You're not non‑technical, you're just lazy

If you are in the workforce, and especially if you work in tech, you must have heard this phrase a thousand times: "I'm non-technical". You may even be the one saying this phrase. And if you are, I have some breaking news for you.

The year is 2026. Intelligence is essentially free. You can have the best teacher at your disposal 24/7. AI can literally write code for you, in any language and for any OS, allowing you to unlock the full power of your computer. And yet you're still repeating "I'm not technical" like a broken record. Well, you're in luck, because this is the sign for you to start if you haven't already. The year is 2026, and you have no excuse not to be technical anymore. And I'm gonna help you do it*.

*If any term in here is new, confusing, or daunting, don't worry. Just paste this article into ChatGPT or Claude and ask them to explain anything that doesn't click or even handhold you through the entire thing. That's the spirit of the whole piece!

No, your job isn't the exception!

You may say "But Luis, my job is mostly in Excel, how could I possibly use AI for this?" Well, you could start dipping your toes into the AI world with the Claude Excel integration, or even just ask ChatGPT to write your macros as a tiny start. But ultimately, Claude Code could build the spreadsheets entirely for you, or even allow you to move out of Excel entirely.

"But Luis, my job is mostly scheduled calls, how could AI possibly help me with this?" WRONG again. Just turn on Google or Teams call transcripts, or if they let you use Granola, even better. Have those files automatically download to your computer, or in the case of Granola, hook up their MCP to your Claude Code. You can ask anything about and do deep analysis on any meeting you ever had, starting today. You can finally make better use of your time.

"But Luis, I mostly do PowerPoint presentations." Have you ever tried Claude Design? I don't even need to develop this one. Just try it out.

After accepting these facts, the next situation most of you will find yourselves in is the following: "My company doesn't let me use any of these tools, what can I do?" And my answer will be controversial, but honest: RUN. Yes, you heard me right. Your company is actively hampering, damaging, destroying your competitive advantage. There is already a 10x difference between the people who have been using these tools for a year and you. Don't let that difference grow bigger.

So let's say you're not running yet. You believe me, you want in, you just don't know where to start. Perfect! The next section is for you!

Claude Code CLI is your in to being technical

You may have noticed a recurring pattern in the paragraphs above. Yes, Claude Code. Your way into the technical world is one $20 monthly subscription and one install away.

Notice I said CLI. This is important. Claude Code itself isn't really the point — it could be Codex, OpenCode, Cursor CLI, or whatever alternative people hype you up for or your company actually allows. But the CLI part is the thing that matters.

First, because you will face the friction of installing this tool from the terminal. You'll hit weird errors and cryptic warnings. Paste them into ChatGPT, Claude, or Gemini on the web and have them walk you through it step by step. Yes, you'll struggle a bit. That's the point. The friction is the lesson. By the time you've authenticated, you'll understand your own computer a little better than you did this morning.

And once you've installed it, you have a helpful teacher and assistant that LIVES IN YOUR COMPUTER. Not in a tab. Not behind a login. On your machine. It has access to whatever files you give it. It can search the web. It can connect to MCPs, to other CLIs, to APIs. It's the all-in-one, all-powerful tool that can take you from zero to 10x in a matter of weeks.

The appendix will explain the difference between CLI tools, REST API endpoints, and MCPs in plain language. But before you turn AI loose on your computer, there's one piece of safety equipment you cannot skip.

The one mandatory tool to accompany AI

There's one tool that's truly mandatory if you're going to work with AI seriously. Not optional, not nice-to-have. Mandatory. It's version control.

The unit you'll work in is called a repository, or repo, which is really just a folder in your computer. A special folder, though, that lives both on your computer and online, with all that history attached. Version control just means the full history of file edits, creations, and deletions is preserved. You can roll back to any point. You can see exactly what changed between yesterday and today.

This tool is essential in the AI era and works as a sort of insurance. Your AI assistant can do a lot of damage in a single command: overwrite the wrong file, delete a folder it misread, refactor something it shouldn't have touched. If it does, it does it permanently. That is, unless you have a repo. Then you can just roll back and you keep going.

The most standard place to host your repos is GitHub. Free for personal use, integrates with basically everything, and your AI tools already know how to talk to it. GitLab, Bitbucket, or Codeberg are perfectly fine alternatives if you prefer them or your company mandates one of them.

One analytics repo to rule them all

Now that we know the fundamentals, let's build the thing. The analytics repo is one of our most powerful tools that you can create to use AI agents like Claude Code on. Its goal is to aggregate all possible data sources (in a safe read only mode) so that AI can go crazy on it: generate dashboards for you, brainstorm new ideas, have the full picture of your company, personal project or even personal life if you want to use it for personal matters and chores.

Your analytics repo will be your aggregator: the one folder where all your different sources of data, documentation, and analysis live. Claude Code will help you set up your repo, you'll just need to create an account on your provider of choice (I use GitHub) and let it do the rest. I organize mine with three things at the top level, and recommend you do something similar:

• CLAUDE.md: a single file where you write down your preferences. How you want code formatted, which libraries to default to, what your data sources are, what your typical analyses look like. Claude Code reads it at the start of every session. You can edit it whenever you want, and Claude Code itself can edit it for you as your preferences evolve.
• playground/: where you go crazy. One-off experiments, throwaway scripts, that quick chart you needed for a meeting in twenty minutes. Gitignore this folder so it never gets pushed online. Use it without thinking.
• integrations/: your incremental documentation. Every time you connect a new data source — a new export, a new API key, a new MCP — you add a short note here explaining what it is, where the data lives, and how to refresh it. Future you will thank present you.

Now let's fill it up with something that's useful to you. Pick a real, low-stakes question you actually want answered. "How has my team's meeting load changed over the last three months?" Or "What does my expense breakdown actually look like this quarter?"

Then:

1. Drop a couple of Excel files into the repo (I would put them in a new folder inside of playground; this is my go to for any new experiment): sales numbers, expense reports, whatever exists. Just download them.
2. Drop in a few meeting transcripts or notes from the last month.
3. Open Claude Code in the repo and say: "Here's the data. Build me a dashboard that answers [your question]. Render it as an HTML page and open it in Chrome (or your browser of choice)."

Claude Code will do whatever it needs in terms of installations, writing the script, running it, saving the output and even opening the HTML in your browser. Done.

It will not be perfect on the first try. You'll iterate. You'll notice a date column is parsed wrong, or the chart you really wanted is a bar not a pie, or two of the fields got swapped. You go back to Claude Code, say what's off, and it fixes it. Within an afternoon you have a thing that didn't exist this morning, and the entire history of how it got built is sitting in the repo.

Start with spreadsheets and local HTML. Then you will start asking "now how do I give Claude live access to this data instead of having to manually download it every time myself?". I'll deep dive into more advanced examples in the appendix section, you can go back to it after you toy around with Claude for a while.

A small note on hallucinations

Every AI can make mistakes. You have read this phrase hundreds of times already from all major AI providers.

These mistakes are usually referred to as hallucinations. You should be wary of this. It's a risk-reward judgment. If you're working on something really critical to your business, something that needs to be right, you should check, double-check, and triple-check the logic, outputs and the data sources. If it's something with no critical impact, something you can easily detect and correct later if it's wrong, don't be afraid to go full autopilot and hand it off to AI.

Appendix: CLIs, APIs, and MCPs

Ok, I'll be honest: I got too lazy to actually write this part. It probably deserves its own article anyways. And yes, you got to this part because I called you lazy, so ironic! Well guess what: I too am lazy, and I became technical when it wasn't this easy. What's your excuse?