It’s Not the AI. It’s the Brief.

Somewhere in the last few weeks, you tried it. You opened ChatGPT or Claude, typed something like “write a follow-up email to a customer,” and hit enter.

Out came something. Grammatically clean. Professionally bland. It sounded like it was written by a polite robot that had never met you, your customer, or your business. Because, in a sense, it was.

So you did what most owners do. You skimmed it, winced, closed the tab, and quietly filed AI under “not good enough yet.”

Wrong conclusion. Right evidence.

Generic In, Generic Out

If you took Where to Start with AI Without Guessing, you already have the frame: AI is a smart-but-brand-new hire. Sharp, fast, tireless — and it knows nothing about your business until you tell it. You learned to brief it in four parts: role, goal, context, format.

This course picks up where that one left off. Because there’s a specific version of the problem that hits almost everyone right after their first experiment. The output isn’t wrong, exactly. It’s mush. Generic. Flat. It could have been written for any HVAC company, any med spa, any law firm, any restaurant in America. It has your task in it and none of your business.

Here’s the mechanism, and it’s worth thirty seconds, because once you see it you can’t unsee it.

AI models are trained on a staggering amount of writing — more or less everything. When you give a vague instruction, the model has no way of knowing which of a million possible answers you want. So it gives you the middle. The average follow-up email. The average social post. The average job ad. And the average of everything ever written is beige.

That’s not a defect. That’s the machine doing exactly what you asked. You asked for “a follow-up email.” It gave you the follow-up email — the one that belongs to nobody.

Before you go further, test the belief most owners walk in with.

Watch the Same Task, Briefed Twice

Here’s a made-up but true-to-life example. Say you run an HVAC company and you want to follow up on a quote that’s gone quiet.

First attempt — the one most owners actually type:

Write a follow-up email to a customer about the quote we sent them.

And here’s the kind of thing that comes back:

Subject: Following Up on Your Recent Quote Dear Valued Customer, I hope this message finds you well. I wanted to take a moment to follow up regarding the quote we recently provided for your consideration. We understand that making decisions about your home comfort needs takes time. Please don’t hesitate to reach out if you have any questions. We would be happy to assist you further.

Nobody reads that email. Nobody answers that email. You wouldn’t send that email — which is exactly why you concluded AI can’t write your emails.

Now the same task, same AI, same day — briefed properly:

You're writing as the owner of a family HVAC company in Mesa, Arizona. Friendly, plain-spoken, zero corporate filler.

Context: Twelve days ago we quoted Sarah $9,400 to replace her 14-year-old AC unit before summer. She said she needed to talk to her husband. It's forecast to hit 100°F this week. We have two install slots left in the next two weeks, and we offer 0% financing for 18 months — she may not know that.

Here's a follow-up I wrote last year that actually got a reply. Match this tone: "Hi Mike — no pressure at all, just wanted you to know we've got one crew opening Thursday if you want that water heater in before the weekend. If the timing's off, no sweat. — Dave"

Write the email to Sarah. Under 90 words. One clear next step: reply or call. Mention the financing once. Do not use the phrases "I hope this finds you well," "valued customer," or "don't hesitate."

The output from that second prompt is an email you could send with maybe one word changed. Same AI. Same task. Same afternoon. The entire difference is what’s wrapped around the ask.

The Three Missing Ingredients

Look at what the second brief added. Not padding. Not magic words. Three specific ingredients:

Every disappointing AI output you’ve ever gotten is missing at least one of these three. Most are missing all three at once.

And here’s the good news hiding in that sentence: the fix isn’t a new tool, a smarter model, or a weekend course in “prompt engineering.” It’s a two-minute upgrade to how you ask. You already know how to do this, because you already brief humans this way. A new tech. A new front-desk hire. A subcontractor. Nobody hands a new hire a sticky note that says “do marketing” and walks away — and if somebody did, we’d blame the sticky note, not the hire.

What This Course Will Do

The next three lessons take one ingredient each. Context in lesson two. Examples in lesson three. The standard for done in lesson four. Each one covers what the ingredient actually is, where to find it in your business in under five minutes, and a worked example you can steal and adapt.

Lesson five hands you a 60-second diagnostic for the moment output comes back bad: which of the three is missing, and how to fix it in the same chat instead of rage-quitting. Lesson six turns your best briefs into a library, so you stop rewriting them from scratch every time.

By the end, “AI output looks like garbage” stops being a verdict on AI and becomes a checklist question: which ingredient did I skip?

Your Test Case

Before you move on, pick your test case — one real task where AI let you down, or one you’ve been afraid to try. You’ll rebuild this exact task, ingredient by ingredient, over the next three lessons, and watch the output change draft by draft.

If you’ve never tried at all, no problem: borrow the quote follow-up from this lesson, or pick the message you dread writing most every week. The dreaded ones make the best test cases.

Next up: ingredient one. Context — and why it doesn’t mean writing longer prompts.