Agentic Software Engineering: Building Reliable Software with AI Agents, Specs, and Human Oversight

By H. D. Liao

Software engineering is entering a new era. For decades, developers wrote most of their code by hand, interpreted requirements, ran tests, reviewed changes, and maintained systems using familiar tools and workflows. Generative AI has already changed that process through code completion, chat-based programming help, and AI-native development tools. Now, AI agents are beginning to participate more directly: they can inspect repositories, edit files, run commands, generate tests, create pull requests, and produce evidence for human review.
Agentic Software Engineering explains this shift and presents a disciplined framework for working with AI agents as active participants in the software development lifecycle. This book does not treat agents as replacements for software engineers, nor does it reduce the future of engineering to mere prompt writing. Instead, it argues that as agents take on more implementation work, human responsibility moves to a higher level: defining goals, writing clear specifications, designing architecture, managing context, supervising tools, reviewing evidence, and deciding what is safe to merge, deploy, and maintain.
This book is written for students, educators, software developers, technical leads, engineering managers, business leaders, technology executives, and anyone preparing for a world in which AI agents become routine members of the software development process. Its central message is simple: the future of software engineering is not code without humans. It is disciplined human-AI software engineering, in which powerful automation is guided by clear specifications, bounded by safe environments, verified by evidence, governed by accountable processes, and ultimately directed by human judgment.

Brief Contents
Part I - The Paradigm Shift in Software EngineeringChapter 1: From Traditional Software Engineering to Agentic Software EngineeringChapter 2: Vibe Coding, Agentic Coding, and Structured Agentic Software EngineeringPart II - Foundations of Human-Agent Software WorkChapter 3: Human and AI Roles in the New Software TeamChapter 4: Agent Workbenches, Tools, Agent Skills, and Harness EngineeringChapter 5: Managed Agents and Serverless Agent RuntimeChapter 6: Artifacts as Interfaces Between Humans and AgentsPart III - Agentic Engineering ProcessesChapter 7: Briefing Engineering and Spec-Driven DevelopmentChapter 8: Agentic Design Patterns and Workflow ArchitecturesChapter 9: Context Engineering, Memory, and Agent MentorshipPart IV - Agentic DevOpsChapter 10: Agentic CI/CD, GitOps, and Deployment PipelinesChapter 11: Policy-as-Code, Eval Gates, and Agent ObservabilityPart V - Quality, Security, and GovernanceChapter 12: Testing and Reviewing Agent-Generated SoftwareChapter 13: Security Risks in Agentic Software EngineeringChapter 14: Governance, Provenance, and Responsible AI UsePart VI - Maturity and the FutureChapter 15: The Agentic Software Engineering Maturity ModelChapter 16: Future Directions and Open ProblemsAppendicesAppendix A: Glossary of Agentic Software Engineering TermsAppendix B: BriefingScript TemplateAppendix C: MentorScript TemplateAppendix D: LoopScript TemplateAppendix E: Spec-Driven Development Packet: requirements.md, design.md, and tasks.mdAppendix F: Merge-Readiness Pack TemplateAppendix G: Sample Agentic Code Review ChecklistAppendix H: Sample Security Review Checklist

• Format: Paperback
• Language: English
• ISBN: B0H7TBH7WM
• ISBN13: 9798994496640
• Release Date: July 2026
• Publisher: Mary's Book Publishing
• Length: 746 Pages
• Weight: 2.80 lbs.
• Dimensions: 1.5" x 7.0" x 10.0"