FOR FOUNDERS AND OPERATORS BUILDING WITH AI FOR EXPERTS WHO WANT AI TO USE THEIR JUDGMENT
Transfer context, judgment, and corrections into AI—one useful workflow at a time. Teach AI your standards, examples, exceptions, and corrections—so the next attempt starts smarter.
I built two companies, sold one, took another public, raised roughly US$20 million, and led an international team of about forty people. The 38X AI Company is the operating system I wish I had when growth still meant more headcount, complexity, and decisions coming back to me.
US$7 PDF + EPUB, one time
Stripe checkout. Instant download. Seven-day guarantee.
Hong Kong: 20 signed, numbered hardcovers, US$20, digital files included.
WHY I WROTE THIS
From the outside, it looked like success: two companies, an acquisition, a public listing, millions raised, and an international team. Inside, every important decision still found its way back to me.
After more than a decade without a month-long holiday, I stepped away, helped install a successor, and began rebuilding around a different question: can a company create real value, make money, and improve without consuming the founder's life?
This book is what I learned from combining that operating experience with AI. It is not a collection of tool tricks. It is a practical method for transferring judgment, checking reality, and making the next attempt smarter.
“I am not teaching AI from the sidelines. I am using it to rebuild the way I wish I had operated from the beginning.”See the full founder journey
FOUR WAYS AI LEVERAGE GOES WRONG
AI can make the wrong work look extraordinarily productive. Which version are you living?
Apps, agents, offers, and automations pile up. You are moving faster than ever, but you cannot point to the business result that improved.
What is really happening: AI removed the cost that used to force you to choose. The bottleneck moved from execution to judgment.
You rewrite the tone, correct the same assumptions, and explain the exceptions that feel obvious to you. The prompt changes; the disappointment does not.
What is really happening: AI has your instruction, but not the standards and hard-earned distinctions behind your craft.
You copy context between tools, rescue broken handoffs, check whether anything actually happened, and keep the whole system alive yourself.
What is really happening: The workflow can execute, but it cannot see enough context or judge its own result.
The AI produces more. Your team produces more. Yet quality, priorities, customer nuance, and the final yes or no still live in your head.
What is really happening: Your business is using AI labor without transferring the judgment or memory that would make it less dependent on you.
LOOK INSIDE
See the real writing, illustrations, and argument—28 pages from the finished book.
A DIFFERENT WAY TO USE AI
The model is widely available. Your advantage is what surrounds it: the job you choose, the judgment and examples you teach, the context you connect, and the corrections you preserve after real work.
THINK
Ask it to question you one answer at a time until the job, business result, evidence, and assumptions are clear. Sometimes the best result is deciding the work does not deserve automation.
TEACH
AI may know more than you about a tool and still know nothing about your customers, standards, approval process, or unwritten rules. Give it a job description, then let its questions reveal what you forgot to explain.
CONNECT
A good sample makes the expected structure visible. Your commentary explains which parts are fixed, why the work is acceptable, and which details belong only to that example.
CONNECT
Ask AI to explain tools and integrations in plain language, then separate verified capability from assumptions. Documentation, a controlled test, or IT confirms access—not AI confidence.
SEE
Break the workflow into small tasks with inputs, outputs, dependencies, and checks. A supplied exchange-rate file can be checked for missing currencies before anyone tries to automate collection, approval, and distribution.
LEARN + REMEMBER
Compare the actual result with the expected one, confirm the lesson, and save it as an instruction, example, mapping, or test where the next relevant run will receive it.
The book shows how to plan the whole job, verify the limits, and complete one useful assignment before you add more automation.
THE 38X FRAMEWORK
You do not need to automate everything. Give one AI colleague a clear job, the right material, a visible standard, and a result you can check. Then use what happened to improve the next attempt.
THINK → TEACH → CONNECT → SEE → LEARN → REMEMBER ↻
WHAT CHANGES AFTER YOU READ IT
Start building the judgment, context, feedback, and memory that let AI create useful business leverage.
