An AI ranked forty-one bids. A human read ten and signed.
Volume 1
Cheap to Make,
Costly to Check
How AI shifts the real cost of decisions — and who in your organization ends up paying it.
Rank 19: never read, fourteen years in the building
The argument
A purchasing system ranks forty-one bids in seconds. The buyer spends two days on the first ten. Her name goes on the contract. The file records a human decision.
What a system produces has become almost free. What it takes to check that output, and to answer for it, costs what it always did. The difference lands somewhere — and in most organizations, nobody decided where.
This book puts a name and a number on that leftover, and asks the two questions about it: who owns it, and can they carry it? The most expensive errors sit where those two answers meet — where someone owns the decision but has no time to check what it rests on, and where nobody looks at all.
It is a book for leaders responsible for decisions that are increasingly prepared by systems they did not choose, cannot fully inspect, and may still have to defend.
How this shows up
The case in the book is not the exception. In most organizations it looks like one of these:
- A tool is approved, and two executives give two different answers to the question of who checks what comes out of it.
- A department has been working with AI for months. Compliance hears about it when the first question arrives from outside.
- Sign-off rests with a person who cannot retrace the system's groundwork in the time available for it.
- There is a policy on using AI. Who decides, who checks, and who answers for it is not in there.
- Where AI is officially in use is easy to name and sits on a project list. Whoever works with it without approval is not on that list, and will not say so as long as it could cost them.
What is on this site
- Sources All ninety references, chapter by chapter, each with its type: what was measured, on whom, and what limits the finding.
- Claims Nine assertions the book makes, how well each is supported, and what would refute it. Anyone who refutes one is doing the author a favor.
- Tools The eleven tools and the twenty-seven things to do, with the chapter each is built in and the time each takes.
- Law The legal instruments the book relies on, with the date each was last checked. A book's legal position ages; this page says how far.
How it is built
Five parts. Part One prices the checking. Part Two asks who owns the decision. Part Three asks whether they can actually carry it. Part Four is the seam: six pairs of decisions, each correct on its own, that cancel each other out where they meet. Part Five is the toolkit.
One case runs through the whole book: a machine builder, a buyer, forty-one bids, and the supplier who called to ask where he lost it. The people are composite figures. The mechanisms recur; the names and numbers do not.
No footnotes. Where a finding comes from someone else's work, the name is in the sentence and the full reference is on this site.
The book
| Paperback | 216 pages, 6 × 9 inches · $24.95 ISBN 979-8-1726-1955-7 |
| Kindle | $9.99 |
| Publisher | Bosbach Edition, published independently |
| Series | Organization under AI, Volume One |
Paperback on Amazon Kindle on Amazon
Both editions are sold through Amazon. The German edition, Wer prüft, zahlt, appears separately at organisation-unter-ki.de.
The short version
Free excerpt
An excerpt from the book, eight pages: the case, the mechanism, and where to start. The full book with every source and the worksheets is on Amazon. No sign-up.
PDF 8 pages Excerpt
Read the short version (PDF) 702 kB
For introducing AI in your own organization, there is also the white paper What AI Tests in Your Rules: ten questions, each with a measure.
Talks and workshops
A talk on the book for executive boards, associations and supervisory boards, and a version with a working session for leadership teams, where each executive classifies one AI use from their own area.
How this book was made
AI supplies material. Whoever answers for it stays a person. That is what this book argues, and it is the arrangement the book was made under.
The German edition was drafted with AI and edited by the author. This English edition was written directly in English by an AI system, working from the author's German manuscript, his research, and his decisions on every question of substance. The author reviewed and revised the result and is responsible for it. Amazon's Kindle Direct Publishing has been notified accordingly.
That statement is in the book itself, not only in a form. A book that argues about who answers for machine-produced work has to say how it was produced.