Introduction to Agentic AI

Document Type: Comprehensive Overview
Version: 1.1 — May 2026
Conceived, designed, and edited by: Ioannis Strikos
Production method: AI-assisted authorship with structured source verification (see Disclaimer)
Status: Living document — regularly updated
Publication location: https://strikos.com/ai/introduction-to-agentic-ai

---

About This Overview

Artificial intelligence is undergoing a fundamental shift. For decades, AI systems waited to be asked. They responded, retrieved, and recommended — but always within the boundaries of a single interaction. Today, a new generation of AI systems is emerging that does not simply respond. It can plan, decide, act, and adapt across extended sequences of tasks, often using tools, memory, and feedback loops under varying degrees of human oversight. This is Agentic AI.

This book-length overview provides a comprehensive survey of Agentic AI — its origins, its architecture, its applications across business functions, and the governance frameworks required to deploy it responsibly. It is written for technology leaders, strategists, and practitioners who already understand generative AI and want to develop a rigorous understanding of what comes next: what agentic systems are, how they are built, how they fail, and what it takes to deploy them at scale.

The book-length overview is organised into seven parts, moving from foundational concepts through to strategic implementation:

• Part 1 establishes the technical foundations — the history of generative AI, the economics of model deployment, and the emergence of multimodal capabilities.
• Parts 2 and 3 examine the architecture of agentic systems and how they connect to existing digital ecosystems.
• Part 4 addresses the risks — cybersecurity, disinformation, and the inherent limitations of autonomous agents.
• Part 5 explores how agents are being deployed across specific business functions.
• Part 6 covers the full deployment lifecycle: crossing the pilot-to-production gap, measuring and improving agent performance, designing human oversight, and building the team that sustains it.
• Part 7 addresses responsible deployment — legal accountability, ethics in practice, and a forward-looking assessment of where enterprise agentic AI is heading.

This is a living document. Chapters are revised when material changes in the landscape warrant an update, and the verification status of each chapter reflects the state of review at the time of its last revision.

---

Authorship and Production

This overview was conceived, structured, and edited by Ioannis Strikos. His contributions include:

• Deciding the purpose, scope, and audience of the overview
• Designing the seven-part structure and the thematic progression across chapters
• Selecting and curating all academic, scientific, and industry sources
• Directing the depth, framing, and editorial approach of each chapter
• Designing the editorial workflow used to produce and verify the content
• Conducting a structured reference audit on each chapter: verifying that cited sources exist, that their metadata is accurately recorded, that the chapter's representation of their findings does not exceed what the sources actually say, and that unsupported inferences are removed or moved to the author's own voice
• Identifying and adding findings from cited sources that were absent from the initial AI-generated text, with targeted additions tied to specific source locations

The text of each chapter was produced using Generative AI, working from those directions and source materials. This is not an unreviewed AI-generated document. It is an AI-assisted manuscript produced through human conception, editorial direction, source selection, revision, and verification. The AI was the writing instrument; the thinking, the choices, and the editorial judgement were human. Responsibility for the final structure, interpretation, editorial choices, and publication of this work rests with the author.

A note on verification scope

The source-checking process described above verifies the relationship between the chapter text and its cited references. It is a more rigorous process than a general editorial review, but readers should understand clearly what it does and does not cover:

What it covers:

• Whether each cited source exists and its metadata (authors, title, journal, date) is correctly recorded
• Whether the chapter's description of a source's findings accurately represents what that source says
• Whether any claims attributed to a source go beyond what the source actually establishes
• Whether material findings in a source are missing from the chapter

What it does not cover:

• Independent verification of the underlying claims in the cited sources themselves — the process checks whether the chapter represents a source accurately, not whether the source itself is correct
• Factual or empirical claims that appear in the chapter without a citation, where there is no referenced source to check against; these represent a verification gap and should be treated with particular caution
• Claims that represent the author's own synthesis, framing, or analytical judgement — these are not subject to citation verification by design, as they reflect editorial contribution rather than sourced findings, and are deliberately distinguishable from cited factual claims
• The validity or quality of the sources selected — that judgement rests with the editor

Verification is also ongoing. At any given time, not all references in all chapters will have completed this process. Chapters are updated as source checks are completed, and the status of each chapter reflects the state of verification at the time of its last revision.

Readers should treat this document as a carefully edited survey — more rigorous than a standard AI-generated text, and more transparent about its limits than most documents of any kind, but not equivalent to peer-reviewed scholarship. Verification of specific claims against the original primary sources before any reliance on this content is strongly encouraged.

Known limitations

• The field of Agentic AI is changing rapidly. Model capabilities, platform features, prices, regulations, and industry practices may change after publication.
• Some claims rely on industry reports, vendor-published material, or rapidly evolving public documentation, which should be read with attention to methodology and possible commercial incentives.
• Source verification checks whether the manuscript represents cited sources accurately; it does not independently validate the underlying research or the future durability of the findings.
• Strategic recommendations reflect the author's synthesis and should be adapted to each organisation's context, risk tolerance, and regulatory environment.

---

How to Use This Overview

Each part is self-contained and can be read independently. Readers with a strong technical background may wish to begin at Part 2. Readers focused on deployment and operations may begin at Part 6. Readers focused on governance, ethics, and strategy may begin at Part 7. Those new to the field should read sequentially.

All chapters include references to the primary academic and industry sources that informed them. Readers are encouraged to consult these sources directly for deeper exploration.

---

License and Citation

This overview is published under the Creative Commons Attribution–NoDerivatives 4.0 International (CC BY-ND 4.0) license.

View full license terms →

Under this license:

• ✅ You may read and share this content freely.
• ✅ You may quote reasonable excerpts with clear attribution, consistent with the license and applicable law.
• ❌ You may not republish this document as your own work.
• ❌ You may not distribute modified or adapted versions of this content.

Required citation

If you use, quote, or reference any content from this overview, you must cite it as follows:

Strikos, I. (2026). Introduction to Agentic AI. Retrieved from https://strikos.com/ai/introduction-to-agentic-ai

Failure to attribute this work when drawing on its content is a violation of the license terms.

---

Disclaimer

This overview was produced using Generative AI under the editorial direction of Ioannis Strikos, who conceived the document, selected and curated all source materials, designed the editorial methodology, and conducted a structured source-verification process on each chapter. Readers should verify factual claims against the cited primary sources before relying on this content for decision-making. Nothing in this document should be treated as legal, financial, medical, cybersecurity, or technical implementation advice; readers should consult qualified professionals before relying on it for consequential decisions. The views expressed are those of the author and do not represent any organisation.