AI · Content System2026

Building an AI Newsroom Assistant Team

Designed and built a dedicated AI content assistant team for GS Caltex Media Hub, a brand journalism platform. Brand tone and voice, content series specifications, a multichannel distribution guide, the operating workflow, and KPIs are unified into a single operating system that supports the full cycle from planning to performance analysis. Built on a Human-in-the-loop principle: AI handles drafts and analysis, people decide and publish.

View Project ↗

Challenge

Media Hub runs many series — press releases, magazines, interviews, long-form features, employee stories, trend columns — across scattered channels, making it hard to keep brand voice consistent while moving fast. Content was produced per channel in isolation, and performance only surfaced afterward as fragmented data. Above all, there was no system to regularly diagnose the existing newsroom's strengths and weaknesses and turn that into improvements.

Approach

Designed an AI content team operating system — unifying the 4 brand voice principles and prohibited expressions, per-series specs (press release / magazine / short interview / unboxing / FACE+TIME / INSIGHT+), and the multichannel distribution guide into a single prompt framework

Hub-Spoke architecture — defined a Core-Derivative distribution flow that radiates from the web original (Hub) to LinkedIn, Instagram, YouTube, newsletter, and internal magazine

Placed AI across a 6-stage production workflow (plan → research → draft → editing & fact-check → multichannel derivatives → publish & measure), with people reviewing and approving — a Human-in-the-loop structure

Used the system directly to build the Artfesta plan — concept, content structure, detailed composition, execution, and per-channel derivatives

Set up automatic weekly content reports and monthly reports, feeding them into weekly and monthly team meeting agendas

Work

Outcome

✓

Established a system that regularly diagnoses the existing newsroom (Media Hub) strategy and connects it to improvement reporting

✓

Settled an operating loop — weekly and monthly reports → team meeting → content direction adjustment

✓

Accelerated Artfesta planning, composition, and execution in collaboration with the AI team, extending one consistent message across channels

✓

Reduced quality variance by auto-applying brand voice, prohibited-expression, and fact-check checklists to every piece

✓

Embedded the Human-in-the-loop principle into operations — AI for drafts and analysis, people for judgment and publishing