Every January I look at the previous twelve months of bookmarks, sticky notes, Kindle highlights, and quietly-bought print copies, and I try to work out which books actually changed how I do my job. Most years the list is three or four books. 2025 was unusual. Half of the list is books that have been on the shelf for a decade, which I re-read on purpose because the field was moving fast enough that I wanted to check whether the foundations had moved. The other half is books written in the last two years, mostly trying to make sense of agentic AI as an architecture concern rather than a model concern.
These are in roughly the order I think a practising EA should read them, not in the order I read them. If you have time for two, read the first two. If you have time for ten, the list compounds. If you have time for none, the elevator-pitch versions are below.
1. The Phoenix Project — Gene Kim, Kevin Behr, George Spafford (2013)
The cheapest org-design lesson on the market and still the right one to hand to a peer who has not internalised the Four Types of Work. The technology is dated, the organisational pathology is not. I re-read it in 2025 and was surprised by how much of it is really about cognitive load before that term existed in the literature.
2. Enterprise Architecture As Strategy — Ross, Weill, Robertson (2006)
The source of the Operating Model quadrant, which I still use weekly as a diagnostic in conversations where the company thinks it is one operating model and is actually another. The case studies are period pieces. The framework is not.
3. The Software Architect Elevator — Gregor Hohpe (2020)
The book I quote most in conversation. The elevator metaphor does sixty percent of the work in any conversation about what enterprise architecture is supposed to be doing. The “architects sell options” framing has changed how I write ADRs.
4. Designing Data-Intensive Applications — Martin Kleppmann (2017)
The only systems book I have re-read on purpose three times. The derived-data chapters have aged into being the most important section. If you read it five years ago and not since, re-read chapters 10 through 12.
5. Team Topologies — Matthew Skelton, Manuel Pais (2019)
Short, opinionated, and the vocabulary has already escaped the book. The cognitive-load framing has changed how I evaluate internal-platform investments. I no longer accept “this will make engineers more productive” as a sufficient justification.
6. Building Evolutionary Architectures, 2nd ed — Ford, Parsons, Kua, Sadalage (2022)
The second edition is the book the first edition wanted to be. The fitness-function concept is the single most useful architectural vocabulary addition of the last decade, and now gets the treatment it deserves. Read this especially if you work on agentic systems; the framing transfers almost without modification.
7. The TOGAF Standard, 10th Edition — The Open Group (2022)
The first version of TOGAF I would actively defend to a sceptical practitioner. The Fundamental Content is small enough to be honest about what it does and does not promise; the Series Guides are explicitly optional. Worth reading even if you have written TOGAF off based on a previous edition.
8. Mastering TOGAF 10 — Roshan Gavandi (2023)
The companion the TOGAF Standard needed. Worked examples that are more concrete than the standard’s own, and practice questions that are calibrated to the actual OGEA-101 / OGEA-102 exam style. The voice is dry; the value is real.
9. Enterprise Architecture in the Age of Agentic AI — Maher Dahdour (2026)
The first sustained attempt at re-framing the EA discipline for a world in which agents are first-class. Uneven, but necessary. The compounded-autonomy risk discussion is the best treatment of that topic I have read in any EA book.
10. Agentic Architectural Patterns — Ali Arsanjani, Juan Bustos (2025)
The first serious patterns catalogue for agent systems. The Verifier pattern alone is worth the book. Pair it with Dahdour for the EA-level framing and with Building Evolutionary Architectures for the underwriting framing both books quietly assume.
The honourable mentions
A few books I read in 2025 that did not make the top ten but deserve a nod:
- Fundamentals of Software Architecture (Richards & Ford). The textbook Building Evolutionary Architectures assumes you have read.
- Accelerate (Forsgren, Humble, Kim). The DORA-metrics book. If you have not internalised the four key metrics yet, this is the cheapest education.
- Domain-Driven Design (Evans). Still load-bearing for any conversation about bounded contexts.
None of these moved my decision-making in 2025, which is why they are not in the top ten. They are still books I would defend.
Reading themes from 2025
Two themes were unavoidable.
Agentic systems are an architecture problem, not a model problem. Half of the most interesting work I read in 2025 was the field collectively realising that the interesting questions about agents are at the system-design level, not at the model level. Agentic Architectural Patterns and the Dahdour book are the front edge of this. There will be much more.
The classics still hold up. The other surprise of the year was how well the classics held up against new pressure. Enterprise Architecture As Strategy still works as a diagnostic for operating models that include autonomous agents. Designing Data-Intensive Applications still gives you the right vocabulary for reasoning about state in distributed systems — including agentic ones. Team Topologies still names the team-design problem correctly. The frameworks survive the re-framing.
If there is a unifying lesson, it is that the altitude of EA thinking has not moved. The specific decisions have moved a lot.
For a thematic take on the classic / bleeding-edge split, see Classic vs Bleeding Edge: how EA reading has evolved. For a study path through the TOGAF books, see TOGAF 10 self-study path.