I’ll buy you a beer if you’re higher than I am.

(Note: if you want to skip straight to the stages, you can click right here [<— Link]. Beware: the thesis of the thing you’d skip is this: Using AI well means building your own house, which requires starting with where you’re at, not reaching with envy towards the things you hear about online or the water cooler. These musings are a part of MY journey towards AI Empowerment, and all those in-between moments are the bricks I’m standing on right now. Skip at your own peril)

Six months ago, and yesterday

Six months ago, if I needed to get some work done and wanted some extra brain power, I’d fire up my chrome browser and log into ChatGPT. I’d say something, she’d say something (my ChatGPT agent identifies female, don’t ask me why), repeat repeat repeat until whatever project I was working on was done. I’d feel seen, heard, and like I had a great idea that just needed a little polish.

Yesterday, I typed a few lines into VS Code that triggered a script that would summon, spawn, invoke (whatever) 20ish different agents, all with different areas of expertise, across three providers to build an AI powered questionnaire that has multiple routes and personalized results. Forgive the foreshadowing.

Even sitting here and typing that last paragraph makes me sort of raise an eyebrow at myself. The workflow for me seems so automatic, but reading all those words up there squished together and I start to confuse myself. So, it’s hard to describe the journey and transformation this AI landscape has taken so many people like me on.

I’m a teacher by trade

By trade, I’m a teacher. Not particularly smart, but I can’t let myself move on from a topic unless I feel I fully understand it. Then, to test my own understanding, I usually go off and see if I can explain this new bit of information to someone else. Rinse, repeat.

So, when it comes to AI, recently I have felt like I can just SEE it. The landscape. What’s possible. How people should be operating with it. So when a friend or a relative comes to me and talks about their problems, unless that problem involves something like “I need to chop this tree down…” my knee jerk reaction was to explain to them exactly how I would solve that problem.

Almost without fail, the result was a blank stare after I started using phrases like “claude code” or “agentic workflow.”

You don’t start at calculus

It seems obvious now… but I obviously when a person is learning something for the first time, like… math … you don’t jump straight to “Well, calculus solves basically all math problems. Let’s start there.”

“Curriculum” really just means take the thing, deconstruct it into some kind of narrative, and get somebody caught up on the thing in a sequence that will be ingestible to them, at a speed they can handle. There’s no one right way to build a curriculum, but there are many wrong ways to do it. For example: using the phrase “agentic workflow” to a person who is chatting with Grok on their phone.

Step 1 is not to copy them

If you’re listening to people talk about AI, and feeling behind the curve, step 1 is not immediately start copying what they are doing. Step 1 is find out where you are at as an AI user, and then identify a few tangible things you can start doing to level yourself up. In my very strong opinion, people who are becoming AI power users are not the ones trying to take shortcuts. They’re the ones who are building their own houses. You have to live through the set up to really appreciate and utilize what you have in the end.

They’re the ones who are building their own houses.

The retrospective

So, with this belief, I did a retrospective analysis of my entire journey from peasant ChatGPT validation gluten to becoming the fully agentic operator that I am today. I dug deep to find every example of something I tried to do with AI. I then distilled out what worked, what needed to change, what needed to get benched, and the next thing that it occurred for me to create after the last.

The result: 11ish stages – and they all have a substantial list of things to look out for while in that stage (in the hopes you can save yourself from the mistakes I made) and, in my opinion, a few things you should start trying to get yourself up to the next level. So here they are:

[DISCLAIMERS: 1. I don’t play the same sport as AI engineers or real developers. They know more about this thing than I ever will, most likely, and will likely have different stages and problems than below. 2. Getting to stage 11 means you need to have a lot of control over your own companies permissions and software. If you work somewhere where permissions are on lockdown, this climb becomes harder as an individual operator and, likely in some cases, could violate workplace policies. Don’t go rogue, don’t ask your AI for permission. Go talk to your manager or tech team if you’re not sure what you’re about to do is safe.]

The 11 stages

1. Stage 1: You’ve heard about AI (but don’t really use it at work).
2. Stage 2: AI answers your questions (chatbot style)
3. Stage 3: AI sees your actual work (you feed it emails, files, schedule)
4. Stage 4: It knows who YOU are (you’ve created a mini-harness so this thing knows who you are)
5. Stage 5: Your first commands (you’ve done the same thing multiple times and it occurs to you… wait there’s probably a faster way to bottle up this workflow).
6. Stage 6: AI does TASKS for you (enter: agents. Stuff gets created and logged while you get a coffee)
7. Stage 7: AI builds your software (seriously do not start here)
8. Stage 8: Many bots, run by a script (agentic workflow I was mentioning earlier. Orchestrated by a script not your “orchestrator agent”)
9. Stage 9: Your own clean data (if all datum were roads, yours would lead to Rome)
10. Stage 10: AI that real people use (the tools you created serve real people every day whether they’re your team or your customers)
11. Stage 11: Agentic operator (you can’t remember the last time you opened a good ol’ fashioned off the shelf software, and when people ask you how you’re so productive it’s hard for you to answer).

The DON’T list

Every one of these stages is fleshed out with the “do next” and the “don’ts”. I also have a list of rules (The DON’T list) that applies to all stages of this ladder. A few examples, “build systems, don’t buy”… “buy software, unless an AI system can do it instead”… “don’t talk to Grok (yet).”

Where to go from here

For the complete list, check out the ladder stages here → [###########LINK##############)

Or, even better, if you want to get a personalized take on where I think you are in your journey, fill out the questionnaire → [############LINK#############]

If any or all of this resonates with you, and you want to know more about how to use AI better, that’s likely because you use the internet and have a job. Meaning: we’re all thinking about stuff like this right now.

