The Use Case Shed Beginner guide
Draft a friendly email from a few facts.
Give a text-chat tool a few facts and ask it to draft a friendly email.
Workflow tested in Codex
A short, friendly email draft with a subject line.
01 / Get ready
Before you start
- A text-chat tool you already have. Availability and limits depend on your account; sign in if your tool requires it.
- The recipient, purpose, facts to include, and anything you cannot promise. Start with the invented picnic reply below.
- A moment to review the wording before using the draft. No email connection is needed.
Safe sample: replying to a picnic invitation
Your neighbor Jo invited you to help set up a community picnic. You can help from 10 to 11 a.m. on Saturday and bring two folding chairs. You cannot offer transport. You need to ask where to meet.
This situation is invented. Use placeholders for private names or details when trying your own draft.
02 / Give it a go
Try the workflow
Open a new text chat.
Keep your facts and limits in front of you. This workflow only needs text pasted into the chat.
Send the complete prompt below.
It gives the recipient, tone, available time, offer, limit, and question, then asks for draft text only.
Read the draft as the recipient would.
Check what it promises. The sample should offer one hour and two chairs, decline transport, and ask where to meet.
Adapt the facts and review again.
For your own message, replace the recipient and facts in the prompt. Edit the draft until it says what you mean. You decide whether to copy it into your email app and send it.
The complete prompt
Original prompt by Use Case Shed. Copy the complete text to try the recorded example.
Draft a short, friendly email to a neighbor, Jo, replying to an invitation to help set up a community picnic. Use only these facts: I can help from 10 to 11 a.m. on Saturday. I can bring two folding chairs. I cannot offer transport. Ask where to meet. Do not promise any other help. Include a subject line. Return draft text only; do not send anything.
03 / Make a correction
If the result goes wrong
It added a promise or changed your availability.
Point out the unsupported commitment and ask for a revision using only your facts. Check the next draft again before copying it elsewhere.
A follow-up prompt to try
Suggested correction prompt. This follow-up was not part of the recorded trial.
Revise the draft using only the facts in my original request. Keep the exact time, day, number of chairs, and transport limit. Remove extra promises and invented details. Keep the question about where to meet, a subject line, and a friendly tone. Return draft text only; do not send anything.
04 / Behind this guide
Sources and evidence
The explanations, sample, and prompts are original work by Use Case Shed. The references below support the method.
- OpenAI: Prompting
Supporting reference for giving context, tone, format, constraints, and revision instructions.
- OpenAI: Use ChatGPT
Supporting reference for drafting and checking important details against the original material.
- OpenAI: Quickstart
Supporting reference for account access and beginning a chat.
What we tested
One workflow run in Codex used the exact prompt shown above. Its draft preserved the time, day, two-chair offer, transport limit, and meeting question. It did not invent an additional commitment or sender name. No email was sent. The exact model identifier was not exposed.
The ChatGPT interface and beginner usability have not been tested. A successful sample does not guarantee that another request or tool will preserve your facts and limits.
Workflow evidence and supporting sources checked .