THIS BOOK IS FOR YOU IF
It is written for entrepreneurs, solopreneurs, experts, and operators who already use AI but still carry the standards, context, feedback, and memory themselves.
It is not:
ONE PURCHASE. A COMPLETE STARTING POINT.
A fixed-layout edition for reading, printing, and annotation.
A reflowable edition for Kindle, Apple Books, and other readers.
See how effectively you use AI now and where your leverage breaks.
A guided interview that finds one worthwhile job and defines its first achievable assignment.
A practical workbook and copy-and-paste prompt library for all six stages of the framework.
Install Keith’s questions inside Codex, Claude Code, Cursor, and other compatible agents.
The principles in the book do not expire. The tools do. The companion library is where implementation resources stay current.
INCLUDED BOOK-BUYER AGENT SKILLS
These are not generic prompt templates. Each skill pushes back on vague answers, inspects the material you provide, separates facts from assumptions, and leaves you with a file you can use. All five are included with the book and delivered in your private buyer library.
Choose one recurring job that deserves AI execution and define what winning means.
Turn your context, examples, exceptions, and unwritten standards into a usable job brief.
Verify access, break the workflow into tasks, and plan the first useful segment.
Compare expected and actual results, then preserve a tested correction.
Run the complete THINK → TEACH → CONNECT → SEE → LEARN → REMEMBER workshop.
WHAT PEOPLE SAY ABOUT KEITH'S TEACHING
These testimonials are about Keith's courses and teaching—not reviews of The 38X AI Company. Book reviews will be shown separately as readers submit them.
LIMITED AUTHOR EDITION · 20 COPIES
Keith is setting aside 20 copies for readers in Hong Kong. Each hardcover is personally signed and numbered, with an optional one-line inscription.
Outside Hong Kong? Order the paperback or hardcover from Amazon.
Standard untracked Hong Kong shipping included
US$20 total · Hong Kong addresses only · Phone number not collectedCancel for a full refund before signing or dispatch. After dispatch, personalized copies cannot be returned for a change of mind; damaged, lost, or incorrect parcels will be replaced or refunded. Full terms · Privacy
START WITH ONE LOOP
Download both formats immediately, then use the companion Assessment and Builder to find the first part of your business to strengthen.
One-time payment
US$7 Secure checkout through Stripe. No subscription.If it does not help you identify where your AI workflow breaks and what to improve next, email keith@rumjahn.com for a full refund.
QUESTIONS
The US$7 direct purchase includes the complete PDF and EPUB. A limited run of 20 signed, numbered hardcovers is available for Hong Kong delivery and includes both digital files. Paperback, hardcover, and audiobook editions are also available from Amazon.
No. The framework starts from the work you already understand and the technical ability you have today. Your first useful loop may be as simple as AI plus a spreadsheet.
No. The book starts where frequent AI users often get stuck: producing more work without transferring enough judgment, context, feedback, or memory for the system to improve. It is about the capability surrounding the model, not beginner prompt syntax.
No. It explains why copying a good prompt often fails to reproduce expert work, then shows how to make your standards, examples, corrections, context, and outcomes usable by AI. The companion library still includes every prompt from the book so you can put the ideas to work.
No. Models and tools change quickly. The book focuses on the durable capabilities a business needs: judgment, context, outcomes, learning, memory, and bounded automation. The companion library is where implementation resources stay current.
No. Start with one recurring piece of work you already understand. The book helps you find the weakest part of that loop and strengthen only that part before adding more automation.
Your purchase includes the AI Fluency Assessment, the guided Build Your First 38X Loop interview, a practical workbook, a copy-and-paste prompt library, and five buyer-only Agent Skills for Codex, Claude Code, Cursor, and other compatible agents.
Stripe sends you back to a private download page immediately after payment. We also email a reusable access link.
Read it for seven days. If it does not help you identify where your AI workflow breaks and what to improve next, email keith@rumjahn.com for a full refund.
THE 38X AI COMPANY
PDF + EPUB · Immediate access · Seven-day guarantee
The 38X AI Company US$7 · PDF + EPUB