@tjrobertson52

It's never been easier to replace yourself with AI. And I don't just mean, like, routine processes that you can write out of the checklist. I mean, like the really difficult stuff, the stuff we thought we'd never be able to turn over to AI.

So I just wanna walk through exactly how I've been doing this and the major unlock I had recently. Because if you're like me, if you're a founder or at least an expert or specialist in your organisation, there's a very good chance that you're the bottleneck. People are depending on that tacit knowledge or intuition that you develop through experience.

And until recently, it really felt like that was something AI couldn't handle. But I'm gonna show you the way.

Let's start by considering why this doesn't feel possible. If you're someone with 5, 10, 20 years of experience, you very likely developed a strong intuition when people come to you with difficult questions. You just have a feeling you've seen a problem like this a hundred times before, and you can take all the context and nuance and confidently put forth a plan for addressing the problem.

The actual decision process or framework is mostly invisible to you. But of course, this isn't magic. There is a decision framework.

The neurons on your brain are processing information and outputting a result.

The difficult part is documenting that framework and all the nuance and edge cases that it comes with. But having done this myself for a handful of decisions that I make on a regular basis, it's incredibly powerful. So let me just walk through the process I'm using right now.

First, you need to identify a problem or category of problem, but try to keep it small. For me, this would be things like keyword research or internal linking, resolving website technical issues, or proposing a site map structure. Pick one of these problems and create a skill in Claude.

Now, you can use ChatGPT, or Gemini. 6. If you've never made a skill before, don't worry.

Claude will walk you through the whole process. Just tell it exactly what you're trying to do. I need to make a skill that solves this problem.

Then you're gonna give it your best first draft of the framework. Carve out an hour and just really spell out the whole process you go through in your head. If you have any documentation or examples of times where you've gone through this process, throw that in there as well.

Claude does a great job with lots of messy context. And then ask Claude to give you any follow up questions it needs to finalize the skill. Claude's gonna take all that information and package it into a nice organized skill file.

That's step one.

Now, every time you're dealing with this problem, use this skill in Claude. If you get a question from a customer, someone in your team, or you're just solving the problem on your own, go to Claude, paste in the context, and link that skill. Say, Claude, help me solve this problem using this skill.

And what you're probably gonna notice very quickly is it's gonna make a lot of mistakes. It's gonna get a lot of things wrong and do things differently than you would. Your job is to itemize each of those discrepancies, explain to Claude what it got wrong and what you would have done instead.

And then at the end of that process, say, Claude, I want you to go back through our entire conversation, take everything you've learned, and update the skill. And that's the magic.

Do this a dozen times, 20 times. Eventually it'll be making these decisions the same way you would 90% of the time. Plus, it has all the advantages of being an AI.

At that point, you can hand it over to your team. Now, for a while, you're still gonna wanna check the output, but I have about half a dozen processes I'm developing this way simultaneously right now, and it's kind of amazing how close it gets.