How to Actually Make Money With AI (Full Guide)
文章目录
Right now, someone doing the same work as you is doing it faster and better with AI.
In an MIT study, people using AI finished writing tasks in 40% less time, and their work got 18% better. And companies are literally paying for it. Jobs that ask for AI skills pay 62% more than the same jobs without them. For some people, that's 5 to 6 figures more a year.
The people learning these skills now get the high-paying jobs, the high-ticket clients and the raises.
The good news is that most people still haven't figured this out. If you start now, you can still get ahead of 90% of people around you.
This article IS for you if:
-
you can see it's getting more powerful, and you want to be on the side that gets paid for it
-
you use AI every week, but mostly for quick questions and first drafts
-
you want skills that actually turn into money
In this article you'll learn:
-
what the data says about AI and money
-
the 5 AI skills I'd go deep on right now, and how to start each one
-
how to pick the right one for you
-
a 30-day plan and a prompt that builds your full learning plan
(And if you find the article useful, a follow would be really appreciated.)
Alright, let's get into it:
Who's Getting Paid in the AI Era
As I said before, a study by PwC, one of the biggest consulting firms in the world, showed that jobs asking for AI skills pay 62% more than the same jobs without them. They looked at over a billion job ads this year.
They also found that the number of jobs asking for AI skills grew by 69% over the last few years, while all jobs grew by only 9%.
And no, you don't need to work in tech. Lightcast, a company that tracks job market data, found that more than half of the jobs asking for AI skills are outside tech.
But AI doesn't help everyone. In 2025, Stanford found that young workers (22 to 25) are losing jobs that AI can do on its own. In those jobs, their numbers dropped by 13% compared to other jobs.
Jobs where AI helps people work better stayed the same or grew. So if you remember one thing from this article, make it this:
The skills that pay the most are the ones where AI makes YOU more valuable.
Here's a quick test for any skill:
-
do people already pay for this result today?
-
does AI make you faster at it, instead of doing it without you?
-
does it need knowledge of a field, like sales, health or law?
If a skill passes all 3, it's worth learning. Simple as that.
The 5 Skills That Will Pay the Most
Okay, now the interesting part. Here are the 5 AI skills I'd focus on if I were starting today. Each one passes the test above, and each one comes with simple steps to start.
1: AI Workflow Design
AI workflow design means taking the work you already do and doing it with AI. It's the safest skill on this list, because you build on experience you already have.
Let's say you're a marketer. AI can write 50 ads in a minute, and honestly, anyone can do that now.
What's rare is knowing which 3 ads are worth testing, why the other 47 won't work, and what the results mean.
AI can do maybe 80% of the work, super fast. The last 20% is choosing the best version, fixing it and catching mistakes.
This last 20% needs your experience, and it's exactly what people pay for.
How to start:
-
write down every task you do in a normal week
-
split your tasks into 3 groups: AI can do it alone, AI drafts and you fix, AI can't help yet
-
for each task in the middle group, set up an AI assistant (skill 2 shows you how)
-
time each task before and after you start using AI
-
write down what changed, like "this task used to take me 4 hours, now it takes 1"
Most people skip this last step, and I get it, it feels boring. But your notes show you where AI really helps, and they prove you can do the same for other people too.
2: Context Engineering
Most people start every AI chat from zero. They explain who they are and what they want again and again, and get a different result every time.
Context engineering means setting AI up once, so it already knows your work before you type anything.
You can do this with Projects in Claude, custom GPTs in ChatGPT or Gems in Gemini.
Paste this setup into the instructions box of your Project, GPT or Gem. It works for any kind of work, like writing, design, data or code:
ROLE
You are my [job, e.g. "brand designer", "sales assistant",
"data analyst"] for [my business or my work]. You've done this
for years and you're picky about quality.
WHAT WE'RE WORKING ON
[the business or project, who it's for, what makes it different]
WHO THE WORK IS FOR
[the customers or viewers: what they like, what they hate,
what they already know]
WHAT GOOD LOOKS LIKE
[upload 3-5 of your best examples: texts, designs, screenshots, reports]
What makes them good: [e.g. "simple layouts with one bold color",
"short sentences, no jargon", "every number has a source"]
RULES
- always: [e.g. "use our brand colors", "write in British English"]
- never: [e.g. "a stock photo look", "made-up numbers"]
HOW TO WORK WITH ME
- if my request is unclear, ask up to 3 questions before you start
- give me [2-3 options / one finished version] unless I say otherwise
- check everything against these rules and fix what doesn't fit
before you show it to me
- when I correct you, suggest a new rule I can add here, so the
same mistake doesn't happen again
The last rule in the setup is the most important one, by far. Every time you correct the AI, it suggests a new rule, and you add the rule to the setup.
Give it a month, and it's kinda scary how close the AI gets to doing the work the way you would.
Context engineering also makes every other skill on this list easier.
3: AI Automation and Agents
Quick difference first. A normal automation follows fixed steps, like: "when someone fills in a form, send them an email".
An AI agent gets a goal and figures out the steps by itself. It can search the web, read your emails, open your files and check your calendar.
Basically, an agent can do work a person used to do, like research, reports and replies. You only check the result.
Here's what a real agent system could look like for a small business. You set it up once, and every Monday it runs without you:
-
8:00: the agent checks 10 competitors' websites and socials
-
8:05: it spots what changed, like new prices, products and posts
-
8:10: it compares everything with last week
-
8:15: it writes a one-page report with 3 moves worth making
-
8:20: the report lands in your inbox, so you can check it before your team sees it
You can build agents with tools like Claude Code, Grok Bot or Hermes Agent, often with little or no code.
