For fifty years software obeyed a simple compact: the same input returned the same output. Then the machine learned to answer back, and a discipline built on predictability found that its ground had moved. Software AI Engineering is a complete account of what comes next - the engineering, alignment, governance, deployment, and philosophy of artificial-intelligence systems - and it makes that account, by deliberate design, without a single line of code.
Its argument is that the hardest decisions in artificial intelligence are not made at the keyboard at all, but in the prior questions of what ought to be built, how its behaviour shall be judged, who answers for it when it errs, and whether, all things weighed, it should exist. Across fifteen parts and sixty-seven chapters the book carries the reader from the foundations of the new discipline through AI product engineering, system architecture, evaluation and testing, MLOps and LLMOps delivery, safety and abuse resistance, and the governance regimes now taking shape under the NIST framework and the EU AI Act, into the domains where AI meets medicine, law, and finance, and at last into the philosophy that was implicit all along.
It is written for the people who must actually answer for these systems - engineers and product managers, founders and executives, board members, regulators, clinicians, and the thoughtful citizen - and it equips each with the questions to ask before a model is trusted to act in the world. Rigorous where it must be and ironic where it can afford to be, it treats responsible AI not as a compliance afterthought but as the very substance of the craft.
A reference, an argument, and an education in judgement, for an age in which the capacity to direct intelligent systems matters more than the capacity to write their code.
