Triage your inbox without automatically replying
Separate messages needing a decision from reference material and low-priority mail.
What you’ll produce
A source-linked inbox review with action items, deadlines, and unsent reply drafts.
Inputs & connections
- A limited date range or mailbox label
- Your priority rules and examples of messages that matter
Start with approved files or pasted excerpts in ChatGPT. If using connected apps, confirm access and permissions first. For recurring monitoring, use a configured task or dot; a single chat prompt does not create a running automation.
Your starting prompt
Replace the bracketed fields. This is our suggested prompt, not a quote from the source.
Help me with this small-business task using only the sources I provide or explicitly connect. Review [mailbox scope] using [priority rules]. Return messages requiring my decision, deadlines explicitly stated in the email, unanswered questions, and optional unsent reply drafts. Link to the original message and distinguish an explicit commitment from an inference. Do not send, archive, delete, unsubscribe, or change labels. Ask for clarification when the sender's intent is unclear.
How to run it
- 01
Connect only the mailbox access you intend ChatGPT to use, or provide approved exports.
- 02
Define the time range and examples of high-priority messages.
- 03
Compare the first brief to the original messages, including excluded items.
- 04
Adjust rules and use an approval step for any later actions.
Review before you act
Check deadlines, recipients, commitments, and any message the assistant may have missed.
Limitations & success checks
- Mail access and available actions depend on your configured account.
- A large inbox demonstration does not prove reliable triage for every mailbox.
- Adapted from a Muse workflow; this guide has not been reproduced with ChatGPT. Availability and integrations depend on your account.
Measure: Missed important messages, false priorities, and corrections to draft replies.
Sources & attribution
The sources below support the reported example or a tool capability. They do not validate our prompt or prove the outcome in your environment.
September 8, 2026. Official demonstration of working with a large email history. Our priority rubric and draft-only prompt are editorial additions. This guide adapts the idea for a small-business owner; any reported outcome belongs to the original tool and source.
Creator / publisher: Muse, featuring Casey and Hal
Official introduction of the personal agent. Connector access, account availability, and action permissions still need to be configured.
Creator / publisher: Meta
Official examples span ChatGPT, Work and Codex. This small-business prompt is our adaptation, not a demonstrated customer result.
Creator / publisher: OpenAI
Editorial review: 2026-10-17. How we label evidence.