Architecting for Autonomy

AI agents promise to accelerate software systems by removing humans from many real-time decisions. Achieving true autonomy for business actions, security, observability, and other mission-critical activities requires a level of trust that agents have yet to earn in large-scale enterprise systems—especially in domains like banking and healthcare where there are rigid external governance requirements. Architecting for Autonomy teaches you how to update, adapt, and rebuild existing systems so that autonomous AI agents can operate safely and responsibly at scale.

Businesses of all sizes are actively building intelligent systems designed to reason over enterprise context, coordinate work, make recommendations, act, and adapt over time. In this important book, enterprise AI experts Anjali Jain and Philip O'Shaughnessy introduce the new architectural principles required for successful autonomous agents in production systems. By the end, you’ll have a working Agentic Enterprise Framework, a domain-agnostic toolkit of patterns, playbooks, and reference models that you can apply to make almost any architecture ready for autonomous agents.

Architecting for Autonomy is organized around the actual decision points facing an enterprise architect. You’ll work through six principles for governing delegated autonomy, and learn to quantify both the value—and the blast radius—of every agentic decision. From there, you’ll move into the applied architecture layers, redesigning value streams around explicit decision surfaces, classifying where autonomous authority should be exercised or reserved for humans, and designing human oversight. Best of all, you’ll learn to recognize anti-patterns like invisible autonomy, escalation deadlock, and agentic silos before they surface and start causing real problems.

As you go, you’ll see relevant examples that reflect the messy reality of enterprise change. You’ll explore industry case studies in autonomous loan approval, payments orchestration, a clinical pathway agent at a 400-bed hospital experiencing model drift, and even a Formula 1 race strategy! You’ll also see the real-world consequences of getting the architecture right, and the cost of getting it wrong.

By the end of the book, you’ll be able to design enterprise architectures for autonomous systems, turning agentic AI from an isolated pilot into full production-grade capabilities. You’ll know how to effectively assess where autonomy creates value, and where human controls still need to be tight. And by constructing an enterprise roadmap that connects strategy, architecture, governance, and production operations, you’ll be able to guide your whole organization into its agent-enhanced autonomous future.