01
SOP & training-doc generator
Recorded calls, chat threads and scattered notes become written procedures your team can actually follow. New staff stop asking the same five questions, and the knowledge stops living in one person's head.
Retention automation for membership and community businesses. I work in n8n, self-hosted on infrastructure I run myself, and I hand over things that keep working after I log off.
What I build
Each of these is scoped to ship in four to eight weeks, with a defined handover. I would rather finish one than start three.
Recorded calls, chat threads and scattered notes become written procedures your team can actually follow. New staff stop asking the same five questions, and the knowledge stops living in one person's head.
Every survey, review, support ticket and comment thread read and sorted into themes, ranked by how often each one shows up just before somebody cancels. You stop guessing which complaint is the expensive one.
The list you stopped emailing, segmented, written to, and routed so replies reach a human instead of an unread inbox. The cheapest revenue in the business is usually sitting in a list nobody has touched since last year.
Built and running
I would rather show you one thing running than list ten I have read about. This is live, and I can walk you through it on a call.
An agent that answers questions on a Facebook Page in Taglish, grounded in a knowledge document so it declines to invent prices or dates, and logs every conversation to a sheet for follow-up. Built node by node: the Meta webhook verification handshake, a filter that drops everything but real text, a Graph API call to identify the sender, and a reply posted back through the Send API.
It runs on infrastructure I set up and administer myself — a VPS running n8n under Docker Compose, with Traefik handling routing and automatic TLS certificates on my own domain. That matters for a practical reason: client demos run on public webhooks that do not vanish when a trial expires.
More goes here as I ship it. I add a build to this page when it is running, not when it is planned.
How I work
Before anything is built, we trace the actual path a member takes and find the point where they go quiet. Often it is not where you expect, and often the fix is smaller than the one you were about to pay for.
One workflow, scoped to one leak, shipped in weeks rather than quarters. You see it running on real data before we discuss anything else.
Every build comes with a written runbook: what each step does, what breaks it, and how to fix it without me. Seven years teaching IT means I write these for the person who has to read them at 2am, not for the person who built it.
Tools
Listed because I use them, not because they are trending.
Working together
I take on one engagement at a time. Limited capacity is a deliberate choice: it keeps each build properly scoped, tested on real data, and documented well enough to run without me. If my schedule cannot meet your timeline, I will tell you that at the outset rather than overcommit.