This website — a sales system that runs itself

Context

AdSys Team is a small shop: there is no one staffing live chat 24/7, and answering the same pricing questions over and over takes time away from the work itself.

Most of what the site sells is digital: n8n workflows packaged in the Workflow Store — orders can arrive at any hour, including midnight, and buyers expect the file almost immediately after paying.

The challenge was to build the exact sales and lead-qualification machine AdSys Team sells to other businesses, and run it for real on its own website — not a demo built just to show clients.

Architecture & stack

The whole site runs on Next.js, with the automation layer behind it running on n8n — the same technology AdSys Team packages and sells back to clients through the Workflow Store.

The sales flow: a signed-in customer clicks buy → the system creates an order with a unique code → a payOS (VietQR) code appears right in the customer portal with the exact amount; the incoming payOS webhook is matched against the order code and amount, duplicate processing is blocked, and the configuration file plus setup instructions open in the customer portal once ready — until then the customer tracks preparation status there (email only confirms that payment arrived).

The lead-qualification flow is a template workflow sold in the store: a six-question quiz scores budget, urgency and company size, hot leads alert the team on Zalo within five minutes, and cooler leads enter a seven-email nurture sequence spaced over 14 days that stops the moment the lead books a call; on this website itself, consulting enquiries are currently handled directly over Zalo rather than through the automated quiz.

A separate error-watch workflow monitors all of the above: if any workflow breaks, it reports straight to Zalo with the failing node's name, so nothing fails silently for days before anyone notices.

Delivery milestones

The workflows were not switched on all at once; they moved to running fully automatically in stages — the Behind the scenes page reflects only what is genuinely running, rather than diagramming something that does not exist yet.

Building the site itself followed infrastructure first, content after: the first version went live on real infrastructure on 28/08/2026, the domain moved to stable HTTPS from 29/08/2026 so payment webhooks and lead forms could work correctly, and only then came the stages that expanded products and content.

A separate customer portal — sign in to look up services, invoices and support tickets — was added later, with its layout fully separated from the sales pages so marketing content never bleeds into a paying customer's workspace.

Day-to-day operation

For digital-product orders: customers sign in and pay through payOS, and the order is matched automatically with no one on duty; files and instructions open in the customer portal once ready, and until then customers track preparation status there — a person only steps in when the transferred amount does not match the order or the customer needs more advice.

For consulting leads: on this website itself, intake currently runs directly over Zalo — the scoring, fast Zalo alerts for hot leads and the automated nurture sequence are the template workflow sold in the store for installing on your own form.

At 7am every Monday, a report combining views, orders, leads and revenue from the past week is sent automatically to Zalo — someone reads the report instead of having to compile it.

Lessons

The clearest lesson: selling clients the exact thing you run yourself forces every marketing claim to stay honest — if a workflow is not genuinely live, the Behind the scenes page has to say it is 'built per system, done as part of a service package' rather than diagram something polished that does not exist.

Clearly separating what a machine should handle from what a person should decide — a mismatched amount, a request for advice — keeps automation from becoming a black box: a person still has a place to step in at the right moment.

A dedicated error-watch workflow for the other workflows turned out to matter as much as the core business workflows themselves: automation that cannot report its own failures lets problems sit unnoticed for days.