Book 1
CONVERGE
From AI Curiosity to Business Reality
How does a business adopt AI?
Publishing 7 December 2026: CONVERGE on AmazonAbout the book: CONVERGE
Book 2 of 3 · The Enceladus AI series
ENGINEER
From Pilot to Production
The pilot worked. Then it met real customers, real data and real edge cases, and it stopped working.
Read it if you are the IT lead, head of engineering or technically minded director who has to commission, build or sign off an AI system, or a CONVERGE reader who wants to know exactly what they are approving.
About the book
ENGINEER is about the distance between a demonstration and a system a business can rely on. It is written by an engineer for engineers, architects and the technically minded directors who sign these systems off, at the moment when machines write much of the code and the engineer's job has moved to specifying, deciding, reading and owning.
It covers what is inside the models, from the transformer to mixture of experts; the workbench of editors, agents and pipelines the work is now done in; when a predictable workflow beats an autonomous agent; retrieval, tools and confidence-based escalation; reference architectures for the quick wins every business starts with; when a language model is the wrong tool; eight archetypes drawn from systems that work at scale; evaluation before scale; security for systems that read the outside world; and the gates from pilot to production.
The worked examples are the author's own: SADIE, a support-agent framework, and Sign2Approve, a product built in a month.
Inside the book
23 chapters in 4 parts.
- Part 1
What you are building, and with what
- The job has changed
- What an AI system actually is
- Inside the models
- The workbench
- From data position to data pipeline
- From brief to specification
- Part 2
Patterns and the parts
- Workflows before agents
- Prompts, context and structured output
- Tools, MCP and least privilege
- Retrieval done properly
- Confidence and escalation
- Agents and multi-agent systems
- A support agent, end to end
- Reference architectures for the quick wins
- When a language model is the wrong tool
- The machine-learning system around the model
- Larger systems, and the archetypes they share
- Part 3
Making it work, and keeping it working
- Evaluation before scale
- Reliability, cost and observability
- Security for systems that read the outside world
- Platform choices, portability and lock-in
- Part 4
Ownership
- The gates from pilot to production
- Handing the system to the business
Book details
- Author
- Ian Lawton
- Series
- The Enceladus AI series, book 2 of 3
- Publisher
- Enceladus AI Publishing
- Publication date
- 7 December 2026
- Paperback
- ISBN 978-1-0655577-3-9
- Hardback
- ISBN 978-1-0655577-4-6
- Kindle
- ISBN 978-1-0655577-5-3
Also in the series
Book 3
GOVERN
From AI Risk to Responsible Advantage
How do you manage AI safely and responsibly?
Publishing 7 December 2026: GOVERN on AmazonAbout the book: GOVERN
