Leadership used to be 100% people. Now half the work is AI. Three skills unlock that half. The third one is the one most leaders refuse to touch.
My coaching session ended at 12:30pm. I had three meetings, payroll, and a lesson to prepare for. By 4pm, I messaged my client to say the website was live.
The three skills that made that possible are the same skills every founder needs to lead in the AI era.
A Coaching Session. An Unused Domain. 4.5 Hours.
Eric and Julien were in my coaching session on Thursday. Eric wanted to know how to set up an AI workflow to manage his new company website and marketing. Julien wanted to understand style guides and Power BI. At the end, they both asked the same thing: "Can you share the projects we worked on? I want to see how you set this up."
I had bought the domain caiocoach.com a year ago. Never used it.
The session ended at 12:30. I had three more meetings, payroll, and a lesson to prep. By 4pm, I messaged Julien: the site is up. The coaching program is live.
I was running two businesses while my AI team built a third.
That is not a productivity story. That is a different way of leading. And it required three specific skills most leaders have never been taught.
The Job Got Bigger
Leadership used to be one thing: lead people. You hired well. You built culture. You held people accountable. That was 100% of the job.
We are in the 50/50 era. Half of every workflow is becoming AI. Half of the work your team does can now be directed to, delegated to, and executed by AI agents, tools, and systems. Not a prediction. It is happening now.
But most leaders only learned to lead people. Nobody taught them how to lead AI. And because of that, most organizations investing in AI are getting almost nothing back. Not because the tools don't work. Because nobody in the room knows how to lead them. We covered the org-chart side of this gap in the four offices of the future. This post is about the skills the leader inside that org chart needs.
What Leading AI Actually Requires
Three skills unlock the other 50%. None of them are technology skills. All three are leadership disciplines.
Skill One: ABC. Always Be Cataloguing.
I preach this constantly. Catalogue everything. Name your files with intention. Build style guides. Document your voice, your process, your standards.
Then I looked at my own image library. No naming conventions. No structure. Files I could not find. Images with no indication of what was in them.
I was telling founders to catalogue while I hadn't done it myself.
That is the gap. Most leaders have knowledge they cannot transfer. Insight locked inside their own heads. Systems that fall apart the moment they step away.
If your AI cannot find it, it does not exist. ABC is the precondition for everything else. Before you build a workflow. Before you run an agent. Before you delegate anything. Get your catalogue in order.
What to catalogue: your brand voice and style guide. Your process documentation. Your image and asset library with clear naming conventions. Your templates and frameworks. Your target audience definitions. Everything your AI team needs to work without you in the room.
Skill Two: Workflow Design Is the New Org Chart
Think about it the same way you think about your org chart. Who does what? In what order? What triggers the next step? Who is accountable for the output? When does a human need to be in the loop?
Those questions used to only apply to people. Now they apply to AI agents too.
When I built caiocoach.com, I was not running one conversation with one AI. I was running parallel threads inside a single project, each with clear instructions, each drawing from the same style guides and context files. One thread handled the structure. One handled the content. One handled the coaching pages. I directed the work. My AI team executed.
The five-step pattern: catalogue and context, then map the workflow, then assign AI threads, then direct and review, then ship. The reason the threads stay coherent is iteration at each handoff, the same skill we wrote about in iteration is the secret weapon of AI power users.
Workflow design is the skill of knowing how to break a goal into components, assign each component to the right resource, and coordinate the output into something coherent. Founders who learn this don't just get faster. They get to a different category of output entirely.
Skill Three: Code
This is the one most leaders don't want to hear.
Three years ago, code was for engineers. The barrier was real. You either learned a language, hired someone who had, or you didn't ship. That barrier has collapsed.
You are not learning to code from scratch. You are learning to direct AI to write code that works. Different skill. Lower barrier. Much higher leverage.
caiocoach.com is not a no-code site. It is a Next.js application with custom Tailwind styling, structured pages, and a proper component architecture. I did not write any of it by hand. I directed AI to write each piece against my style guides and content briefs. The structure thread produced the layout. The content thread produced the copy. The pages thread assembled coaching offers and program flows. Every piece was code. Every piece was written by AI, directed by me.
This is not a contradiction of what an AI Officer does. An AI Officer designs the program. Code is what lets a creative generalist actually ship the program at speed. Without code, you draw on a whiteboard and pay someone else to build. With code, you direct, review, and deploy. If you're still living in the "prompts only" world, prompts are dead covers why that's no longer enough.
You don't need to be an engineer. You need to be able to ship.
Where to start with code, in order
Learn to write prompts that produce consistent, usable output. The floor. Everyone in the AI era operates here.
Rules that govern how your AI behaves across a project: voice, format, what to do and not do. This is where you stop treating AI as a chatbot and start treating it as a team.
Claude Routines automate the recurring work. Claude Code writes scripts, instructions, and integrations on demand. Two Anthropic-native tools. No engineering background required.
Next.js sites, internal tools, custom dashboards. The top of the stack. Possible for any leader who has done the three steps above.
Pick the lowest level you have not mastered. Start there.
4.5 Hours. Two Businesses Running.
Coaching session ends at 12:30. Three meetings. Payroll. A lesson to prep. A domain that had sat unused for a year.
By 4pm, the site was live. The coaching program was published. Everything Eric and Julien asked about was up and ready to share.
That result needed all three skills.
Without ABC, the AI team had nothing to work with. Style guides ready. Brand voice documented. Process structured and tested.
Without workflow design, there was no way to coordinate parallel work. Parallel threads. Clear instructions. Coordinated output.
Without code, there is no Next.js site. There is no live coaching program. There is a slide deck about a coaching program and a domain that keeps sitting unused.
The creative generalist with these three skills is not 10% more productive. The leverage is a different order of magnitude entirely.
Dave Hajdu is the founder of Edge8 AI and the AI Officer Institute, where he trains the next generation of AI leaders across Southeast Asia and beyond.