Back to all articles

AI Strategy

From AI overwhelm to AI mastery: what I’d do if I were starting today

What I’d do if I were learning AI today: work deeply with one AI, give it real context, build repeatable systems, and turn ideas into things you can use.

By Mollie Amkraut Mueller · · 7 min read

From AI overwhelm to AI mastery: what I’d do if I were starting today, with a steps motif

Updated September 2026 to reflect how I’d recommend learning AI today.

“I know I need to get better at AI. I just don’t know where to start.”

I hear some version of this all the time.

A year ago, my advice was basically: try ChatGPT, experiment with some image tools, play with an agent, build something in Lovable.

I still believe in learning by doing. But after another year of using AI constantly, my advice has changed.

I would spend much less time trying to learn “AI,” and much more time learning how to work really well with one AI.

You do not need to understand every new model. You do not need a giant prompt library. And you definitely do not need to try every tool people are yelling about on LinkedIn.

If I were trying to get good at AI from scratch today, here’s exactly what I’d do.

1. Pick one good AI and use the hell out of it

ChatGPT or Claude. Pick one.

Yes, they have differences. Yes, people on the internet will happily spend 4,000 words telling you which benchmark one model beat another one on.

For most people learning AI, this matters much less than you think.

What matters is using one of them enough that it becomes a normal part of how you work.

For the next few weeks, bring it into everything:

The goal is not to memorize what AI can do.

It’s to develop the instinct to think: “Could AI help me with this?”

That instinct is one of the biggest differences I see between people who occasionally use AI and people who are getting enormous leverage from it.

2. Stop treating it like Google

The most common mistake I see is people giving AI a two-sentence prompt and expecting magic.

They type:

“Write a strategy for my company.”

They get a generic answer.

And they conclude that AI is kind of mediocre.

I do almost the opposite.

I give AI the mess.

The half-formed idea. The document I’m reacting to. Screenshots. Background on the business. What I tried already. What I hate about the obvious answers. The weird constraint that seems irrelevant but actually changes everything.

Think about how you would bring a smart colleague into a problem.

You wouldn’t walk into their office, say “develop my strategy,” and then silently judge whatever came out of their mouth.

You’d talk.

Treat AI more like a teammate and less like a search box.

And don't worry nearly as much about writing the perfect prompt.

One of the most valuable skills you can build is simply getting good at reacting to what AI gives you:

“Too generic.”

“Number three is interesting. Take that further.”

“I hate all of these. Here’s why.”

“You’re making this too complicated.”

“This doesn’t sound like me.”

That back-and-forth is where a lot of the magic happens.

3. Make giving work to AI ridiculously easy

This one has changed how much I use AI more than almost anything else.

Typing can become a bottleneck.

So I use voice constantly.

I talk to AI while I’m walking. I dictate a messy thought instead of organizing it first. I brainstorm out loud. I hand over an idea while it’s still forming instead of waiting until I’m back at my computer and have time to write the Perfect Prompt™.

The easier it is to hand something to AI, the more often you’ll think to use it.

So remove friction.

Put the app on your phone. Put it on your desktop. Use voice. Upload the file instead of explaining what’s in it. Send the screenshot. Drop in your rough notes.

Get the distance between “I have a thought/problem/task” and “AI is helping me with it” as close to zero as possible.

4. Give it access to your actual world

AI gets dramatically better when it has context.

Not abstract context. Your context.

Your business plan.

Your past writing.

Your customer research.

Your meeting notes.

Your calendar.

Your project documents.

Your previous thinking on the same problem.

This is also why I still recommend something like Granola as an easy early AI win. Your meetings contain an enormous amount of useful information that normally disappears the second everyone leaves the Zoom.

Start capturing that information.

And when appropriate, give your AI access to the materials it needs to help you properly.

One caveat: obviously use judgment around confidential, sensitive, personal, or company information. “Give AI more context” does not mean “blindly upload everything you have.”

But in general, the more AI understands the world you’re operating in, the less it feels like a clever chatbot and the more useful it becomes.

5. When something works, turn it into a system

This is probably the biggest upgrade from how I used AI a year ago.

Don't keep having the same great conversation over and over.

If you figure out a good way to use AI for something you do repeatedly, save it.

Maybe that becomes:

If AI helps you prepare brilliantly for one sales call, ask: how do I make this happen before every sales call?

If it gives you a useful weekly analysis, ask: how do I get this every Friday without rebuilding it from scratch?

If you develop a great process for turning meeting notes into follow-ups, make that the process.

This is where you start moving from using AI to changing how you work because AI exists.

6. Build something, even if you are “not technical”

This is the part I wish more non-technical people would try.

Go to something like Lovable and build a tiny app.

Not because you need to become a software developer.

Because building something changes your mental model of what is possible.

Make a calculator you wish existed. A little assessment. A tool for your team. Something for your kids. A dashboard. A silly game. A prototype of an idea you’ve had for years.

The important thing is that it should be something you actually want, not a coding exercise somebody assigned you.

Building with AI is one of the fastest ways I know to experience the shift from:

“I can use AI to help me do my work”

to:

“Wait. I can make things I literally could not make before.”

That is a much more profound skill shift.

7. Stay informed without turning AI into a second job

You do need some awareness of what’s changing.

You do not need to know everything that happened in AI today.

I’d pick:

Then stop.

There will always be a new model, tool, feature, agent framework, benchmark, or viral demo.

Chasing all of them can actually make you worse at AI because you spend all your time starting over.

Depth beats novelty.

Get really good at working with one AI first.

Then expand when you have a reason to.

The goal isn't AI mastery

Actually, I don't love the word “mastery” anymore.

This stuff changes too quickly for any of us to master it.

What I want instead is agency.

I want you to see a new capability and think, “Interesting. I could probably figure out how to use that.”

I want you to have an idea and feel capable of making it real.

I want you to encounter a problem at work and have an entirely different set of tools available for solving it.

And I want you to be good enough at AI that you're participating in the decisions about how it changes your work, rather than having those decisions made around you.

You don't get there by reading another 50 articles about AI.

You get there by using it.

So if you’re feeling behind, don't start by trying to catch up on everything you've missed.

Pick one AI. Give it a real problem. And start.