The Hardest Part of AI Is Going From Zero to One

I keep coming back to the same thing when people ask me how to start with AI. The tools aren’t the hard part. Going from zero to one is.

Once you use it for a single, low-stakes task, the rest gets easier. Until then, it sits on your phone like every other app you downloaded and ignored.

Here’s what I think is coming. Once Apple Intelligence and Microsoft Copilot are sitting inside the operating system, using AI gets forced on people the same way the pandemic forced remote work. You can wait for that. Or you can pick one free tool now, use it for one boring task, and build the habit on your own terms.

ChatGPT is enough. Open it tonight and ask: “What did I learn today?” That’s the whole start. Journaling. Nothing high-stakes. Nothing you can mess up.

Why this feels different from every other tool

Writing a book used to mean pen and ink. Then the typewriter showed up and sped up the physical act of typing. You still had to do the work. The author still sat there and produced every sentence.

ChatGPT is the next version of that, and it throws people off because the tool now does the typing. You give it the intent (about 50 words), and it produces the output (about 1,000 words) in a voice and style you pointed it toward. You still provide the creative direction. The drudgery of generating the draft is what got automated.

If you’ve been telling yourself “I can just type it myself,” try the 50-word prompt experiment. Write 50 words of intent. Let it produce the draft. Then look at how long that would have taken you to type from a blank page.

Stop sounding like a robot

The fear I hear most is that everything comes out generic. That’s a prompting problem, not an AI problem.

Here’s the method. Call it Analyze and Apply.

1. Paste something you already wrote

An email. A blog post. A chapter. Anything that actually sounds like you.

2. Ask it to analyze the writing

Tone. Sentence structure. Grammar. Have it tell you what it notices before it writes anything new.

3. Give it a new topic and tell it to write in that analyzed style

That’s it. You’re not asking it to “be professional” or “write well.” You’re asking it to copy your patterns.

If you want this to stick, create a Custom GPT with a knowledge base of your past writing so the voice is locked in. You can toggle between formal and casual as needed instead of re-explaining yourself every session.

One more thing: treat the draft as a 70% starting point. Editing is faster than writing from scratch. You’re not looking for a finished piece. You’re looking for something you can fix in a fraction of the time.

Create the Custom GPT today if you already have a pile of old emails or posts sitting around. That’s the whole asset.

Two buckets (this one is from Brooks)

Brooks splits AI into two buckets, and once you hear it you can’t unhear it.

Active AI is anything you explicitly open to do a task. ChatGPT. Copilot. You sit down, you prompt, you get an output.

Background AI is the invisible kind that anticipates what you need. Google Calendar adding a flight from an email is the clean example. You didn’t open a tool and ask. It just happened.

The future of productivity is the growth of Background AI. The less you have to actively do, the better the experience gets.

So audit your workflow for passive automation. Which tools do you need to learn (prompting, Custom GPTs, camera input)? Which ones do you just install and trust? Most people spend all their energy on Active AI and ignore the Background stuff that’s already sitting in the apps they use.

Use the camera. I’m serious.

If text prompts still feel abstract, open the ChatGPT app and use the camera.

Photograph the laundry symbols on a tag instead of hunting for a manual. Photograph the error light on a coffee machine. Photograph a board game state (Catan works) and ask for strategic advice.

You’re turning the phone into a real-time consultant. Research that used to take 30 minutes drops under 5.

Download the app and take a photo of a confusing label. That’s the fastest way to feel why this is useful, because it leaves the chat box and hits a problem you actually have.

Search was matching words. AI is matching intent.

Traditional search engines match keywords. Literal. If the word isn’t on the page, you don’t find the page.

AI tools work off intent and context. Semantic understanding.

Example: you search your notes for “meeting notes with a URL.” Traditional search fails because the word “URL” might never appear in the text. AI understands the concept of a link (http, www, href) and still finds the notes.

That’s why AI search feels smarter than Google when you’re looking through your own stuff. Try searching a knowledge base for a concept, not a keyword, and you’ll see the gap immediately.

If you already have a giant archive in Evernote, check the app for the new AI search. You can ask things like “Find meeting notes from yesterday” or “Notes with app recommendations.” It turns a static filing cabinet into something you can actually talk to.

I’m not putting Evernote in a top-tier recommendation slot. The brand has had performance issues. If you’re starting from scratch, I wouldn’t pick it on AI search alone. If the notes are already there, this feature makes it a viable option again. Try it on the archive you have. Don’t switch your whole system for it.

Do this tonight

Pick ChatGPT. One prompt. “What did I learn today?”

Don’t build a Custom GPT yet. Don’t audit your whole workflow. Don’t migrate out of Evernote. Go from zero to one on a task so small you can’t talk yourself out of it. The forced-adoption wave is coming either way. You might as well have the habit before the operating system makes the choice for you.