You've done it. You typed "write me a blog post about spring lawn care" into ChatGPT and got back 400 words of nothing. Generic advice. No voice. Could've been written by any lawn company in any state. You copy-paste it anyway because you're busy, and it sits on your website doing absolutely nothing for you.

That's an input problem, not a ChatGPT problem.

Why most prompts fail

"Write me a blog post" gives the model nothing to work with. No audience, no voice, no point. So it gives you the most generic version it can come up with. It reads like Wikipedia wrote it after a long shift: bland, correct, forgettable.

The model isn't bad at writing. You're bad at asking. If you want the short version of what's actually happening under the hood when you type a prompt, this breakdown of what an LLM is covers it in plain terms. Every complaint we hear about AI output ("it sounds fake," "it's too generic") traces back to a one-line prompt. Better input means better output, almost every time.

The fix is a better habit, and it costs about fifteen extra seconds per prompt.

The Context-Constraint-Goal framework

We use three questions with every client, on every prompt, for every task, because completeness beats cleverness every time.

Context. Who are you, who's reading this, and what's already happened? "I run a 12-person landscaping crew in Gilbert. This customer has paid on time for three years." That single sentence tells the model more than most people put in an entire prompt.

Constraint. Tone, length, what to include, what to leave out. "Keep it under 150 words. Warm, not corporate. Don't threaten legal action." Constraints work like a fence, keeping the model from wandering into generic-land.

Goal. What should the reader do or feel when they finish? "I want them to pay within a week and still feel good about us." Without one, the model is guessing at what "done" looks like.

Stack all three and a two-line request becomes a real brief. That's the whole framework, and it doesn't require a new app or a course.

Real examples from a landscaping business

We built this framework working with a landscaping company in the East Valley. Here's what it looked like in practice.

A longtime customer was 60 days late on an invoice. The owner didn't want a form collections letter. He wanted to keep the relationship. The prompt: "I run a landscaping company. This is Dave, a customer of three years who's always paid on time. His invoice is 60 days overdue, $340 for two mowing visits. Write a short, warm email reminding him: no threats, no legal language, under 120 words. I want him to pay this week and still feel like a valued customer, not someone we're chasing." That's context, constraint, and goal in four sentences. The email that came back sounded like a person who knew Dave, not a collections bot.

The crew lead used to text a client a two-line scope note and call it an estimate: "front and back yard cleanup, mulch install, trim three trees." Now that note becomes the context for a prompt that also specifies constraint (one page, itemized, plain language, no jargon) and goal (the homeowner should understand exactly what they're paying for and feel confident signing). The output is a proposal that looks like it came from a company with an office, not a guy with a truck.

Crew leads talk faster than they type, so one of them records a two-minute voice memo walking through how they winterize an irrigation system, runs it through a transcription tool, then feeds the transcript into a prompt: "This is a voice memo from our senior crew lead explaining how we winterize irrigation systems. Turn it into a step-by-step SOP a new hire could follow without supervision. Number the steps. Flag anything that's a safety issue. Keep it under one page." Context is the transcript itself. Constraint is the format. Goal is that a first-week employee can follow it alone.

Same three questions. Three completely different documents.

What changes when you use it

The output sounds like the business on the first try. You're not rewriting half of it or sending it back for a fifth pass. That's the actual time savings people miss when they talk about AI productivity: you stop editing your way out of a bad prompt.

Once the framework is muscle memory, it stops feeling like a framework. You just start writing prompts that include who you are, what you need, and why, because leaving those out now feels obviously incomplete, the same way sending an email with no subject line would.

And it scales past writing tasks. Pricing a job, drafting a policy, summarizing a meeting, answering a Google review: the same three-part scaffold covers all of it. You're not learning a new trick for every use case. You're learning one habit that applies to hundreds of them.

Make it stick

Write the three questions somewhere you'll actually see them: a sticky note on the monitor, the notes app on your phone, whatever gets them in front of you before you start typing. If you've tried this before and still ended up with something you had to rewrite, here's what usually went wrong and what to do differently. If you want help building this into how your team already writes emails, estimates, and SOPs, reach out and we'll walk through it on your actual work, no pitch attached.