People ask me how I learned to do the things I do with AI. Building working applications, wiring up automations, all of it, without a computer science degree.
The honest answer is slightly embarrassing. I asked AI.
Josh and I got to this independently on the podcast, from opposite directions, and landed in the same place. If you are stuck at the starting line, the move is to describe your job to an AI and ask how you could use AI in it.
Why this works better than it should
Most advice about getting started with AI fails for one reason: it's generic.
"Use it to summarize documents." Fine. Which documents? Yours are a specific shape, arriving in a specific way, needing a specific thing done to them. The gap between generic advice and your Tuesday is exactly where good intentions die.
Every job is different. Two accountants at two firms do genuinely different work. An article written for both of them is useful to neither.
But specificity is what these tools are actually good at. Give it the details of your actual role and it will produce a list aimed at your actual work. That isn't a party trick. That is the thing it does best.
How to do it properly
Do not type "how can I use AI." You will get a listicle.
Give it the real texture of your week:
I'm a bookkeeper at a small firm with about a dozen clients. Most of my time goes to categorizing transactions, chasing missing receipts, and preparing monthly reports. Clients send me documents by email, usually as photos, sometimes with handwriting on them. What are a few specific ways I could use AI in this work, and for each one, tell me what could go wrong and what I would need to check?
That last clause matters more than the rest. Ask for the failure modes alongside the suggestions. You will get a more honest list, and you'll find out which ideas require caution before you have built anything on top of them.
Then do one. Not five. One, this week, on something small.
The loop
The part that surprises people is that it compounds.
You ask AI how to use AI. You try one thing. Trying it teaches you what these tools are shaped like, which makes your next question sharper, which produces a better answer.
It is a strange loop and it does work. It is how I learned, and I still use it whenever I'm somewhere unfamiliar.
The one caution
Everything on that list needs checking, because the tool will confidently suggest things that do not work.
It might recommend an integration that doesn't exist. It might describe a feature from a year ago that has changed. It might suggest something technically possible and completely inappropriate for your data.
So treat the output as a list of leads, not a plan. Especially if the suggestion involves client information, in which case the question is not "can I" but "should I, and where does that data go."
Why I keep giving this advice
Because the barrier for most people isn't capability. It is not knowing that the door is unlocked.
There is a real gap between people who poke at new tools and people who put them down and get back to work. That gap isn't intelligence and it isn't age. Often it is just that nobody ever showed them a first step small enough to take on a busy day.
This is that step. It costs ten minutes, it requires no purchase, no training course, and no permission from anyone.
You don't need to become a technical person. You need one specific idea, aimed at work you already understand, small enough to try before lunch.
Go describe your job to it and ask what it would do.