This is the pipeline when the source is the inbox. A small web-design agency for local Australian businesses ran everything out of a single Outlook inbox, and the owner, Michael, was losing hours a day to it. I built an AI triage layer that reads every incoming message, pulls out what matters, classifies it, and either drafts the reply or schedules the follow-up, logging it all to a sheet. Same core job as the rest: extract structured data from one channel and route it where the work actually happens. It cut his inbox time from about 3.5 hours a day to roughly 30 minutes.
Results at a glance
- Founder time cut from ~3.5 hours/day to ~0.5 hours/day
- 20–30 emails/day auto-handled, plus scheduled follow-ups
- Lead replies in seconds; invoice nudges sent on time, every time
- Prototype in 1 day, live in 5
The problem
Michael ran the business from one Outlook inbox, no automation, no CRM. He needed to reliably categorize incoming mail (leads, invoices, support, spam), send instant personalized replies where it was safe and draft the rest for quick approval, keep lead and invoice follow-ups on schedule without tracking them by hand, stay compliant with Australia's Spam Act, and keep it all inside Outlook.
What I built
An n8n workflow that watches the Outlook inbox, classifies each message, drafts or sends a response, and schedules follow-ups, logging every action to a Google Sheet for review.
- Classifies each message: lead, quote, invoice/payment, support, spam, or out-of-office
- Replies: instant send for low-risk, high-confidence cases; one-click-approve drafts for anything nuanced
- Lead follow-ups: instant first reply, then nudges at +2 and +5 business days, auto-stopping on a reply or booked call
- Invoice follow-ups: pre-due, due-day, +3 and +7 business days, pausing after two unanswered
- Personalization: business name, suburb, service requested, and prior thread context, in local tone and spelling
- Operations: sends only during AEST/AEDT business hours; idempotent processing, retries with backoff, and failure alerts
The results
Founder time dropped from roughly 3.5 to 0.5 hours a day. Lead response fell to seconds in straightforward cases, and drafts cut the rest to minutes. Invoices got consistent, on-brand reminders with no manual tracking.
| Before (manual inbox) | After (AI-assisted flow) |
|---|---|
| Missed and late follow-ups | Scheduled nudges for leads and invoices |
| Slow, inconsistent replies | Instant replies or ready-to-send drafts |
| No system of record | Google Sheets log with status and next step |
| High founder time cost | 3+ hours/day saved |
"My inbox used to run my day. Now it's the other way around. Replies go out fast, follow-ups never slip, and I only review the tricky ones. I got back about three hours every day without losing the personal touch."
Compliance: built to Australia's Spam Act 2003, sender identification and an opt-out line on non-transactional follow-ups, no cold sending, and DKIM/SPF/DMARC validated through the Microsoft 365 tenant.
Built with: n8n, Microsoft 365 Outlook (Graph API), and Google Sheets.
Want your inbox to run itself? Tell me your stack and goals and I'll send back a build plan and timeline.