Automations and AI tooling that let one marketer do the work of several.
Agents
draft content at scale
AI creative
on-brand, localized
Personas
pre-launch testing
I build the stack, not just use it: agents and automations (Claude, n8n, live data feeds) that draft newsletters, product updates and research; AI creative production wired to the brand book, including localized video; and synthetic-audience personas I use to pressure-test a campaign from the buyer's side before a single asset is made.
The persona testing is the part people underrate. Simulating the exact person I am writing for and letting them poke holes in a campaign catches the weak angles before launch, while they are still cheap to fix.
At the centre sits an LLM-first knowledge base, a wiki-style repository that acts as the brain of the whole operation. Every workstream and recurring task has its own living context, improved and updated daily with new research, call transcripts, product-development notes and competitor analysis, so nothing has to be re-explained twice.
The setup is deliberate: Claude as the main model, the wiki as its library, Obsidian as the front-end, and MCP connections into every tool I use, from Search Console, Analytics and the marketing Postgres database to the CMS, publishing tools, Google Docs, Premiere Pro and Asana. The result is a model that carries as much context about our operations and strategy as I do, if not more, and knows how to drive the tools itself.
None of it replaces taste. The stack removes the busywork so the time goes into judgement: the story, the hook, the thing that actually makes someone stop scrolling.
Impact
It is how a lean team keeps pace with a much larger one: more shipped, faster, with the craft intact. The analytical, technical side of marketing is where I am most at home.