Blog (case study)Source
From 3.5 Hours a Day to 30 Minutes: AI Email Triage in Outlook
by Noah Hailegiorgis · 2025-08-07
✓ Opened with WebFetch on 2026-10-15
Read the original ↗Part 1 · What the source did
An agency owner's single inbox gets an AI layer that sorts every message, drafts or sends replies, schedules lead and invoice nudges, and logs everything to a sheet.
- Classifies mail as lead, quote, invoice/payment, support, spam or out-of-office.
- Sends only low-risk, high-confidence replies; everything else becomes a one-click draft.
- Schedules lead nudges at +2 and +5 business days and invoice reminders around the due date, stopping on reply.
- Sends only during business hours, with retries, failure alerts and a Google Sheets log.
Summary written by Agent Cookbook. See the original for the full context.
Part 2 · Our copy-paste version
Adapted by us from what the source described. Fill in the [BRACKETS].
The prompt
For each new email, return JSON only:
{"category": "lead|quote|invoice|support|spam|ooo", "confidence": 0-1, "customer": "", "service": "", "summary": "one line", "draft_reply": "in our tone, under 120 words", "follow_up_days": [2,5] or [], "send": false}
Rules: "send" may be true only if confidence >= 0.85 AND the email mentions no money, complaints or legal terms. Never invent prices or dates. Use the thread history if present.
🛡
Approval ruleStart with everything as drafts. Allow auto-send only for one narrow, low-risk category after two weeks of clean drafts.
Set it up
- Connect the inbox trigger in n8n (or your agent's email tool).
- Paste the classifier prompt into the model step.
- Route by category: log to a sheet, create drafts, schedule follow-ups.
- Run on a copy of last week's mail before going live.
Common failures
⚠
What usually goes wrong
- Follow-ups keep firing after the customer replied. Stop sequences on any reply.
- Marketing follow-ups need an opt-out line where the law requires it.
Variations
- Grok Bot version: run as a routine with Gmail and 'draft only'.
- Add a weekly sheet summary of response times.