Architecture paper
The Four-Part System: a reference architecture for governed enterprise AI
By Richard White ·
Free, no form.
Abstract
The reference architecture itself. Enterprise AI needs a dependable route from business intent to information, permitted execution and an observable result. This sixteen-page paper organizes that route around four responsibilities (the brain, the nervous system, the body and the immune system) and the contracts that connect them, with a control plane tying decisions to evidence.
It is written for enterprise architects and the data, technology and security leaders who must review a target state. It works through information contracts, execution, environments, security and control evidence, then follows one illustrative claims request through all four parts, step by step.
It is vendor neutral by design: a component is acceptable when it preserves the contracts and the control behavior, and the paper ends with six questions you can review your own architecture against.
In the paper
Three passages, as published
The execution boundary
The brain may propose a plan. Access and action authority are enforced here, in the execution path, and not above.
Recovery
A last approved version is a recovery option only if it still meets current policy and dependency requirements.
Substitution
A similar interface alone does not demonstrate that equivalence.
Next step
Discuss this paper against your own program.
A free sixty-minute working session with Richard on your top gaps. You leave with a one-page summary within three business days, and three next questions.