Most marketers treat AI as a faster typewriter. Throw a prompt at Claude, get 2x the copy in half the time, declare victory. That's not leverage—that's just outsourcing repetition.
Real leverage lives in systems. Agents that route work based on context. Retrieval pipelines that ground LLMs in your actual product docs so hallucination goes to zero. Evaluations that catch quality drift before it ships. A prompt library your whole team uses consistently.
I've put this into practice on the blog — building systems where AI handles structure and drafts, with human validation at every step. The time saved goes back into quality assurance and more strategic work. The tools matter less than the architecture.
Models & Chat
- Claude — Daily driver. Context window, instruction-following, reliability. Primary for everything.
- Gemini — Nano for on-device/lightweight tasks, Flash for speed-sensitive workflows, Pro for complex reasoning. Exploring long-context use cases.
Agents & Workflows
- Claude Code — Building this page, automating code review, managing GitHub workflows. Core daily tool.
- n8n — Custom automation. Webhooks, integrations, cross-system orchestration.
- Custom Agents — Team workflows. Prompt library capture. Evangelized adoption across APAC peers.
Project Management
- Linear — Issue tracking, sprint planning, and roadmapping. Daily driver for managing backlog, priorities, and long-term blog enhancements and planning.
The blog is a live experiment — every essay goes through a structured workflow where AI handles the heavy lifting and I validate, add references, and assure quality and accuracy at each stage. Not theoretical. It works.
- Jun 2025 — Prompt engineering rigor. Building eval rubrics that catch quality drift in content pipelines before humans review.
- Sept 2025 — RAG at scale. Moving beyond vector databases into hybrid retrieval for long-context system design.
- Dec 2025 — Evaluating Hex for natural-language analytics. Can we make data self-serve for the team?
- Jun 2026 — MCP servers and tool-use patterns. Connecting LLMs to live systems (GitHub, Slack, analytics) to enable agentic flows without hallucination.
- Dec 2026 — Building long-term memory systems for work and personal life. Structuring persistent context so AI tools stay useful across sessions and projects.