Models expire.
Your system compounds.
Learn the fundamentals of AI and build a system that turns what you know into excellent work, on any model. 49 lessons, 7 projects, self-paced.
Everyone's catching up with AI.Catching up is how you fall behind.
New model, new tool, new prompt pack. And most of what AI hands you is good enough: rarely excellent, and too risky to trust blindly with your clients, your money or your reputation. Every conversation starts from zero, and nothing accumulates.
The people who get the best out of AI own something the tools can't take back: a system. A framework for their context, their processes, their definition of excellent, and the checks that hold every output to it. It runs on any model, and it gets better every time they use it.
Most AI courses teach you the tools. This course builds the layer the tools run on.
The layer between you
and any AI.
It also controls quality. You know why an output is good, and you catch it when it isn't.
Seven modules.
Seven things you own.
Zoom Out
Where AI stands right now, the four things that won't change, and a first look at the layer you'll own, in a form any AI can use.
Clarity OS
Research, analysis, a one-page brief, a five-minute deck, all checked before they ship.
Nucleus OS
One home for your SOPs, prompts, specialists and evaluations. Then the same system as plain files any AI can read.
Specialists
Assistants built from your own instructions, like a Brand Voice Editor that rewrites your drafts in your voice. Tested and registered in your Nucleus, on ChatGPT, Gemini or Claude.
Autopilots
Automation is a privilege earned through proof. In the lessons: a lead comes in, the AI qualifies it and drafts the right email, and nothing goes out until you approve it.
Agents
A research team of three agents: a manager, a researcher, and a challenger that hunts for gaps. When not to build an agent at all. Then an agent working inside your own folder, under your rules.
Make It Trustworthy
Define good, test everything you built on cases you chose, improve from the evidence, and put the human in the right place.
Hi, I'm Charafeddine.
I led AI rollouts for 4,000+ employees and have shipped 60+ AI systems. I'm an AI scientist and a professor, and I've trained 200+ professionals live since 2023.
What people built with it.
Version 2 of the course opened in October 2026. The bootcamp taught the same framework.
“After this bootcamp, we were able to split in half Board readiness timing and ensure built-in audit trail in 3 finance performance initiatives (cash management, shareholders reporting, budget and forecast).”
“In my final pitch project, I've built a team of 3 Agents who now handle a key (marketing) business process, has an 80% reduction in friction and has reduced the amount of time spent from 11 hours to 2.2 hours.”
“The bootcamp enabled me to build, test, and select the tools best suited to my specific use cases. I'm no longer guessing which AI tool to use — I have a system for evaluating them.”
“I now ship resilient, layered systems that I actually know how to debug and fix when they hit edge cases.”
You won't practice alone.
The course opens three months in the Engine Room, Cohorte's community of people who use AI at work every day. They bring real problems, compare what worked, and keep the method sharp long after the last lesson.
People doing the same work.
Operators, engineers, analysts and managers, from bootcamp alumni to course buyers. Ask how someone applied a lesson, compare notes, find who has already solved your problem.
Live, every other Monday.
An open Q&A on Zoom at 18:00 Paris time with Charafeddine, who teaches the course. Bring the prompt that disappointed you or the decision you're stuck on. Can't make it? Leave your question in the thread.
The Helpdesk.
Post a real question from your work. Charafeddine answers in writing, members add what worked for them, and every answer stays searchable for the next person.
You don't. The training rewarded it for agreeing with you. … Generate in one session, judge in a separate one, and ban “is this good?” as a prompt.
Charafeddine's answer, first lines.After the three months, staying is optional: €29 a month or €249 a year. How the Engine Room works
This course isn't for everyone.
This is for you if:
Don't take this if:
Get 20% off and lifetime access.
- 49 lessons across 7 modules, 33 videos
- 7 projects, ending with your Trust Report
- Every template you'll use: the Nucleus OS for Notion, the workbook, the Prompt Vault and the Trust Toolkit
- The demo operational layer: a fictional consultant's complete system, to practice on before your own
The live AI OS Bootcamp includes this course and adds six weeks of live work on your own workflows. Applications open soon.
14-day money-back guarantee.
Start the course. If it isn't right for you, email us within 14 days of purchase and we refund you in full, no questions asked, as long as you haven't completed the first module.
Start building your AI OS- How is this different from the live bootcamp?
- Same framework, same rigor. The bootcamp adds live sessions, written feedback and a cohort over six weeks. If you want live coaching, do the bootcamp. If you want flexibility, do the course.
- I already use ChatGPT or Claude every day. What does this add?
- A system. Your context, your method and your checks stop living in your head and in old chats. A teammate can work your way, and when a better model comes out you plug it in.
- Do I need technical skills?
- No. The course is built for managers, analysts, marketers, founders and professionals who use AI as a tool. No code, no engineering background. When Module 5 puts an agent on your own folder, the lesson walks you through every step.
- How much time does it take?
- Plan for 4 to 6 hours per module, plus the project. At one module a week, you have a working system after four weeks and finish in about seven. Access never expires.
- What tools will I need, and will my IT team approve them?
- One paid AI plan (ChatGPT Plus, Claude Pro or Google AI Pro, around €20 a month), Notion and Loom (free). Modules 4 and 5 add Zapier and n8n (14-day trials) and about €10 of API credits. For IT: you choose what data goes in, a fictional consultant's system is included so the Module 5 agent runs on it before it touches your files, and your Trust Report can use fictional or sanitized cases. A guide shows the same system on Microsoft 365, Google Workspace or plain files.
- Can my company pay for this?
- Yes. Email [email protected] with your company's details and we send a quote and an invoice. Bank transfer accepted. For several seats or a team program, see Teams.
- Can I use this with my own clients?
- Yes. What you build is yours, and you can build it for a client and hand it over. The course is in English.
- Is there a refund policy?
- Yes. 14 days. If you haven't completed the first module, email us and we'll refund you in full, no questions asked. After that, the course is yours.
Models expire.
Your system compounds.
The complete system, and the proof of what it can be trusted to do. Self-paced. €397 until 13 November, then €497.
Start building your AI OS