Builder

Knowing what these systems cannot do is more useful than knowing what they can.

I build agents, AI products and agentic processes for marketing and growth. In practice: prompt architectures that encode campaign logic and brand voice, agentic workflow systems for content and demand generation programmes, and AI tools that let small marketing teams operate at a scale that would otherwise require significantly more headcount.

That has meant building prompt architectures that function as codified marketing intelligence: structured instructions that give language models the context, constraints and judgement criteria to produce consistent output across paid media, content, ABM and events. It has also meant designing end-to-end agentic processes for programmes that would previously have required a team to coordinate manually, from content production pipelines through to full campaign distribution workflows.

What makes this useful rather than just interesting is the underlying position: I build these systems because I understand where they break down. An agent that automates execution without a human capable of interrogating its output is a liability. The build and the strategic judgement have to sit together.