Skip to content
TriedPrompts

More in Business admin & workflows.

Automate a small shop's support inbox with ChatGPT Plus after Claude fixes the design

WorkedEvidence: Article or interviewPosted Last checked

After failed ChatGPT designs, Claude reworked a shop's support inbox flow; of 1,689 backlog emails, only 11 needed a personal reply.

How to use this

  1. Tap Copy prompt.
  2. Open ChatGPT.
  3. Paste it into the message box.
  4. Send it, then read the answer before acting on it.
New to this? What's a prompt?

A prompt is the message you send to the AI tool. Copy the one on this page, put your own details where it asks, and read the answer before you act on it.

Prompt

Recreated prompt

The source didn't share its exact prompt, so we wrote one from what it describes. The result shown is the source's, not ours.

What the prompt labels mean

The person didn't share their exact words, so we wrote this from what they described. We haven't tested it, and they didn't use this exact version.

Here is a full export of our support inbox. Learn how messages have been filed and answered, write a rulebook for classifying new mail, and design a simple daily workflow: ChatGPT files messages and drafts replies from templates, and a scheduled Apps Script sends only qualifying drafts and confirms they went out. Anything unclear goes to a Human Attention Required folder.

Try this in:

Sends the Recreated prompt shown above. ChatGPT and Grok open with it filled in, and ChatGPT may send it straight away. Gemini and Claude open empty, so we copy the prompt for you to paste. Every button also copies it.

Opens in a new tab. We don't see what you send.

These links are only a shortcut. We're not affiliated with OpenAI, Google, Anthropic or xAI, and your results may not match the source.

Optional stronger version

Our rewrite with extra checks. The poster didn't test it.

Before you start, ask me 1–3 short questions about the details below and wait for my answers. Once I've answered, confirm the plan in one line and run the whole task start to finish.

Goal: Cut a small business support inbox down to the few messages a person really needs to answer.

Questions to ask me first:
- Which mail types arrive (customers, orders, payments, vendors, spam)?
- Which replies are safe to send automatically?
- Who should get urgent or payment-problem alerts?

Inputs (fill in or confirm):
- Inbox export or sample: [FILE]
- Existing labels and templates: [LIST]
- Accounts and plans available: [CHATGPT / APPS SCRIPT / OTHER]

Rules:
- Derive rules from real history before designing anything.
- Keep the design as simple as the business needs.
- Draft first; only send drafts that meet clear, testable criteria.
- Leave anything unconfirmed in the inbox for a human.

Checkpoint: Run one real day of mail through the rulebook in draft-only mode and show the results.

If you get stuck: If a step can't run unattended, say so instead of simulating success.

Verify before you finish: Confirm sent replies appear in Sent and originals are labeled; spot-check ten filings.

Output: A rulebook, the scheduled task setup, the send script and a weekly error report.

Done when: Only messages that need a human are left, and every auto-reply is confirmed sent.

AI tools

ChatGPTClaude

Outcome

David Gewirtz set out to automate customer support email for a spouse's small e-commerce shop selling digital hobby goods, where only about 1–2% of incoming mail came from real customers. For almost four weeks ChatGPT designed a multi-tier system around a shared spreadsheet that grew to 73 tabs and had Gewirtz copy instructions between conversations. A second try turned it into a skill with a 12-stage pipeline that passed 59 unit tests and 139 regression tests but made no real Gmail changes when scheduled. Gewirtz then had Claude, on a $100/month Claude Max plan, review a dossier of ChatGPT's work. Claude read a Google Takeout export of 88,777 messages in under an hour, wrote a Mail Bot Rulebook, and split the job: a daily scheduled ChatGPT task on the $20/month Plus plan files mail and drafts template replies, and an hourly Google Apps Script sends qualifying drafts (at least 10 minutes old, up to 10 per run) and confirms each one. After 15 small rulebook revisions it runs unattended. In one 1,689-message backlog, only 25 were from real customers, leaving 11 that needed a personal reply.

Evidence: Article or interviewA newsletter, journalist or interview account of what happened.

Source

Posted
by David Gewirtz (ZDNET)
Last checked
Method:
How we verify

https://www.zdnet.com/innovation/using-chatgpt-plus-for-small-business-customer-service

Context: Written by a ZDNET columnist about a family business; also syndicated on Yahoo Tech. Earlier ChatGPT-only attempts failed; the working setup still routes unclear mail to a human.

More in Business admin & workflows