Businesses pay for agents because every hour an agent saves is an hour their team gets back.
How to start:
-
pick one task someone does by hand every week, like a report, research or sorting messages
-
write down every step of the task, including the decisions a person makes
-
build your first agent with the guide below
-
make the agent send its result to you first, and only go further after you approve it
-
run the agent for 2-3 weeks, fix what breaks, and count the hours it saves
If you've never built an agent, start with this guide. It walks you through your first Grok Bot step by step, and Elon Musk called it a "helpful guide to hiring your first Grok Bot employee":
Once the agent works for you, you can sell the same agent to other businesses with the same problem.
4: AI-Assisted Development
With tools like Claude Code, Cursor or Codex, you can build a simple version of an app in one afternoon. Which is insane, but it also means lots of people can do it now.
What people actually pay for is turning that simple version into an app people use every day.
A real app needs logins, payments and a place to save data. You also need to fix bugs and improve the app based on what users say.
How to start:
-
write a one-page plan first: who the app is for, the problem it solves, the 3 features it needs and what you're leaving out
-
build one feature at a time, and test each feature before you start the next one
-
save every working version, so you can go back when something breaks (ask your AI to set this up for you)
-
ask the AI to explain any code you don't understand, because you need to know how your own app works
-
show the app to 5 real people in the first week and watch them use it
If you want to go deeper, this full guide covers the mistakes most people make when they build with AI, and how to fix them. It has over 1M views:
5: Content and Distribution
Now that anyone can build something in a weekend, many more people are competing for the same customers. The ones who win are the people others already know and trust.
Skills 1-4 give you plenty to post about: an agent you built, a task you sped up with AI, the exact setup you use, etc etc.
People save and share posts like these, and the posts prove you can do the work.
AI can help you research and write, but your experience, opinions and results have to come from you. There's no shortcut there.
How to start: pick one platform where the people you want to work with spend time (X or LinkedIn for most people).
Post 3-5 times a week about what you're building and what you're learning.
And if you're starting from zero, this guide walks you through how to grow your audience, step by step:
Content and distribution is the slowest skill on this list. It usually takes a few months before it pays off, and yeah, that part is annoying.
But it makes every other skill worth more, because people start coming to you instead of you chasing them.
Choosing Your Skill Stack
You don't need all 5 skills (please don't try to learn all 5 at once). Most people do best with two skills that work well together:
-
if you have a job you like: learn AI workflow design and context engineering, so you get faster and better at the work you already do
-
if you want to freelance: learn AI automation and agents plus content and distribution, so you can build agents for businesses and get clients to find you
-
if you want your own product: learn AI-assisted development plus content and distribution, so you can build the app and get people to use it
-
if you're not sure yet: start with context engineering, because it makes every other skill easier, then add one more
Your 30-Day Action Plan
-
week 1: pick ONE skill and one real problem you want to solve with it
-
week 2: build the first version, like an AI assistant, an agent or a simple app
-
week 3: use it on real work and write down how much time it saves you
-
week 4: share how you built it, and offer it to 3 people who could use it
A few tips:
-
don't get attached to one tool, because tools change every few months
-
learn by solving real problems, courses alone won't get you there (trust me)
-
someone with two of these skills is much harder to replace
-
give it at least 3 months before you decide if it's working
I could make this list very long, but you get the point.
The AI Learning Plan Prompt
Once you've picked a skill, this prompt builds a learning plan around what you already know. Fill in the brackets, paste the prompt into your AI and answer its questions:
I want to get really good at [the skill] so I can [your goal, e.g.
"get a better-paid job", "sell AI agents to small businesses",
"launch my own product"] within [time, e.g. 3 months].
WHERE I AM NOW
- what I do now: [your job or studies]
- how I use AI today: [e.g. "ChatGPT every day for writing and research,
never built an automation"]
- what I already know about this skill: [be specific]
- tools I already use: [e.g. Claude, Notion, Figma]
- time I can spend: [hours per week]
- budget for courses and tools: [amount, or "free only"]
- how I learn best: [videos / reading / building projects]
Act as an expert who uses [the skill] for paid work every day and has
trained a lot of people to do the same.
Build me a learning plan with:
1. The few parts of this skill that give most of the results
2. A week-by-week plan for [time], with what to learn and one real
project to build each week
3. The best resources for each week (name real courses, channels or
docs, and tell me if you're not sure something exists)
4. 3 portfolio projects that prove I have this skill, from easy to hard
5. The mistakes people at my level usually make, and how to avoid them
6. How I'll know I'm ready: what I should be able to do by the end
7. How people make money with this skill: who pays for it, what they
pay for, and how I can find my first paying client or job
RULES
- skip the basics I already know, start from where I am
- be honest about what's hard and how long things really take
- everything has to fit my time and budget
Before you start, ask me up to 5 questions about anything that's unclear.
After that, I'll tell you every week what I did, and you'll adjust the plan.
Save the chat and come back to it every week. It's honestly like having a personal teacher who always knows where you are.
Okay, quick recap:
-
everyone uses AI now, so the money goes to people who make AI do real work others pay for
-
AI workflow design: do your own work with AI, and show people the time you saved
-
context engineering: set AI up once, and let every correction make it better
-
AI automation and agents: build agents that do boring work for businesses
-
AI-assisted development: turn simple AI-built apps into apps people use every day
-
content and distribution: share what you build, so people know and trust you
-
pick 1 or 2 skills, follow the 30-day plan, and use the prompt to plan the rest
You don't need to know what the world will look like in 5 years. You just need one skill people pay for, and AI to make you better at it.
And that's it :)
If you have any questions, drop them below, I’ll try to reply to every single one. And tell me what I should write my next article about.
