Project Skynet — Building DLG's AI Operation from 0→1
I led Project Skynet — DLG's internal AI initiative — from a blank page to a fully operating system that the whole agency uses every day. Not a prototype, not a slide. A working platform: a public-facing AI Hub (dlg-ai-hub.com) with live Report / Brief / Knowledge modules, an 8-Gem AI Studio, a three-layer Sites/Drive/NAS file architecture, a China social-media playbook, and a 37-use-case portfolio with phased validation — plus the manuals, workshops, and governance that made the whole agency actually adopt it.
Role
Project Lead · 0→1 architect · platform owner
Year
2024 – now
Outcome
Phase 1 live — 7 use cases validated and in production. 37-UC portfolio approved by leadership. DLG is now positioned to luxury clients as an AI-forward partner, with defensible internal capability behind every pitch.
AI Hub — live modules
01 · p.1
02 · p.2
03 · p.3
04 · p.4
05 · p.5
06 · p.6
07 · p.7
08 · p.8
What I built
Six load-bearing pillars.
AI Hub · dlg-ai-hub.com
Public portal with three live modules — automated Client Reports (BigQuery → LLM insight → VBA export), AI Brief Generator, and a Brand Knowledge SQL chatbot. Shipped end-to-end.
AI Studio · 8-Gem Matrix
Role-based Gemini Gems (Strategy / Account / Copy / Design / Motion / Data / BD / Ops) that guide prompting without requiring prompt skills.
Sites · Drive · NAS Architecture
Three-layer knowledge system — public Sites for clients, Drive for team assets, NAS for heavy brand data and model checkpoints.
China Social Media Guide
Six platforms × ten modules — the living playbook every account lead now uses when briefing AI or pitching luxury clients.
Phase 1 Workshop · 101 pages
All-staff workshop with leadership. Moved AI from abstract discussion to delivered proof — live case studies across 8 UCs.
Every use case is scored on survey signal, time freed per month, team impact, and priority. Phase 1 (Quick Wins) is already live; Phase 2 restructures workflows; Phase 3 institutionalises what works.
