I am not a trader, a quant, or a developer.
This is the documented story of how I taught myself n8n, Claude Code, and AI agent orchestration well enough to build a fully governed, audited trading system from scratch — while between client engagements in 2026.
It is not a success story. It is a method story. The system went through multiple versions. One was rejected. The final version was frozen, audited, and governed. The document shows the real numbers, the real decisions, and what they reveal about governing any autonomous AI system.
The same method I bring to every client: question, test, reject what doesn't hold up, freeze what does, govern the rest.
What's inside.
- The starting point — zero trading knowledge, zero coding background, and a decision to build something real.
- The build process — how Claude Code, n8n, and agent orchestration were used to construct the system.
- The rejected version — what failed, why it failed, and the decision to scrap it.
- The frozen model — how the final version was validated, audited, and locked down.
- The governance layer — what controls were put in place and why they matter for any autonomous AI.
- Real numbers — actual performance data, not marketing claims.
- What this means for your organisation — how the same method applies to any AI system that makes decisions on your behalf.
Best read alongside The Executive's Guide to AI Governance and The Executive's Guide to the AI Product Lifecycle. The case study shows the method in practice; the guides give you the framework to apply it.