You want a dripping tap repaired. Your notes say where it is, when you noticed it, and how to contact you. An AI draft turns that into a confident explanation of a broken valve and a promise that someone can enter any time. The message sounds helpful, but it now says things you did not establish or authorize.
The writer's problem is turning observations into an actionable repair request while keeping diagnosis, urgency, and access arrangements separate. This example complements our protected-terms guide: here, the details to protect include what happened and what you have actually agreed to. The scenario is fictional and concerns writing, not plumbing instructions or tenancy advice.
Collect the facts the recipient can act on
Before requesting a polished email, write a short record in your own words. Include the location, the observable problem, when you noticed it, and the effect you can describe. If you have already reported it, include the date or reference you actually have. Leave out guesses about causes.
- Kitchen cold tap drips into the sink after the handle is closed.
- First noticed on November 10; still happening on November 12.
- No repair attempted. Cause unknown.
- A photo is available, but has not been attached yet.
- Request: arrange an inspection or repair visit.
- Availability: Tuesday afternoon may work, subject to confirmation.
- Reply through the existing maintenance portal to arrange a time.
These notes do not establish a leak beneath the sink, a flood, a faulty washer, or permission to enter while the resident is absent. The writer should not add any of those details merely because they would make the message more complete.
Ask for a draft and a separate question list
Try a prompt that tells the model where its authority ends:
“Turn these notes into a concise maintenance request with a descriptive subject line. Describe only the observations supplied. Do not diagnose the cause, invent previous reports, add a deadline, or grant access permission. Keep tentative availability tentative. Put missing information in a separate list of questions for me, outside the message.”
Use placeholders for an address, unit number, or contact details while working in a general writing tool. Add necessary identifying information yourself in the appropriate reporting channel. There is no benefit in giving an editing model private details it does not need to improve sentence structure.
Compare what each version commits you to
Overstated draft: “Urgent: faulty kitchen valve. The broken valve has been wasting water for days. Please replace it immediately. I have attached photos, and your contractor can enter on Tuesday afternoon even if I am out.”
The draft invents a diagnosis, chooses a repair, claims an attachment exists, upgrades tentative availability, and grants access. Editing only the opening adjective would leave most of the problem untouched. Each sentence needs to be compared with the notes.
Source-preserving draft:
Subject: Kitchen cold tap continues dripping after closing
Hello, the kitchen cold tap continues to drip into the sink after I close the handle. I first noticed this on November 10, and it is still happening today, November 12. I have not attempted a repair and do not know the cause.
Could you arrange an inspection or repair visit? Tuesday afternoon may work for me, but please confirm a time through this portal before scheduling. I can provide a photo if helpful. Thank you.
This version tells the recipient what to investigate and how to coordinate. It does not pretend the writer has diagnosed the fault. If the portal already has separate fields for the location or appointment preferences, use them consistently instead of adding conflicting information in the message.
Keep urgency factual
A message can be clear without automatically using “urgent,” “dangerous,” or “uninhabitable.” Equally, do not let a request to sound polite erase a serious observation you actually supplied. The task is to preserve the situation accurately, not to make every report sound mild.
If the real situation requires an urgent reporting route, use the relevant established procedure rather than waiting for an AI draft. This example does not determine which problems are emergencies. For an ordinary written report, describe the observed impact and avoid adding an unsupported deadline or legal claim.
Check attachments and appointment language separately
Before sending, verify that every promised attachment is present and corresponds to the issue. “I can provide a photo” and “I have attached a photo” describe different actions. A generated hero illustration, like the one on this page, cannot serve as evidence of a real repair problem.
Read the appointment sentence as though you were the person booking the visit. Does “Tuesday afternoon may work” still invite confirmation, or has it become “Tuesday afternoon is confirmed”? Does the message accidentally authorize entry? Keep availability, booking, and access permission as separate decisions.
Make follow-ups add information
If you follow up, retain the original report date and add the current status you can verify. For example: “Following up on my November 12 report: the tap is still dripping as described. Could you confirm the next step?” Do not let a model invent previous unanswered calls or a worsening condition to make the message more persuasive.
Our guide to preserving uncertainty in AI drafts is useful when observations remain incomplete. You can also use AI Undetectable to explore clearer wording, then compare the result with your notes. A detector score cannot verify the facts or permissions in a message.
The final check is practical: can the reader identify the problem, understand what you are requesting, and arrange the next step without relying on invented information? Once those points are clear, the message is ready for your review and submission.