New Release — Digital Edition
The ABCs of Agentic AI
Controlling Autonomous Action at Runtime
By Adam DiStefano · Creator of the ACR Standard
$9.99
PDF + EPUB · Instant download
AI agents don't just generate answers. They take actions — and those actions have consequences.
When an AI system can plan multi-step workflows, invoke external tools, access production databases, send communications, and operate with increasing autonomy, traditional governance frameworks fail. Static policy documents, periodic risk assessments, and pre-deployment checklists cannot govern systems that make decisions and execute them in real time.
This book provides the definitive implementation guide for controlling autonomous AI action at runtime. It introduces the ACR Standard — the mandatory control plane through which protected AI actions must pass before execution — and translates it into practical, enforceable governance for enterprise environments.
What's Inside
Runtime Enforcement Architecture
How to build the control plane that intercepts, evaluates, and gates every consequential AI action before it executes.
The ACR Standard in Practice
The complete implementation doctrine — identity, purpose, policy, risk, and human authority enforced at execution time.
Agentic Threat Defense
Prompt injection, RAG poisoning, tool misuse, autonomy drift, and chaining attacks — with concrete countermeasures.
Evidence-First Governance
Audit-grade explainability: traceability, rationale, and evidence you can defend to customers, regulators, and boards.
Key Topics
- Defining consequential action classes and enforcing human decision rights at execution time
- Building kill switches, rollback mechanisms, safe mode, and containment drills
- The foundational rule: model output is not authority — policy must be enforced at the point of action
- Tiering AI use cases from low-risk to mission-critical based on impact and autonomy level
- Mapping runtime governance to NIST AI RMF, ISO/IEC 42001, and enterprise compliance frameworks
- Designing audit-grade evidence: traceability + rationale + proof of human authority
- Governing multi-agent systems, tool chains, and cross-boundary workflows
Free Governance Toolkit
Access interactive tools built from this book — ACR Maturity Assessment, Agent Risk Classifier, Runtime Incident Scenarios, and downloadable document generators.
Open the ToolkitReader Reviews
No reviews yet. Be the first to share your thoughts.
Try the Interactive Tools
Built on the same ACR controls covered in the book — assess, classify, and test for free.
Get the Book
The definitive guide to governing agentic AI in production. PDF + EPUB, instant download.