June 2026 · Case Study #5 · Emanuel Cira, AI Automation

June 2026 · Case Study #5 · Emanuel Cira, AI Automation

From One Inbox to Two Pipelines: Website + Order Automation for Engravit

0%

of inquiries logged automatically

0%

of inquiries logged automatically

0

back-and-forth needed just to understand what a customer wants

1

structured intake pipelines (orders vs. questions)

1

structured intake pipelines (orders vs. questions)

Image

The Problem

One inbox, every kind of message, zero structure.

The old GoDaddy site funneled everything: questions, orders, vague one-liners; into the same email.

01.

No separation

A price question and a real order looked identical. Both had to be read and sorted by hand.

02.

Incomplete orders

No material, no quantity, no timeline. Every order meant a reply just to find out what was actually being asked for.

03.

Nothing tracked

No log, no record. A buried email meant a buried order.

04.

No confirmation

Customers submitted and waited, no idea if it was even seen.

The result: every inquiry, big or small, cost the same thing… a read, a guess, and a reply.

How It Works

Two pipelines. One source of truth.

From the moment someone lands on the site to the moment their order's tracked, nothing touches an inbox until it's already structured.

Frontend

3-page site with a 25+ photo gallery and embedded order forms. Customers see the work before they fill anything out — fewer "what do you even make?" emails.

Automation

Every submission logs to Google Sheets and triggers two emails — an internal alert with the full order, and a confirmation to the customer. Nothing gets missed, nothing waits.

Backend

The form splits questions from orders, and orders come with the fields that matter: material, quantity, finish, timeline, file upload. JS validates it, n8n receives it.

Why This Works

The form does the sorting the inbox couldn't.

Splitting questions from orders fixes the real problem. Each one now routes to where it's useful, instead of competing for the same read.

Specific fields beat open text. Customers don't know what's relevant, the form does. That turns vague requests into job-ready specs.

Logging beats remembering. Every order is recorded the second it's submitted, not whenever someone gets to it.

Fewer steps, not more software. What took 3–4 emails per order now takes one form and, usually, zero replies beyond a confirmation.

Tech Stack

Logo

HTML/CSS/JS

site + dual-pipeline forms

Logo

n8n

routes, logs, automates

Logo

Gmail + Google Workspace

live order tracker, instant alerts + confirmations

What This Unlocks

Intake's automated. Next: the work itself.

Two follow-on builds came directly out of conversations with the owner. Both aimed at the two things that still eat his time once an order is logged: answering repeat questions, and pricing the job.

01

AI Replies

RAG agent for repeat questions

Most questions repeat. An agent trained on Engravit's services and pricing can answer them directly. Anything it can't, goes straight to the owner.

RAG agent for repeat questions

Most questions repeat. An agent trained on Engravit's services and pricing can answer them directly. Anything it can't, goes straight to the owner.

02

AI Quoting

Research + quote agent

Takes an order's specs, finds the materials, runs the pricing formula, drafts the quote. From request to ready-to-send quote, no manual lookup.

Image