01 / 09
AgentOS
An operating model for enterprise marketing, run by AI agents.
- Role
- Architect & Product Lead
- Context
- Strategic AI initiative, Emirates account, 2026
- Stack
- Claude, Gemini, Pinecone, Node.js, Express, React, PostgreSQL
AgentOS — the film
- 14specialised agents
- 4end-to-end workflows
- 3governance layers
The problem
Enterprise marketing at the scale of 100+ markets is bottlenecked by manual, repetitive work spread across disconnected tools. The question wasn't 'which tool' — it was whether the operating model itself could be rebuilt around autonomous agents.
What I did
- 1Designed a 14-agent framework spanning the full lifecycle — campaign management, CRM, content, segmentation, automation, QA, legal, brand governance, analytics and competitive intelligence.
- 2Structured the platform into three layers — Strategic, Execution, Control & Validation — with clear governance boundaries and human-in-the-loop oversight.
- 3Engineered the stack: LLM orchestration (Claude, Gemini), RAG retrieval over Pinecone, real-time multimodal capabilities, on a Node.js + PostgreSQL backend.
- 4Defined four end-to-end workflows mapped to measurable outcomes: Campaign Creation Engine, Post-Campaign Intelligence Loop, Loyalty Upsell Engine and NPS Recovery Automation.
Outcome
- A scalable enterprise blueprint bridging traditional marketing automation and agentic operations.
- Presented to senior stakeholders as the reference architecture for AI-led marketing transformation.