I have a slightly embarrassing habit when I'm testing prompts.
I keep score.
Not in a formal way. Just a running note on my phone where I write down how many edits I had to make after each prompt. It started because I kept feeling like one prompt was better than another without any real evidence, and feelings are a terrible way to judge anything.
So now I count. Every edit is a mark. Fewer marks wins.
This post is one of those tests, written up properly. Same task, three different prompts, and an honest count of what each one cost me in editing time.
I picked a task I do constantly: turning rough notes into a short update email for my manager. It's the kind of thing that should take five minutes and somehow takes twenty.
Let me show you all three prompts, all three outputs, and the final tally.
The Task
Here's what I was working with. This is a real set of notes from a project check-in, with names and specifics changed.
"carrier delay again - 3 days this time - customer not happy, wants update
we can partially ship Friday, rest Monday
Mike checking with the warehouse
need to tell the customer before they ask again
Dana wants to know if this affects the Q3 numbers
probably does but small
also the template rollout got pushed a week
follow up next Tuesday"
Nine lines. Fragment soup. About 90 words. And I needed to turn it into a short update email for my manager, who has roughly thirty seconds to read anything I send her.
That's the task. Now let's see what three prompts did with it.
Prompt One: The Lazy Version
This is what I used to write. I'm including it because I want you to see the difference, and because I suspect most people are still here.
"Summarize these notes into an email for my manager."
That's it. That's the whole prompt.
What Came Back
Subject: Project Update
Hi [Manager],
I wanted to provide a quick update on the current status of our project. There has been a carrier delay of approximately three days, which the customer has expressed some concern about. We are working on a partial shipment for Friday, with the remainder expected to go out on Monday. Mike is coordinating with the warehouse to confirm details.
Dana has raised a question about whether this will impact our Q3 numbers. It likely will have a minor impact, though the extent is still being assessed. Additionally, the template rollout has been pushed back by one week. We will follow up next Tuesday.
Please let me know if you have any questions.
Best,
Alex
My Reaction
It's not wrong. That's the frustrating part. Everything in there is accurate.
But look at what it cost me.
Edits I made:
Deleted "I wanted to provide a quick update" — filler.
Deleted "Please let me know if you have any questions" — filler.
Rewrote "has expressed some concern about" — softer than reality. The customer was annoyed, and my manager needs to know that.
Changed "approximately three days" to "three days" — I know the number, why hedge?
Rewrote "It likely will have a minor impact, though the extent is still being assessed" — this is nine words of nothing. The actual answer is "probably yes, small."
Moved the template rollout line — it's a separate item and shouldn't be buried in the same paragraph as the customer issue.
Cut the whole thing from 145 words to about 90.
Total edits: 7. Time spent editing: about 12 minutes.
Which means this "time-saving" prompt cost me more time than writing the email myself would have.
Prompt Two: The Better Version
For this one I added a little structure. Nothing fancy—just told it who, what, and how long.
"Turn these notes into a short update email for my manager. She's busy and wants the key facts fast. Keep it under 100 words. Use bullets for the items. Don't add anything that isn't in my notes."
What Came Back
Subject: Quick Update — Carrier Delay & Template Rollout
Hi [Manager],
Two updates:
• Carrier delay of 3 days. Customer is unhappy and waiting for an update. We can partially ship Friday, remainder Monday. Mike is confirming warehouse details.
• Q3 numbers: Dana asked about impact. Likely small, but real.
• Template rollout pushed back one week. Following up next Tuesday.Alex
My Reaction
Much better. The bullets work. The length is right. It reads like something a person would actually send.
Edits I made:
Deleted "Two updates:" — a header for three bullets is unnecessary.
Moved the Q3 line out of the middle — it's a question, not an update, and it reads oddly as a bullet next to two concrete items.
Changed "Customer is unhappy and waiting for an update" to "Customer is unhappy — needs update today." Tighter, and it makes the urgency explicit.
Added the word "today" — ChatGPT couldn't know that, and it's the most important detail in the email.
Total edits: 4. Time spent editing: about 5 minutes.
Better. Real improvement. But still not free.
Prompt Three: The Five-Part Version
Here's where I applied the structure from my Five-Part Prompt post. Role, Task, Context, Constraints, Format.
Role: You are an operations coordinator writing a status update to a manager.
Task: Turn my raw notes into a short update email.
Context: My manager is extremely busy and reads email on her phone. She cares about three things: what changed, what's at risk, and what she needs to know. She dislikes long emails and dislikes hedging language. The customer situation is urgent; the template rollout is not.
Constraints: Keep it under 90 words. Lead with the most urgent item. Do not add information that isn't in my notes. Do not use phrases like "I wanted to provide" or "please let me know if you have questions." Do not hedge — if something is uncertain, say "unclear" rather than softening it.
