Agentic systems build
When the plan calls for building, I build — AI that does work, not demos.
The record: an agent-operated knowledge system that interviews its owner using the Critical Decision Method; two complete companies designed end-to-end by AI agents working from that system — research, brand, product, website, films — as a documented experiment in how far autonomous building can go; multi-model data and content pipelines in n8n, including an Olympic NOC data scraper and a LinkedIn engine that chains four AI models between scrape and publication; a modular AI skills architecture with a routing layer, built so the system stays maintainable as it grows; and a multi-stakeholder CRM and relationship-weighted forecasting architecture designed for the sports industry.
The company builds were never launched — that was the experiment’s constraint, not its failure — but every artefact and decision log exists, and the working systems run daily. This site was written by one of them.
Questions
What kinds of systems do you build?
Agents, automations and multi-model pipelines: knowledge systems that elicit as well as store, n8n data and content pipelines, multi-agent research and validation machinery, and the architecture that keeps it maintainable as it grows.
Do you advise on builds, or do them?
Both, but this service is the doing — hands-on, documented, handed over working.
How do I know any of it works?
The building page documents the record — including the system this site was written by.