I’m often asked how to get started with AI. It’s easier thank you think.
Insight: AI is the easiest tool to learn…you’ve been ready since you were 2 years old.
So what? Stop overthinking and just start talking. Messy thoughts are the best way to just start using AI.
I was running a workshop last month, and afterward one of the transit directors pulled up ChatGPT on her laptop. She stared at it like it was a tax form. Then she started typing, very slowly, one word at a time: "How... do... I..." She got about five words in and stopped, backspaced, tried again. This went on for a solid thirty seconds.
Then she turned to me, threw her hands up, and just started talking: "Okay, what I actually want is to take this board report from last quarter and pull out the stuff that shows my team is actually working miracles here — because I've got a meeting Thursday and my Chair keeps grilling on what we've actually accomplished and I know it's in there somewhere but I don't have time to dig through forty pages of ridership stats and make it pretty and just get him off my back."
I'd quietly hit the voice record button on my phone about halfway through. I handed her the transcript — typos, run-on sentences, and all. "Send that. Exactly that."
She looked at me like I'd suggested she Facetime her boss in her pajamas.
But here's the thing: Google trained us to curate our thoughts. Five keywords max, no fluff, think before you type. But AI is the opposite. It actually works better when you give it—as any writer calls it—the “shi*ty first draft.” The rambling, unpolished, "I haven't figured out what I'm asking yet" version? That's the prompt that actually gets you out of analysis paralysis, gets you out of your own head, gets you into real action.
It’s as easy as being 2 years old again.
Here's proof. In drafting this post, Claude asked me to give an example of someone overcoming this mental block. The story above all came when I started dictating this messy, ugly beast (reader be warned):
One moment or example I was meeting with a transit leader recently who, after a workshop, they were experimenting with ChatGPT or Claude and they were being very careful and typing very slowly, only like 5 words because they weren't quite sure how to ask it. Then they turned to me and asked, "What I really want is to be able to analyze this board report to show so-and-so that I'm actually doing a really good job blah blah blah." When they started talking to me, I just held down my recording button and it pasted their entire transcript and I gave that to them and said, "Send that directly to the AI," and that was the moment where they realized they've been overthinking.
Just like we've been taught with Google (use 5 words in a query is better than 500), but with AI you don't have to curate your thoughts and we have to sort of untrain and unlearn those things. It was that moment where I'm seeing it was with many people. People don't know how to ask them, so they turn to me and they just start voicing exactly what they want. My answer to them is literally exactly what you just said to me is what you should be sending to AI.
So then the contrast to draw is probably like with old Google vs the new one. You're used to curating your words very tight, but actually you can get out your shitty first draft as a journalist as soon as possible and then trim it down. My tone is always encouraging and warm.
A couple iterations later, Claude and I had 90% of this written. My hardly-coherent rambling became a fun little story. Insane, right?
You've been asking questions and rambling since you were two years old. You don't need a certification to use AI, you just need to stop editing yourself before you start. So if you’re wondering how to get started with AI, my advice to you is embrace the sh*tty first draft.
Get unstuck and just get moving.
Transit Intelligence is where I share practical AI insights for transit leaders. Subscribe if you haven’t. Forward to a colleague who needs it. And hit reply — I read everything.