26 Days, One Person: Taking a Trilingual Growth Engine from Zero to Live

Background: a one-person company, and a claim that had to be tested on ourselves first

Wooshai is a one-person company serving local businesses in Portugal and Southern Europe — real-estate agencies, immigration and law practices, schools, stays and tourism businesses, private clinics. What we sell is an AI growth engine: content that keeps publishing, enquiries that get caught, and a record behind every step.

But before pitching that to any business owner, there was a question we couldn't walk past. If we claim AI can take over most of the execution work of a content and acquisition team, the honest way to prove it isn't a polished slide deck — it's to run our own business through the engine first. One person, on the clock, everything on the record.

So we decided: before finding a client, be the client.

What we built: an eight-step pipeline

The core of those 26 days is a pipeline running from topic to distribution in eight steps. Machine rules guard every step; humans keep the judgment calls — AI executes, a person decides.

1. 1

   Topic selection

   Topics are chosen around the real operating pains of business owners in five Southern-European industries; every candidate has to pass three gates — will the reader stop, recognise themselves, and be able to act.

2. 2

   Master article

   AI drafts, machine rules inspect line by line — a title's promise must be delivered in the body; invented numbers and invented experiences are rejected. Failing drafts are rewritten with the reviewer's notes attached.

3. 3

   Images

   LinkedIn infographics are typeset programmatically rather than AI-painted — pixel-precise, no garbled text, drawn from a library of 232 layouts.

4. 4

   Audio

   Narration is voiced separately in all three languages, each with an engine suited to it; script and narration always come from the same source text.

5. 5

   Video

   A self-built ffmpeg pipeline assembles footage, narration, per-line subtitles and branded intro/outro — with zero third-party fees in the render-and-compose stage.

6. 6

   Human review

   Every piece of content for every platform is individually approved by a human before it can ship; edit the master article and all prior approvals expire automatically.

7. 7

   Trilingual site publish

   Chinese, English and Portuguese pages go live together. The English and Portuguese are not machine translations — each is rewritten for its own reader.

8. 8

   Multi-platform distribution

   One topic becomes LinkedIn, YouTube, Instagram and email content, each piece carrying its own source-tracking link.

Alongside the pipeline sits a set of pre-publish machine gates: whatever a title promises, the body must deliver; no invented numbers, no unsourced anecdotes, no quantified outcome promises. Those rules governed our own content first — and they will govern every piece produced for a client.

Where it stands: as of 26 July 2026 — every figure below is verifiable

Each item has a paper trail: feature releases map to merge records in the repository, articles and videos to publicly accessible pages and the YouTube channel, WhatsApp clicks to an attribution log with a timestamp and source on every entry.

And in the same spirit of honesty: a click is a click, not a sale. What 85 WhatsApp clicks prove is the chain itself — content got seen, landing pages got opened, the attribution loop works. They are not a revenue claim, and dressing clicks up as clients is exactly what our own rules forbid.

What this means for your business

This case answers a plain question: how far can one person with no team get on this engine? The answer — in 26 days, trilingual content, landing pages, enquiry capture and attribution tracking, all working end to end. And your business holds something we didn't have: real clients, real word of mouth, real local stories. The same engine attached to a business with stories to tell starts from higher ground.

How the engine plugs into each of the five industries is written up as five method playbooks — illustrative scenarios, clearly labelled as such, never client engagements. Real client cases will be written the way this one was: after they actually happen.