I still remember the exact moment I gave up on writing the email myself.
It was a gray Tuesday afternoon in Peoria. A vendor had missed a delivery window for the third time in six weeks. The delay was starting to affect two downstream customers, and my manager had asked me to “handle it firmly but keep the relationship intact.” In other words: don’t start a fight, but don’t let them walk all over us either.
I opened Outlook, stared at the blank message window, and typed three different versions. The first one sounded like I was apologizing for their mistake. The second one sounded like I was ready to cancel the contract. The third one was so stiff it could have been written by a robot wearing a tie. None of them felt right.
I had heard people talking about ChatGPT. I had even opened the site once and closed it again because I didn’t know what to ask. That afternoon I decided to try something small. I copied the facts into a private note, removed every company name, account number, and identifying detail, and typed a plain request:
“Rewrite this email so it is clear, professional, and firm without being rude. Keep it under 150 words.”
What came back was not perfect. But it was better than anything I had written in the previous forty minutes. I edited three sentences, added one specific date, and hit send. The vendor replied the next morning with a revised schedule and an apology that actually addressed the problem. That was the first work email I let ChatGPT rewrite. It was also the moment I realized the tool could be useful if I treated it like a drafting assistant instead of a magic solution.
I’m Alex Rivera. I work in operations and project support at a regional distribution company. I’m not an AI expert and I don’t claim every email should go through ChatGPT. I only write about methods I have tested at my own desk, with real constraints, and with the same privacy rules most office workers should follow. This post is the full story of that first email—what I tried, what worked, what needed fixing, and how much time it actually saved.
The Situation That Forced the Experiment
The facts were simple but the tone was not.
A supplier had missed an agreed delivery window. The delay was creating scheduling problems for two of our customers. My manager wanted a written record that we had raised the issue clearly. At the same time, this supplier still handled other reliable work for us. Burning the relationship over one recurring problem would create more headaches than it solved.
I had already spoken with the account manager by phone. The email needed to do three things:
Confirm the missed window in writing
State the impact without exaggeration
Request a concrete recovery plan with dates
I also needed the message to sound like it came from a human who expected a professional response, not a form letter or a threat.
Why the Manual Drafts Failed
My first draft spent too many words explaining how busy we were. It diluted the point. The second draft used phrases like “unacceptable pattern” and “serious concern.” Accurate, but it landed like a formal complaint. The third draft tried to split the difference and ended up sounding like corporate filler.
I was stuck in the classic office trap: trying to be clear and polite at the same time, and ending up with neither.
The Exact Prompt I Used
I cleaned the facts first. No company names. No account numbers. No internal project codes. Just the sequence of events and the request.
Then I wrote this:
“I need to send a professional email to a supplier who has missed a delivery window for the third time in six weeks. The delay is affecting two customers. I already spoke with them by phone. The email should confirm the missed window, state the impact clearly, and ask for a specific recovery plan with dates. Keep the tone firm but respectful. Do not sound angry or overly formal. Keep it under 150 words. Here are the facts: [cleaned facts].”
That was the entire prompt. No special techniques. No long list of role-play instructions. Just a clear task, the constraints, and the raw material.
What Came Back
The first output was clean and direct. It opened with the confirmation of the missed window, stated the customer impact in one sentence, and closed with a request for a revised schedule by a specific date. The language was professional without the stiffness I had been producing.
It was not perfect. One sentence sounded slightly more formal than I like. Another sentence repeated a point I had already covered in the phone call. But the structure was usable. I spent less than five minutes editing.
The Test Card for This Email
Here is the full breakdown I use for every practical test.
Task
Rewrite a firm but professional email to a supplier about repeated late deliveries.
Prompt or workflow
Cleaned facts + clear request for tone, length, and required content.
Starting conditions
Three failed manual drafts. Forty minutes already spent. Privacy filter applied (no names, no account numbers, no internal codes).
Time before ChatGPT
Approximately 40 minutes of drafting and rewriting with no usable version.
Time after ChatGPT
4 minutes to generate + 5 minutes to edit and send. Total: 9 minutes.
Editing required
Minor. Adjusted one formal phrase, removed one redundant sentence, added the exact date we needed the recovery plan.
What went wrong
The first version repeated information from the phone call and used one phrase that felt slightly stiff for our normal working relationship. Nothing factual was invented.
Privacy notes
All company names, customer identifiers, and internal reference numbers were removed before pasting. The final email was reviewed by me before sending.
Who should not use this method
Anyone dealing with legal notices, formal contract disputes, or situations that require exact legal language. Also anyone who cannot take five minutes to edit and verify the output.
What I Changed Before Hitting Send
I keep a short personal checklist for any ChatGPT-assisted email:
Does the opening state the point in the first two sentences?
Is the tone consistent with how I normally write to this person?
Did I remove or rewrite any phrase that sounds like generic AI language?
Are the dates, numbers, and requests exact?
Would I be comfortable if this email were forwarded to my manager?
In this case I softened one phrase, cut a repeated point, and made sure the requested response date was unambiguous. The final version still carried the firmness the situation required, but it sounded like it came from me.
Why This Small Win Mattered
The supplier’s reply arrived the next morning. They acknowledged the pattern, offered a revised delivery schedule, and included a short explanation of the internal issue that had caused the delays. The conversation stayed professional. No relationship damage. No escalation.
That outcome was not magic. It was the result of a clear request, a usable first draft, and human editing. The time savings were real. More important, the mental load dropped. I no longer had to fight the blank page and the tone problem at the same time.
I started keeping a simple note: “Email rewrite – firm but respectful – works when facts are cleaned and length is limited.” That note later became the seed for several other desk workflows.
Lessons I Still Use
A few practical rules came out of this first test and have held up:
Clean the input before you ask for a rewrite. Privacy is not optional.
Tell the tool the tone you want and the length you need. Vague requests produce vague results.
Always edit. Even a good draft usually needs one or two human adjustments.
Measure the real time. If the process does not save time after editing, it is not worth repeating.
Keep the final voice yours. Readers can usually tell when an email was written entirely by a tool.
I also learned that “make this better” is almost never a useful instruction. Specific constraints produce better first drafts.
How This Fits the Rest of the Desk Fixes Series
This email was the first practical test that survived a real workday. Later posts will cover meeting notes, report summaries, and longer threads. The pattern stays the same: show the prompt, show the conditions, show the time numbers, and show the limits.
I still write many emails without any AI help. Some messages need a personal touch that no draft can supply. Others are so short that typing them directly is faster. ChatGPT earns a place only when it reduces the time or the friction without creating new problems.
If you have a difficult email sitting in your drafts folder right now, try the same approach I used. Clean the facts. State the tone and length. Generate. Edit. Send. Then decide for yourself whether the nine minutes I spent would have been worth it at your desk.
I tried it at my desk so you don’t have to. No magic. Just a workflow that survived Monday. Useful beats impressive. And if it saves time, I’ll show you how much.
