Skip to content

Email & Inbox

AI Auto-Reply Drafter

Drafts a human-sounding reply for every email — you hit send after a quick glance.

Saves about ~3 hrs per week

What it does

  • Reads each incoming email and drafts a reply in your voice
  • Fills in answers from your calendar, pricing, and FAQ notes
  • Leaves the drafts in your inbox for a one-click approve or edit
  • Never sends without your say-so — it's a drafter, not an autopilot

How it works

  1. New email triggers the script
  2. AI drafts a reply using your context notes
  3. Draft lands as a reply in your drafts folder
  4. You review, edit if needed, and hit send

What's included

  • Draft prompt setup
  • Context note template
  • Gmail drafts routing

Set it up yourself

Follow these 4 steps in order. You will need accounts for: Gmail.

Step 1: Write your context notes

  1. Create a Google Doc named "Reply Context" and keep it open while you build.
  2. Write a short section on your pricing: what you charge, what's negotiable, and where to point people for quotes.
  3. Add a section on availability: your working hours, how bookings happen, and the link people should use.
  4. Add a section of FAQ answers you give on repeat (timelines, revisions, scope).
  5. Keep it under ~1,500 words — the AI reads this every time, and huge docs slow it down and dilute focus.

Watch out: Anything you do not want AI to answer (legal, contracts, sensitive pricing) should be explicitly labeled "Do not answer — escalate to me".

Step 2: Set up the Gmail trigger

  1. Create the OAuth credentials in Google Cloud as in the Gmail Inbox Triage guide (APIs & Services → Library → Gmail API → enable; then Credentials → OAuth client).
  2. Open https://script.google.com and create a new project named auto-reply-drafter.
  3. Paste the drafter script from the guide into Code.gs.
  4. In Project Settings → Script Properties, add the DOC_ID property equal to the ID of your Reply Context doc (the long id in the doc's URL between /d/ and /edit).
  5. Save and run once to authorize.

Watch out: Without the DOC_ID property the script cannot read your context and drafts will be generic.

Step 3: Add the drafting prompt

  1. In the Apps Script editor, find the REPLY_PROMPT constant at the top of the script.
  2. Replace the placeholder with the prompt template from the guide.
  3. The prompt should instruct the AI to: acknowledge the specific concern, answer in 2 sentences from the context doc, include one clear next step, stay under 100 words, professional but not formal.
  4. If you want a signature, append "Sign off as <your name> using the signature in the context doc".
  5. Save.

Watch out: Keep the prompt outcome-focused ("draft a reply that…") rather than personality-focused ("be witty") — the latter produces unusable drafts.

Step 4: Route drafts back to Gmail

  1. In the script, confirm the draft function uses GmailApp.createDraft(thread) so the reply lands inside the existing thread.
  2. Add the trigger for the function that processes new mail — time-based trigger running every 10 minutes.
  3. Authorize Gmail scopes when prompted.
  4. Send yourself a test email, wait for the next trigger cycle, and open Gmail → Drafts: you should see the AI reply attached to the thread.
  5. If drafts are missing, open View → Logs in the editor and check for authorization or quota errors.

Watch out: Because this only creates drafts, it is safe to leave running — nothing is ever sent without you.

Frequently asked questions

What does the AI Auto-Reply Drafter automation do?

Drafts a human-sounding reply for every email — you hit send after a quick glance. Reads each incoming email and drafts a reply in your voice

How much time does AI Auto-Reply Drafter save?

It is estimated to save ~3 hrs per week. Actual savings depend on your volume.

Which tools does AI Auto-Reply Drafter use?

It works with Gmail.

What does AI Auto-Reply Drafter include?

Draft prompt setup, Context note template, Gmail drafts routing.

Can I set up AI Auto-Reply Drafter myself?

Yes. The do-it-yourself guide on this page walks through 4 steps, from getting access to testing the finished automation.