel scan bot" as a free lead magnet: you'd enter your URL, AI crawls your funnel, screenshots each page, and gives specific feedback like "Your offer is weak on page 3." You could use AI to give more free value, then upsell the recurring tools/services they’ll need.
2. Build "Action Systems" Not "Information Systems". Most business software tells you what to do: your analytics says "conversion dropped 15%," then you go execute manually. But what if your AI system actually did the thing, instead of just recommending it? This is all possible today, we just haven’t adjusted to this reality.
Final Thoughts
I hope this helped you understand how to use AI to work smarter, thanks for being here. Let me know if you have any questions!
Thaddeus
P.S. If you want help building AI systems like these for your team, check out my agency
Reply