Available nowPaperback and kindle
The Contextual Agentic Enterprise
From systems of record to governed AI agents and enterprise outcomes
Most enterprise AI stalls after the demonstration. This book is about what comes next: designing the context, authority, controls and evidence that let an agent take part in real work.
- Give agents the organisation's own context to act on, not documents retrieved and hoped for.
- Grant, scope and revoke authority explicitly, with controls that do not depend on the model behaving well.
- Produce the evidence a reviewer needs as a by-product of the work, not an afterthought.
Published by Polymath Signal Press. Buy links go to Amazon; the price you see there depends on your store.
Who it is for
- Enterprise and solution architects designing agent systems that touch systems of record
- Engineering leaders responsible for what those systems do in production
- AI governance, risk and assurance practitioners who have to review them
- Platform and security engineers implementing the boundaries
What it assumes
- Working knowledge of enterprise systems and integration
- Familiarity with how large language models are used in applications; no research background assumed
- No prior experience building agent systems is required
Read a sample first
The Introduction in full, plus the opening pages of the chapters on enterprise agent design patterns and Agentic Flow Mining. Free, no sign-up.
Enterprises possess enormous amounts of data, knowledge, process history and technical capability, yet struggle to assemble the right context, evidence and authority at the moment a consequential decision must be made.
What you will learn
- Design a contextual substrate an agent can act on, rather than retrieving documents and hoping
- Model authority explicitly — granted, scoped, delegated and revocable — instead of implying it in a prompt
- Apply Zero Trust principles to agent actions, including the write path
- Produce evidence as a by-product of agent work, so assurance is not archaeology
- Place human oversight where it can still change the outcome
- Reason about the economics without the denominator sleight of hand
About the book
Most enterprise AI programmes stall at the same place. The demonstrations work. The pilot is convincing. Then someone asks what happens when the agent is wrong about a customer’s contract, who authorised it to make that change, and what the auditor will be shown — and the answer turns out to be that nobody designed for any of that.
This book is about the part that comes after the demonstration. It argues that an agent becomes useful to an enterprise at exactly the moment it is given authority over real work, and that authority is an architectural problem, not a prompting one.
Two things have to be true at once. The agent has to be grounded in the organisation’s own context — its semantic model, its entitlements, its records — rather than inferring the business from documents it happened to retrieve. And the boundaries around it have to be enforced by systems that do not depend on the model behaving well: deterministic controls around variable behaviour, not a promise that the variation will stop.
The book works through what that means in practice. How to build the contextual substrate an agent reads from, and what it costs to keep it true. How authority is granted, scoped, delegated and revoked, and why the write path deserves more design attention than the read path. How to enforce boundaries with Zero Trust principles that were built for exactly this problem. How to produce evidence as a by-product of the work rather than reconstruct it under pressure afterwards. How to keep humans in a position to oversee something faster than they are. And how to be honest about the economics, including the denominator that most published figures quietly omit.
It is written for people who will have to build, review or defend these systems.
Contents
- Preface
- Introduction
Part I — Historical and conceptual foundations
- Chapter 1 The enterprise intelligence problem
- Chapter 2 From expert systems to agentic computing
- Chapter 3 Deterministic, probabilistic and agentic systems
Part II — Knowledge, context and semantic foundations
- Chapter 4 Enterprise knowledge and organisational memory
- Chapter 5 Context engineering and in-context adaptation
- Chapter 6 Retrieval-augmented generation, hybrid retrieval and semantic search
- Chapter 7 Ontologies, knowledge graphs, evidence and time
Part III — Agentic architecture and interoperability
- Chapter 8 Enterprise agent anatomy
- Chapter 9 Enterprise agent design patterns
- Chapter 10 Multi-agent coordination, frameworks and enterprise orchestration
- Chapter 11 MCP, A2A and capability interoperability
- Chapter 12 Agentic Flow Mining (AFM)
Part IV — Trust, security, governance and accountability
- Chapter 13 Zero Trust for the agentic enterprise
- Chapter 14 Threat modelling, identity, authorisation and tenant isolation
- Chapter 15 AI sovereignty, data residency and sovereign inference
- Chapter 16 Governance, explainability, evidence and human accountability
Part V — Engineering, integration and operations
- Chapter 17 Evaluation, reliability, observability and replay
- Chapter 18 Connecting the existing enterprise: Legacy integration and incremental modernisation
- Chapter 19 Cloud and platform architecture: Cloud-neutral, AWS and Azure
- Chapter 20 Enterprise model portfolios, routing, benchmarking and intelligence economics
- Chapter 21 Memory, continual learning and adaptive enterprise intelligence
Part VI — Economics, operating model and transformation
- Chapter 22 Operating model, FinOps, cost and productivity
- Chapter 23 From experimentation to enterprise capability
Part VII — End-to-end implementation
- Chapter 24 Three applications of the Contextual Agentic Enterprise
- Chapter 25 Building Enterprise Digital Brain, the companion CAE Intelligence Fabric
Part VIII — Research frontiers and the next enterprise
- Chapter 26 Open problems and the next enterprise
Appendices
- Appendix A Enterprise agent pattern cards
- Appendix B Zero Trust agentic security checklist
- Appendix C Architecture decision guide and example ADRs
- Appendix D Evaluation and testing cookbook
- Appendix E Enterprise model, router and memory benchmark scorecards
- Appendix F Cloud implementation mapping
- Appendix G Enterprise structured agent skill standard
Formats
Three formats, each linked to its Amazon listing. Check the listing for the current price in your store.
| Format | What it is | Status | Where to buy |
|---|---|---|---|
| Paperback, black and white | Black-ink interior, 7 × 10 in. Diagrams print in black and white. 7 × 10 in, 666 pages, black ink ISBN 978-1-7649957-0-2 |
On sale | Amazon |
| Paperback, premium colour | Premium-colour interior, 7 × 10 in. Diagrams keep their colour. 7 × 10 in, 666 pages, premium colour ISBN 978-1-7649957-1-9 |
On sale | Amazon |
| Ebook | Reflowable ebook for Kindle. ISBN 978-1-7649957-2-6 |
On sale | Amazon Kindle |
Companion material
The book points at a companion repository rather than reprinting long listings. That is a deliberate trade: code that is going to change belongs somewhere it can change, and a printed path cannot be corrected once it is printed.
One address is printed in the book. Everything else is reached from there, so a change of arrangement costs a commit rather than a reprint. The public repository holds the full listings that were abridged in print, the material of Appendices C, E, F and G (decision records, scorecards, cloud mapping and the skill standard), and two small runnable samples.
All of that is public and needs no sign-in. The listings are illustrative and the samples are small; the complete Enterprise Digital Brain environment is not supplied. Anything offered separately to registered readers is optional and is described, with its current status, on the companion code page.
Updates on the book and its companion material
A short note when companion material is released or the book is updated. Nothing else.
The email list is not open yet, and rather than take an address we cannot yet store, here is what works today: the article feed carries every piece as it is published, collects nothing and asks for nothing. You can also follow the Contextual Agentic Page on LinkedIn (opens in a new tab). Or write to contact@contextualagentic.com and we will add you by hand when the list opens.
No tracking, no address stored, no account.