Agentic AI: From Foundations to Frontiers explains the concept of intelligent systems that can act autonomously to achieve specific goals. It begins by covering the core foundations such as machine learning, natural language processing, and reinforcement learning. The book describes how AI agents perceive their environment, make decisions, and execute tasks efficiently. It also explains the architecture of agentic systems, including planning, memory, and tool integration. Various real-world applications in healthcare, education, and business are highlighted to show its practical impact. The discussion extends to advanced topics like multi-agent systems and collaborative AI environments. It emphasizes the growing role of large language models in enabling intelligent agents. Ethical considerations such as bias, safety, and accountability are also addressed. Additionally, the book explores challenges like reliability and high computational requirements. It concludes by presenting future directions where AI systems become more autonomous and transformative across industries.
Preface
1. Evolution of Artificial Intelligence and the Emergence of Agentic Systems1
1.1 Introduction: From Intelligence to Execution1
1.2 Phase I: The Analytics Era — Descriptive and Diagnostic Intelligence3
1.3 Phase II: The Machine Learning Era — Predictive Intelligence10
1.4 Phase III: The Large Language Model Era — Linguistic Intelligence13
1.5 The Execution Gap: Why Enterprises Need More Than LLMs16
1.6 Phase IV: Agentic AI — From Answers to Actions24
1.7 Defining Agentic AI25
1.8 Human–AI Collaboration as a Design Principle27
1.9 Why Traditional Automation Fails Where Agentic AI Succeeds29
1.10 Chapter Summary31
2. Agents vs Chatbots and the CORE² Execution Model35
2.1 Introduction: Why “Agents” Are Not Chatbots35
2.2 Chatbots: Conversational Response Systems37
2.3 Why Chatbots Fail in Enterprise Workflows38
2.4 Agents: Goal-Directed Execution Systems41
2.5 Agents Are Systems, Not Models42
2.6 Chatbot vs Agent: A System-Level Comparison43
2.7 Why Execution Requires a Behavioral Framework44
2.8 The CORE² Execution Model: Overview45
2.9 Context: Interpreting Intent and Constraints47
2.10 Orchestrate: Planning and Coordination47
2.11 Run: Deterministic Action Execution48
2.12 Evaluate: Validation and Quality Control48
2.13 Escalate: Human-in-the-Loop Governance49
2.14 CORE² as a Closed-Loop System50
2.15 Why CORE² Reduces Hallucinations50
2.16 Agents as Accountable Digital Workers51
2.17 Chapter Summary55
3. Single-Agent Architectures: Reasoning, Tools, Memory, and Validation59
3.1 Introduction: From Behavior to Architecture59
3.2 The Single-Agent Architectural Stack60
3.3 Context Interpreter62
3.4 Reasoning and Orchestration Engine65
3.5 Tool Execution Layer68
3.6 Memory Subsystems70
3.7 Validation and Evaluation Layer73
3.8 Escalation and Governance Interface75
3.9 End-to-End Flow of a Single Agent-A Complete Walkthrough77
3.10 Common Failure Modes in Single-Agent Systems82
3.11 Designing for Reliability83
3.12 Chapter Summary83
4. Multi-Agent Systems: Coordination, Roles, and Topologies87
4.1 Introduction: Why One Agent Is Not Enough87
4.2 What Is a Multi-Agent System?88
4.3 Why MAS Are Essential in Enterprise Contexts89
4.4 Canonical Agent Roles in MAS91
4.5 Mapping CORE² to Multi-Agent Systems94
4.6 Communication Between Agents96
4.7 Coordination Topologies96
4.8 Conflict Detection and Resolution105
4.9 Synchronization and State Management108
4.10 Failure Handling in MAS110
4.11 Detailed Multi-Agent Workflow: Contract Review System111
4.12 Design Principles for Reliable Multi-Agent Systems116
4.13 Chapter Summary117
5. Memory, Knowledge Grounding, and Retrieval-Augmented Generation119
5.1 Introduction: Why Memory Matters More Than Models119
5.2 Memory as a First-Class System Component120
