At every company I’ve led, I’ve enforced a deceptively simple rule: If you can’t teach someone else how to do your job, then you don’t really know your job.
This isn’t just a philosophy—it’s an operational requirement. Every time I spend time doing something—solving a problem, setting up a workflow, handling an edge case—I treat that time as an investment. And the return on that investment comes when I convert that work into something teachable: a structured, reusable topic that someone else can learn from.
When done right, you don’t just end up with notes—you build a living compendium of training topics. Each topic becomes a mini-package of knowledge: something that can be used to onboard, delegate, and scale. It becomes part of the company’s brain.
Better yet, these topics can now be handed to an AI agent. That’s what we’ve done. The training topics we’ve built over time—what used to just be SOPs and onboarding docs—have now become the building blocks for our AI work assistants. These agents use our real work topics to guide new team members, handle support queries, automate task handoffs, and enforce company-specific standards. It’s the most effective knowledge transfer we’ve ever achieved.
But when this approach isn’t taken? You get fragility. Institutional knowledge becomes tribal. Delegation turns into friction. And every time someone moves on, they take half the system in their head with them. Training becomes improvisational theater instead of structured progression.
And here’s the hardest part: ensuring people actually do this.
Every time I hire a new head—whether in product, ops, engineering, or sales—I sit them down and explain this philosophy. Every time, without fail. They nod. They agree. They say it makes perfect sense.
And yet, when they move on or transition up, I almost always find a dearth of organized, comprehensive documentation. The topics may exist—scattered across chats, meeting notes, task systems—but they aren’t compiled, refined, and made teachable. They aren’t ready for handoff.
That’s why I’m especially excited about the next evolution: AI work assistants monitoring this very process. Soon, our agents won’t just use training topics—they’ll track whether they’re being produced. As new work emerges, our AI will check whether it’s been converted into a topic, who it’s assigned to, and what’s still pending. The AI won’t just support the work; it will encourage people to complete the loop—to formalize what they’ve done so others (and AI) can take it forward.
The goal is simple: No work should be done that isn’t teachable. No knowledge should stay trapped in one person’s head. And no company should grow without growing its brain alongside it.
…