Playing with AI is fun. Knowing how to make it useful is better.
I recently appeared on the Joyful Business Podcast, and naturally, AI came up.
When the episode drops, I’ll share it here. I already know I’m going to have another think piece after I listen back because there were parts of that conversation I want to dig into separately.
But one thing has been sitting with me since we recorded.
I think we're making learning AI harder than it needs to be.
You don't have to learn everything.
Actually, please don't.
You will drive yourself fucking crazy.
Every week there's another tool, update, agent, model, feature, acronym, or person on the internet telling you that if you haven't mastered it by Thursday, your career is over.
Relax.
You don't need all of AI.
But you do need to get past the surface.
Cute Prompt. Now What?
Most people start with the fun stuff.
Write me an Instagram caption.
Draft this email.
Make me a picture.
Give me 20 content ideas.
Great.
That's how a lot of us started experimenting.
But if you're a freelancer and that's still the extent of what you can do with AI, I want you to go further before 2026 is over.
Because the opportunity isn't becoming the person who knows every AI tool.
It's becoming the person who knows how to make AI useful inside actual work.
And there are four things I'd focus on first.
1. Give AI a Damn Job
When you're using AI for writing, stop throwing naked instructions at it.
Give it a role.
Instead of:
“Write a post about personal branding.”
Try approaching it as:
“You are a content writer specializing in personal brands for service-based entrepreneurs...”
Writing a newsletter?
Give it the role of a newsletter writer and tell it who the newsletter is for, what its job is, what voice it should use, and what information it has to work with.
Writing a blog?
Same thing.
“You are a blog writer for...”
Then give it context.
Audience.
Purpose.
Voice.
Source material.
Constraints.
Examples.
Desired output.
A role alone isn't magic. But it gives the model a much better starting point than tossing one sentence into the box and hoping it reads your mind.
Because it can't.
And frankly, neither can your clients.
2. Stop Asking for “A Pretty Picture”
Image prompting is another place where people stop too early.
“A beautiful Black woman working at her desk.”
Okay.
What kind of lighting?
What camera?
What lens?
What angle?
What's happening in the background?
What's she wearing?
What time of day is it?
What does the image need to feel like?
Editorial?
Cinematic?
Candid?
Vintage?
High-fashion campaign?
Early-2000s music video?
Those details matter.
I think about AI image prompting more like directing a photoshoot.
You're not just describing the person.
You're directing the scene.
Tell it the lighting.
Tell it the lens.
Tell it the camera perspective.
Tell it whether you're shooting close-up, overhead, low-angle, full-body, wide, or over-the-shoulder.
Tell it the environment.
Tell it the mood.
Give it enough visual direction to understand what you're seeing in your head.
“Make it cute” is not art direction.
3. Your Agent Does Not Need 19 Jobs
Agents are where things get really interesting.
They're also where people can build themselves a beautiful little digital disaster.
If you're starting to build agents, give each one a defined responsibility.
Don't create one mega-agent responsible for content strategy, customer service, lead generation, invoices, research, onboarding, your grocery list, and finding your next husband.
That's too much responsibility for anybody.
Including the robot.
Decide what the agent owns.
Decide what information it needs.
Decide what tools or sources it can use.
Decide what requires human review.
Decide where its responsibility ends.
Then test the hell out of it.
I'd rather have several smaller agents with clearly defined jobs than one giant mystery machine where I have no idea which instruction caused something to go sideways.
And when something does go sideways?
I want to know exactly where to look.
4. Document the Damn Thing
This might be the least sexy part of AI.
It's also one of the smartest.
Keep a knowledge base.
Save your prompts.
Document your instructions.
Keep your source files organized.
Record what changed.
Keep examples of good outputs.
Write down why something exists.
Because three months from now, when you decide to change one piece of the system, you don't want to be staring at something you built thinking:
What the fuck did I do here?
Documentation makes changes easier.
It makes troubleshooting easier.
It makes testing easier.
And for freelancers, there's another reason I care about this so much.
That documentation can become money.
Build It for Yourself Before You Sell It to Somebody Else
This is one of my favorite parts.
Become your own proof of concept.
Build the thing.
Use it.
Break it.
Fix it.
Test it again.
Document what you learn.
Now you have more than an AI skill.
You have a process you've actually used.
And eventually, that process may become something you can offer a client.
Maybe you implement an AI-supported content workflow.
Maybe you build an internal knowledge base.
Maybe you create an agent for a specific business function.
Maybe you help a client document and improve something they're already doing.
When it's time for knowledge transfer, you're not scrambling to explain what you built.
You already documented it.
And if you notice you've solved the same problem enough times?
There might be an offer sitting there too.
That's where I want freelancers thinking bigger.
Don't just learn AI so you can say you know AI.
Learn something useful.
Use it in your own business.
Get good at it.
Document it.
Then figure out whether somebody would pay you to help them do the same.
I Made the Prompt Part Easier
I've spent the last few years experimenting with AI while also working with freelancers and helping people build businesses and brands.
And I kept seeing the same problem.
People knew AI could help.
They just didn't always know what the hell to ask it.
So I started collecting the prompts and structures that made AI more useful for actual business and content work.
Eventually, that collection became The Prompt Vault: AI Prompts for Business + Content.
It isn't about memorizing everything AI can possibly do.
It's a starting point.
Because I would much rather give you something useful to build from than have you reinventing the wheel every time you open a new chat.
The Prompt Vault: AI Prompts for Business + Content
You Have Until the End of 2026
Not to become an AI expert.
Not to learn every tool.
Not to build an agent army.
Not to start calling yourself an AI consultant because you made one really good picture.
Just go deeper.
Learn how to write better instructions.
Learn how to direct an image.
Learn how to give an agent a defined job.
Learn how to document what you build.
Pick something.
Get good at it.
Then use it.
Because knowing a little bit about 75 AI tools doesn't impress me nearly as much as knowing how to make one of them solve a real problem.
CURRENT MOOD
“She Works Hard for the Money” by Donna Summer.
Because that's exactly where my head is with AI.
If I'm going to spend my time learning this shit, it needs to work.
Not just make cute pictures.
Not just write captions.
Not just give me another shiny tool to play with for three days before I forget the password.
I want AI helping me build something useful.
Something I can test.
Something I can document.
Something I can use in my own business and eventually turn into something valuable for somebody else's.
Work smarter.
Make the robot work harder.
Watch “She Works Hard for the Money” by Donna Summer on YouTube
Everything else in the blog stays the same.
PEAK OF THE WEEK
Walking away from the Joyful Business Podcast thinking about all the things I still want to say about AI.
That's usually how I know I enjoyed a conversation.
The recording ends.
My brain does not.
So when the episode drops, I'll share it here.
And knowing me, there will probably be another XXX File waiting right behind it.
#MakeTheRobotUseful
XXX,
M💋
0 comments