Skip to content
← All cases
RUCV

CASE 06

DK Nades

My own project, where I went from leading a team of contractors to a system in which automation does the routine work and a person only makes the call.

DK Nades — dknades.com
~75 hof manual work replaced by the system in 12 days

Problem

Production did not scale: every piece of content was recorded by hand — set up the scene, record, edit, file it — and then the contractor left. Publishing was manual, decisions about what to release got lost in chat, and every change went straight to production. All of it resting on one person with no budget for a team.

What I owned

Founder and sole operator. 2024–2025: built and led a team of five — strategy, SEO, design decisions, product logic, briefs, advertiser relations. 2026: took the product on myself — designed the production and publishing system, briefed every part of it and signed each one off, with AI tools used in the development.

System

  • 01A platform in 8 languages: built locally, released to production by protocol
  • 0289 automated production checks, off-site backups with a verified restore
  • 03Unattended production: recording, editing in two formats, automatic quality control
  • 04Finished work lands in a messenger — a person approves it in one tap
  • 05What is approved becomes a page and publishes itself to four channels
  • 06Human corrections are logged, and the system learns from that log
  • 07Work is split by role across AI workers; anything irreversible needs sign-off
  • 08Overnight shifts run for around 10 hours with nobody watching

WHAT THIS MEANS FOR YOUR COMPANY

Automating manual workI find the most repetitive work and turn it into a system, keeping people only where a decision is needed. Proven twice: the invoice workflow at work and video production here.
Rolling out AI with controlAI as a managed operating model rather than a chatbot: roles, rules, quality checks, a budget per model, and sign-off for anything irreversible.
Approval and acceptanceFlows where the system prepares and a person signs off in one tap, with defects caught before anyone has to look. The same shape works for invoices, contracts and contractor work.
Honest metricsSeparated real visitors from bots — the real figures were half of what was reported — and made the weekly report arrive on its own.
Continuity and securityOff-site backups with verified restores, a review after every incident, key-based access. The same work as the Microsoft 365 migration, continued.

Result

Producing one lesson used to take about 30 minutes of manual work; now a person taps once. In its first 12 days the pipeline delivered around 150 finished videos in two formats — roughly 75 hours of work that nobody has to do any more. The catalogue went from 176 to 218 lessons across 8 languages in 9 days, publishing to four channels runs on a schedule, and pages load two to four times faster. The system is running and still being built out. Stated plainly: the project's audience is small today — this is a case about building a system, not about growing reach.

Stack

dknades.com8 languages · 2,234 pagesUnattended video productionPublishing to 4 channelsRole-based AI agents

Next case

Odoo ERP