If you’re interested in comparing notes, climbing a few rungs to this ladder, or adding new ones to the top, apply to a cohort [#########LINK##########].

And if I owe you a beer, let me know and we’ll make good on that.

I’ll buy you a beer if you’re further up this ladder than I am. I mean it.

Here’s the short version of why I built it. Six months ago I used AI one way. Today I use it in a way that would have looked like science fiction to me back then. The whole climb happened in small steps, and almost none of them were the steps I saw other people bragging about online. This is the map of those steps, so you can find the one you’re standing on.

Six months ago, and today

Six months ago, when I needed help with work, I opened a web browser and logged into ChatGPT. I typed something. It typed back. We went back and forth until the task was done. It felt great. I felt heard, like my idea just needed a little polish.

Today I do something that still makes me raise an eyebrow at myself. I type a few lines into the plain text window where I now do most of my work. (It’s called VS Code. Think of it as a control panel for talking to AI.) Those few lines kick off a saved instruction. That instruction wakes up about twenty different AI workers at once, each one a specialist, running across three different AI companies. Together they build something. Last time, the something was a questionnaire that asks you a handful of questions and hands back a personal result.

Reading that back, even I get a little lost in it. The work feels automatic to me now. But the words, stacked up like that, sound like a foreign language. And that’s the whole problem with talking about AI. The people who are deep in it forget how far away the starting line looks.

I’m a teacher, so this bugs me

By trade, I’m a teacher. I’m not the smartest person in the room. But I can’t leave a topic alone until I understand it, and the way I test whether I really understand something is to explain it to someone else. If they get it, I got it.

So lately I’ve felt like I can just see the AI landscape. What’s possible. How people should be working with it. And when a friend or a relative tells me about a problem, my first instinct is to explain exactly how I’d solve it with AI.

Almost every time, I get a blank stare. It shows up the second I say something like “Claude Code” or “agentic workflow.” I lost them, and that was my fault, not theirs.

You don’t start a math student on calculus

Here’s the thing I should have seen sooner. When someone is learning math for the first time, you don’t open with “calculus solves almost everything, let’s start there.” You’d lose them on day one.

A curriculum is just this: take a big thing, break it into a story, and walk someone through it in an order they can absorb, at a speed they can handle. There’s no single right order. But there are plenty of wrong ones. Saying “agentic workflow” to someone who’s texting with a chatbot on their phone is a wrong one.

The first move is not to copy the person ahead of you

If you hear people talk about AI and feel behind, the first move is not to copy what they’re doing. The first move is to find out where you are standing right now. Then pick a few small, real things you can do to climb one rung.

Here’s what I believe, about as strongly as I believe anything about this:

The people becoming real AI power users aren’t the ones taking shortcuts. They’re the ones building their own house.

You have to live through the setup to understand what you’re holding at the end. A house you built, you know how to fix.

How I found the rungs

So I went back through my own six months. I dug up every single thing I’d tried to do with AI. For each one I asked three questions. What worked? What had to change? What did I have to give up on? And every answer pointed me at the next thing to try.

What came out was a ladder. Eleven rungs. Each rung has its own list of traps to watch for, so you can skip the mistakes I made, plus a few things to try when you’re ready to climb.

Two honest warnings first. One: I’m not an AI engineer or a software developer. They know far more than I do and would draw this ladder differently. This is an operator’s map, not a programmer’s. Two: the top rungs need you to have real control over your own company’s tools and permissions. If you work somewhere that locks those down, the climb is harder, and some moves could break your workplace’s rules. Don’t go rogue, and don’t ask your AI for permission. Ask your manager or your tech team.

The eleven rungs

Read this bottom to top. Find the highest rung that still sounds like you, because that’s where you’re actually standing, not where you wish you were. The goal isn’t to leap to the top. It’s to climb one rung.

1. You’ve heard about AI but don’t really use it at work.
2. AI answers your questions. You ask, it answers, like a chat.
3. AI sees your real work. You start feeding it your emails, files, and calendar.
4. It knows who you are. You’ve written it a short profile, so it answers like it knows your job and what you’re after.
5. Your first shortcuts. You’ve done the same task by hand enough times that you finally bottle it into a saved instruction you can run again.
6. AI does tasks for you. Now it goes and does the work while you get a coffee, and leaves a record of what it did.
7. AI builds your software. It can write working tools for you. (Seriously, don’t start here.)
8. A whole crew, run by an instruction. Many AI workers at once, kicked off and steered by a saved instruction instead of you babysitting each one.
9. Your own clean data. All your information is tidy and connected, so the AI is never working from a mess.
10. AI that real people use. The tools you built serve people every day, whether that’s your team or your customers.
11. Operator. You can’t remember the last time you opened a piece of regular, off-the-shelf software. When people ask why you’re so productive, you struggle to answer.

The one list that covers every rung

Each rung has its own do’s and don’ts. But I also keep one list of rules that holds at every rung. I call it the DON’T list. A few from it: build systems instead of buying them. Keep the software you already pay for, unless an AI setup can do the same job better. And don’t talk to Grok (yet).

The thread running through all of it is the same one from the start. The thing you’re missing is almost never a new tool. It’s a cleaner, more connected setup underneath the tools you already have.

Where to go from here

If you want the full ladder, with the traps and the next steps spelled out for every rung, the whole map is here. It’s also where you can get a read on which rung you’re standing on right now.

And if you want to compare notes, climb a few rungs alongside other people, or add new ones to the top, come join a cohort.

If any of this lands, it’s probably because you use the internet and you have a job. Which means you’re already thinking about this stuff. So am I.

And if it turns out I owe you a beer, find me. We’ll make good on it.