Format: A one-line subject, then a short opening sentence, then three bullets max. No closing paragraph.
What Came Back
Subject: Carrier delay — customer needs update today
Hi [Manager],
Customer is unhappy about a 3-day carrier delay and is waiting on an update. Partial shipment goes Friday, remainder Monday; Mike is confirming warehouse details.
• Q3 impact: likely small — unclear how small
• Template rollout: pushed one week, following up Tuesday
• Need your input: none yetAlex
My Reaction
This is the one.
Edits I made:
Deleted the "Need your input: none yet" bullet — it's technically true but adds nothing.
Total edits: 1. Time spent editing: about 90 seconds.
That's it. One deletion and it was done.
The Score
Prompt | Words In Output | Edits | Editing Time |
|---|---|---|---|
Lazy | 145 | 7 | ~12 min |
Better | 88 | 4 | ~5 min |
Five-Part | 61 | 1 | ~90 sec |
I want to be honest about what this doesn't show, though.
The prompt writing time differs. The lazy prompt took me five seconds to write. The five-part prompt took me about two minutes. So the real comparison isn't "12 minutes versus 90 seconds"—it's "12 minutes versus 90 seconds plus 2 minutes of prompting."
Even with that added, the five-part version wins by a wide margin. Two minutes of prompting saved me more than ten minutes of editing.
One test isn't proof. I've run this pattern a handful of times now, and the results are consistent, but three prompts on one task is a single data point. Your task might behave differently.
The five-part version isn't always the right choice. If I'm writing something throwaway—a two-line note to a coworker—I don't build a full prompt. The structure pays off when the output matters and you'd otherwise spend real time fixing it.
Why the Third One Won
It's not because the prompt was longer. It's because the prompt made decisions so ChatGPT didn't have to guess.
Look at what the five-part version specified that the others didn't:
Who's reading it — a busy manager on her phone
What she cares about — what changed, what's at risk, what she needs to know
Tone rules — no hedging, no filler phrases
Priority order — most urgent first
Explicit no-go list — the exact phrases to avoid
The lazy prompt left every one of those to chance. ChatGPT filled the gaps with the average of everything it's ever read, which is exactly what a generic email sounds like.
The five-part prompt left almost nothing to chance. And when nothing is left to chance, there's nothing to fix.
What I Took Away From This
Three things, and I've applied all of them since.
Count Your Edits
If you're not counting, you're guessing. I now keep a rough tally in my head whenever I test a prompt. It takes no effort and it's the only way to know whether a change actually helped.
The Prompt Is the Work
The lazy prompt feels faster because you write less. But you're not saving work—you're moving it. Every second you don't spend writing the prompt, you spend editing the output, usually at a worse rate.
Two minutes of prompt writing to save ten minutes of editing is a trade I'll make every single day.
Specify the Reader, Not Just the Task
The single biggest improvement between prompt two and prompt three wasn't the format rules. It was telling ChatGPT who was reading and what they cared about.
"Short email" is vague. "A busy manager on her phone who cares about what changed, what's at risk, and what she needs to know" is a target. Given a target, ChatGPT hits it far more often.
The Test Card
Task: Turn rough project notes into a short update email for a busy manager.
Prompts tested: Three versions—lazy (one line), better (basic structure), and five-part (role, task, context, constraints, format).
Starting conditions: Nine lines of unstructured notes, roughly 90 words, no existing draft.
Time before ChatGPT: Writing the email from scratch: about 15 minutes including review.
Time after ChatGPT: Lazy prompt: 12 minutes editing. Better prompt: 5 minutes editing. Five-part prompt: 2 minutes prompting plus 90 seconds editing.
Editing required: 7 edits, 4 edits, and 1 edit respectively.
What went wrong: The lazy prompt produced filler phrases and hedged language. The better prompt got the structure right but buried the priority. Only the five-part prompt got the ordering and tone correct on the first pass.
Privacy notes: Names replaced with generic roles before pasting. No client names, account numbers, or specific project identifiers included. Checked against company policy before using an external tool.
Who should not use this method: Anyone writing something too short to justify two minutes of prompt writing. Anyone who won't count their edits—without the count, you're just trusting a feeling. And anyone whose notes contain details they can't describe generically.
The Bottom Line
Same task. Same notes. Same tool. Three prompts, and a difference of more than ten minutes of editing between the worst and the best.
The winner wasn't the cleverest prompt. It was the one that made the most decisions up front—who's reading, what matters, what to avoid, what shape to use.
I tried it at my desk so you don't have to. Useful beats impressive, and if it saves time, I'll show you how much: eleven minutes, on a single email.