5.3 Types of Memory in Agentic Systems121
5.4 The Problem of Hallucination123
5.5 Knowledge Grounding: The Core Principle125
5.6 Retrieval-Augmented Generation (RAG)126
5.7 Components of a RAG System127
5.8 RAG Within the CORE² Model131
5.9 Complete RAG Implementation: Retail Pricing Agent131
5.10 Evaluation and Grounding Validation135
5.11 Failure Modes in RAG Systems136
5.12 Memory Governance and Safety137
5.13 End-to-End Example: RAG in the Retail Pricing Agent139
5.14 Designing Memory-Aware Agents: A Practical Framework141
5.15 Chapter Summary141
6. Tool Calling and Action Execution in Agentic AI145
6.1 Introduction: From Knowledge to Action145
6.2 Why Language Generation Cannot Execute Work146
6.3 Defining Tools in Agentic Systems147
6.4 Properties of Enterprise-Grade Tools148
6.5 The Tool Execution Layer150
6.6 Tool Selection and Authorization155
6.7 Complete Tool Implementations: Retail Pricing Agent156
6.8 Error Handling and Recovery162
6.9 Tool Calling Within the CORE² Model164
6.10 Tool Outputs as Evidence165
6.11 Why Tool Calling Reduces Hallucinations165
6.12 Security and Safety in Tool Execution166
6.13 End-to-End: Tool Execution in the Retail Pricing Agent167
6.14 Design Principles for Actionable Agents169
6.15 Chapter Summary170
7. Reasoning, Planning, and Autonomy Levels in Agentic AI173
7.1 Introduction: Intelligence Is Not Autonomy173
7.2 Reasoning in Agentic Systems174
7.3 Why Reasoning Alone Is Insufficient177
7.4 Planning as Structured Decision-Making178
7.5 Planning Within CORE²180
7.6 Autonomy as a Controlled Design Choice181
7.7 The Five Autonomy Levels182
7.8 Autonomy Is Not Uniform Across Actions184
7.9 Confidence, Risk, and Autonomy185
7.10 Escalation as an Autonomy Boundary186
7.11 Failure Modes in Reasoning and Planning187
7.12 Designing Responsible Autonomy: A Framework189
7.13 Chapter Summary189
8. Safety, Governance, and Explainability in Agentic AI193
8.1 Introduction: Why Safety Is an Architectural Concern193
8.2 The Risk Landscape of Agentic Systems195
8.3 Governance as a System Design Principle198
8.4 Policy Enforcement in CORE²199
8.5 Safety Boundaries and Guardrails199
8.6 Explainability: Beyond Model Interpretability201
8.7 Three Forms of Explainability202
8.8 Auditability and Traceability203
8.9 Human-in-the-Loop Governance205
8.10 Safety in Multi-Agent Systems206
8.11 Common Safety Failures and Anti-Patterns208
8.12 Designing Trustworthy Agentic Systems: A Checklist210
8.13 Chapter Summary211
9. Failure Modes, Evaluation, and Monitoring of Agentic Systems215
9.1 Introduction: Failure Is Inevitable, Collapse Is Not215
9.2 Taxonomy of Failures in Agentic AI216
9.3 Common Failure Modes in Single-Agent Systems219
9.5 Evaluating Agentic Systems: Beyond Accuracy224
9.6 Task-Level Evaluation225
9.7 Step-Level Evaluation225
9.8 Confidence and Calibration226
9.9 Monitoring Agentic Systems in Production227
9.10 Incident Detection and Response228
9.11 Stress Testing and Simulation229
9.12 Continuous Improvement of Agentic Systems231
9.13 Anti-Patterns in Evaluation and Monitoring231
9.14 Designing for Resilience232
9.15 Chapter Summary233
10. Enterprise Case Studies and Applied Patterns in Agentic AI237
10.1 Introduction: From Theory to Practice237
10.2 Case Study 1: Document-Driven Compliance Review (Legal / Regulatory)238
10.3 Case Study 2: Healthcare Information Synthesis (Clinical / Regulated)243
10.4 Cross-Cutting Architectural Patterns249
10.5 Choosing Autonomy Levels by Domain252
10.6 Measuring Success in Enterprise Deployments252
10.7 Lessons Learned from Enterprise Practice253
10.11 Chapter Summary255
Appendix258
Appendix A258
Appendix B260
Appendix C260
Appendix D261
Appendix E261
Appendix F262
Appendix G262
Appendix H263
Appendix I263
Appendix J264
Reviews
There are no reviews yet.