Enterprise AI Architecture Patterns

Enterprise AI Architecture: The Pombaline Cage is a two-volume reference for architects, engineers, and technology leaders building AI systems that must work under the weight of real enterprises — regulated industries, multi-jurisdictional compliance, auditability, vendor risk, and the operational reality that a system must keep running long after the team that built it has moved on. Volume One develops the foundations: a six-layer and a four-plane reference architecture, six reusable structural patterns, the advanced agentic and frontier patterns, the full engineering stack from inference infrastructure and prompt caching to knowledge architecture and foundation-model strategy, and the governance operating model that ties it all together — illustrated throughout with historical analogies from John Snow's cholera map to the Franklin Expedition. Volume Two applies every pattern to twelve industries, from banking and insurance through mining, telecommunications, manufacturing, healthcare, education, public sector, energy, retail, media, and professional services, each with its own data architecture, use cases, governance treatment, case study, and dedicated African chapter. The African perspective runs through both volumes as a first-class concern, not an appendix — covering sovereignty, mobile-money integration, multilingual retrieval, cost-engineered inference, and the architectural adaptations that African deployments demand and that, once made, produce systems more robust than their Northern equivalents. Written by Kennedy Chengeta, PhD, and published by Century Publications, the book is designed to be carried into architecture reviews, governance boards, and vendor negotiations as the reference that answers the question every enterprise eventually asks: we have the models — now how do we build the system around them that we can trust for the next ten years?