If your prompt sounds like something you'd type into a search bar, we need to talk.
I have watched people open ChatGPT and type:
“Write me a process.”
Then AI gives them something painfully generic.
And they're disappointed.
Baby, what exactly did we give it to work with?
That's like walking into a restaurant, telling the waiter, “Bring me food,” and getting mad when a turkey sandwich shows up.
Technically, they did what you asked.
This is why one of the first things I teach people about prompting is role-based prompting.
Before you tell AI what to do, tell it who the hell it's supposed to be.
Give the Robot a Job
Think about how you would delegate something to an actual person.
If I need somebody to write a newsletter, I want a newsletter writer.
If I'm working through an operations problem, I want an operations strategist.
If I need a blog written, I want a blog writer who understands the brand, audience, voice, and purpose of that blog.
So my prompt might begin:
“You are an Operations Strategist supporting a seven-figure founder.”
We've already given AI more direction than:
“Make me an SOP.”
But the role is only the beginning.
Now Tell It What the Hell You Want
Once I've assigned the role, I define the job.
Let's say my client sends me a voice note explaining how her team handles onboarding.
It's messy.
She jumps around.
Half the process exists in her head.
Welcome to operations.
I could ask AI:
“Write an onboarding process.”
Or I could say:
“Review this workflow description and turn it into a step-by-step SOP.”
Much better.
Now it knows what source material it's working from and what I'm expecting it to produce.
But we're still not finished.
Put Some Guardrails Around It
This is the part people skip.
Constraints.
Who owns each step?
What tools are involved?
Are there deadlines?
What shouldn't AI change?
What information has to remain exactly as supplied?
How detailed should the answer be?
Who is going to use the finished document?
Constraints aren't there to make the prompt unnecessarily complicated.
They're there because context matters.
If I'm building something for a client, I don't want AI wandering off and making assumptions because I couldn't be bothered to explain the assignment.
I want it working inside the brief.
Tell It What the Finished Thing Should Look Like
This one will save you so much editing.
Don't just ask for information.
Tell AI how you need that information delivered.
Maybe I need:
A checklist the client's team can follow tomorrow.
A table with owners and deadlines.
A client-facing SOP.
A short email.
A 30-day content calendar.
A decision tree.
Those are very different deliverables.
If I know what I need, I'm going to say so.
Otherwise, I might get six paragraphs when what I actually needed was something my client's team could check off line by line.
Now I'm editing something I could've specified from the beginning.
No, thank you.
And What Are We Actually Trying to Accomplish?
This might be my favorite part.
Tell AI why you're doing the task.
Not some dramatic mission statement.
The actual business outcome.
Maybe the SOP needs to reduce onboarding mistakes.
Maybe the checklist needs to help a new team member complete the process without asking the founder 12 questions.
Maybe the newsletter needs to drive people to a specific offer.
Maybe the research needs to help your client make a decision.
That outcome changes how I construct the prompt.
Because I don't just want something that looks finished.
I want something useful.
So What Does the Whole Thing Look Like?
Instead of:
“Write me a process.”
We can build:
“You are an Operations Strategist supporting a seven-figure founder. Review the workflow below and turn it into a step-by-step SOP. Include the owner, tools, and timeline for each step. Format it as a checklist the client's team can use immediately. The goal is to reduce onboarding errors and allow the team to execute without relying on the founder.”
Now we're talking.
Role.
Task.
Constraints.
Format.
Outcome.
That's the anatomy.
And once you understand those pieces, you can use the same thinking for almost anything.
Content.
Research.
Client communications.
Operations.
Project management.
Documentation.
Strategy.
You don't need to memorize one giant magic prompt.
You need to understand how to build one.
This Is Why I Care About VAs Learning AI Properly
I don't want VAs learning AI because it's the trendy thing everybody is talking about.
I want you learning it because your job already requires you to take information, understand what somebody needs, organize it, solve problems, and communicate clearly.
Prompting is awfully familiar when you look at it that way.
You're delegating.
And good delegation has never been:
“Hey, do the thing.”
You provide context.
You explain the outcome.
You establish boundaries.
You define what finished looks like.
Then you review the work.
AI doesn't eliminate those skills.
It makes them even more valuable.
The Prompt Is Only as Smart as the Person Giving Directions
This is the part I want people to understand.
You don't become valuable because you know how to open ChatGPT.
Everybody can open ChatGPT.
Your value is knowing what to ask.
Knowing what information matters.
Knowing when the answer is wrong.
Knowing what context is missing.
Knowing how to take an output and turn it into something a client can actually use.
That's why I don't think the goal is to “master AI.”
Learn how to direct it.
That's a much more useful skill.
And the next time AI gives you some bland-ass answer?
Before blaming the robot, go look at what you asked it.
You might've ordered the turkey sandwich.
CURRENT MOOD
“Upgrade U” by Beyoncé featuring Jay-Z.
Because that's really what we're doing here.
We're not throwing away the skills you already have.
We're upgrading how you use them.
You already know how to communicate.
You already know how to delegate.
You already know how to understand a client's needs.
Now learn how to translate those skills into better instructions for AI.
Same brain.
Better tools.
Listen to “Upgrade U” by Beyoncé featuring Jay-Z on YouTube
PEAK OF THE WEEK
Realizing that after all these years of delegating to humans, a lot of what I know about prompting came from the exact same place.
Give clear directions.
Provide context.
Explain what good looks like.
Check the work.
Apparently I've been training for this AI shit longer than I thought.
#GiveTheRobotAJob
XXX,
M💋
0 comments