GlobalAssist

A content generation and publishing pipeline for a network of country travel booking sites, built entirely inside the client's own Zapier organisation.

 

## The client

GlobalAssist runs a network of country-specific travel and excursion booking websites on WordPress, read by travellers looking for activities, guides and transport in that country. The network covers six countries (Madagascar, Tanzania, Kenya, the Philippines, Thailand and Indonesia) and aims for thirty. Every site follows the same content pattern: pages for a service, for a service in a specific location, and for offers, until now written by hand one country at a time.

The client's rule for the engagement is that the whole system lives inside their own Zapier organisation: no external servers, databases or hosted code, so that generation, data and publishing stay on one platform the client already runs. That rule is why this is not a Drupal build.



 

 





 

## At a glance

ClientGlobalAssist, a network of travel and excursion booking sites on WordPressPlatformZapier (Tables, Zaps, Interfaces), publishing to the client's WordPress sitesScopeSix countries so far, with thirty as the targetStatusIn progress. Data tables and prompt templates are in place for all six countries; automated generation and publishing to the live sites has not shipped yet.

 

 





 

## What we built

- Zapier Tables holding every combination of location and service for all six countries, about 800 rows, following the client's own field names so their existing workflow still reads them
- 24 reusable prompt templates rebuilt from the client's hand-written originals, for services, locations, posts, local services and offers, parameterised by country, brand and domain so that one template serves every site
- A shared content taxonomy: a tag for the type of service, such as guide, excursion, airport transfer or car rental, and a category for the location
- A designed generate-to-publish flow: build a prompt from a table row, generate the content through the client's Google Gemini connection, create a WordPress draft, then set its SEO title and description, its URL and its tag and category



 

 





 

## Generation and publishing

The generation half has been proven end to end: a local-service prompt run for a real location and service in the Philippines produced all 21 of its content fields as clean JSON. The publishing half waits on an authenticated connection to the client's WordPress sites, so no page has yet been generated and published automatically to a live site.



 

 





 

## Getting the data right

On tables this wide, loose instructions to Zapier's table tools produced values in the wrong columns. Every write now states exactly what each column must contain and clears everything else, and every row is read back after it is written to confirm it matches.

The shape of the data was checked against the client's live sites rather than assumed: the older Tanzania site already carries a complete grid of locations and services, which confirmed that the gaps on the newer Madagascar site are content still to write, not a wrong model.



 

 





## Start a project.

Tell us about your content workflow and the platforms it has to live inside.



[Get in touch →](/contact)