The AI conversation today is dominated by comparisons among ChatGPT, Claude, Gemini, Grok, and whatever new versions of these models were released this week. Those are interesting discussions, but I don’t think they’re the ones that will determine who actually wins in the AI era.
The biggest productivity gains won’t come from buying a better model. They’ll come from uploading your brain.
Unplugged, But Connected
Now, before anyone thinks I’m talking about plugging your mind into a machine, I’m not. I’m talking about externalizing the way you think.
All knowledge workers make many decisions every week. They maintain mental playbooks for everything from customer onboarding to qualifying leads to writing proposals to troubleshooting issues to managing projects to preparing for presentations…the list goes on and on. But over time, we all develop shortcuts, like standards and procedures, so we can move fast and avoid repetitive work.
The problem is that most of this knowledge exists only in our heads.
The Real Reason AI Doesn’t Get It
When people say AI doesn’t understand their business, they’re usually right, but it’s not because AI isn’t smart enough. It’s because we’ve never taught it how we think. We’re asking AI to replicate years of experience without giving it access to the experience itself.
Imagine for a second that you hired the smartest employee ever. They are a quick learner, never get tired, and work 24/7. But you refuse to show them your documentation, standard operating procedures (SOPs), history, what’s worked and what hasn’t, and how you come to decisions. It would be wrong of you to expect this employee to meet your expectations, right? Or even comparing them to someone who’s been at your company for a decade, right? This lack of “onboarding” is exactly how companies are deploying AI today.
They’re giving AI intelligence without context, and context is king.
This is why the concept of a “Second Brain,” popularized by productivity expert Tiago Forte, has become so relevant in the age of AI. Originally, the idea was about creating an external system to organize your knowledge so you could think more effectively.
AI doesn’t create expertise out of thin air; instead, it amplifies expertise that already exists. If your best practices, decision frameworks, and institutional knowledge have never been documented, AI has very little to work with beyond generic information it learned during training.
From Prompts to Knowledge
This is why I believe we’re entering a shift away from prompt engineering and toward knowledge engineering. For the past two years, we’ve obsessed over writing better prompts. Going forward, the organizations that gain the greatest advantage will be those that build better knowledge repositories. They’ll document their sales playbooks, capture customer conversations, organize project templates, define decision trees, preserve meeting notes, and explain not just what they do, but why they do it.
Only then can AI begin to do more than answer questions. It can begin executing meaningful work.
Research is increasingly pointing in this direction as well. Experts have highlighted that one of the biggest challenges facing enterprise AI is capturing tacit knowledge, the expertise employees develop through years of experience that rarely gets written down. AI can only reliably leverage knowledge that has been made explicit, structured, and accessible. In other words, AI scales what you’ve documented, not what you’ve memorized.
That’s why I tell clients that before we automate anything, we first have to map how they think. Building the AI agent isn’t the first step. Building the SOP the agent will rely on is.
Before You Buy the Software
Winning with AI looks like an organization that recognizes that institutional knowledge is one of its most valuable assets, not software.
So, before you ask which AI platform to invest in next, ask yourself a simpler question:
If your best employee walked out the door tomorrow, how much of their thinking would leave with them?
Your answer to that question may tell you more about your organization’s AI readiness than any software purchase ever will.
Because AI doesn’t scale intelligence, it scales documented intelligence.
Advice for Executives
Adjacent to using AI tools, your biggest AI unlock right now should be to conduct an internal audit for winning use cases.
A winning use case looks like a highly repetitive, high-impact problem that your team can map a full SOP to. So this means, if you don’t know the process of the problem you are trying to solve, it doesn’t matter how impactful it is to your business – and AI won’t save you. Thus, it’s this locked knowledge that must make its way to a written process.
I’ve done dozens of these SOP writeups for AI systems, and it’s not pretty – typically you have workers describing a process they’ve held in their heads for a decade, so understandably it’s hard, especially since AI needs the details. One miss, and the AI doesn’t know what to do.
Some suggest that one way to start to chip away at this SOP capture issue is to use something called a “Skill Creator” in Claude, which will allow the company to start to build out procedures that AI can use as a slash command to execute from. The issue, however, compared to an audit, is that this can take longer without employee guidance, can possibly capture information incorrectly if done in a silo, and worst of all, it doesn’t necessarily guide teams towards solving the problems that are dragging them down in the first place, as there hasn’t been a thorough review yet.
Take my advice. Audit first, buy AI second.
