See the closed loop
Learn how a system acts, changes the evidence, and treats the consequences of its own decision as independent proof that it was right.
The Loop
When the numbers look best, the system may be most wrong.
How AI and Automated Systems Learn to Repeat Their Own Mistakes
How AI turns its own outputs into proof, and how to break the loop before the echo becomes the truth.
What the book reveals
A system makes a decision. The decision changes the world. The changed world produces new data, and the system mistakes that data for independent proof that the original decision was right. That is the loop. It can begin with a reasonable goal, a useful metric, or a tool designed to make work faster and fairer. It becomes dangerous when the system can no longer see the cases it excluded, when its own output becomes the evidence used to evaluate the next cycle, and when improving numbers give leaders less reason to look outside the dashboard.
The Loop follows this pattern across hiring, lending, content feeds, policing, customer service, and the newer systems that write, summarize, recommend, and act on our behalf. Rad Stephens shows how a chatbot can learn from customer surrender as though it were resolution, how a hiring tool can validate itself using only the people it allowed through, how machine-generated language enters permanent records as if it were independent evidence, and how AI agents can remove the pause where human correction was supposed to occur. The book is not anti-AI. It is a practical warning about self-reference, speed, opacity, and the absence of independent correction. It also gives readers a working set of safeguards. Build evidence the system did not create. Name the person who owns the exception. Preserve provenance. Log decisions in a way that can be reconstructed. Install breaker switches before an automated action becomes irreversible. Treat near misses as evidence. Give human reviewers the time, authority, training, and institutional permission required to disagree. Concise, operational, and grounded in real organizational experience, The Loop helps leaders and teams ask the question that can expose a dangerous system before an outside correction does: where would the truth come from if the system were wrong?
Recognize the pattern
Follow the same failure through hiring, lending, content feeds, policing, customer service, generative AI, benchmarks, copilots, and autonomous agents.
Build the correction
Use independent evidence, exception ownership, provenance, evidence logs, breaker switches, and near-miss reporting before confidence outruns truth.
What readers will understand
See the system more clearly.
How automated systems turn the consequences of their own decisions into apparent proof that those decisions were correct.
Why hiring, lending, feeds, policing, customer service, generative AI, benchmarks, and autonomous agents can share the same hidden failure pattern.
How independent evidence, exception ownership, provenance, breaker switches, and near-miss reporting can interrupt the loop before harm compounds.
Who this book is for
For business and technology leaders, AI and data teams, operations managers, risk and compliance professionals, human-resources leaders, policymakers, consultants, auditors, educators, journalists, and anyone responsible for approving, deploying, reviewing, or living with an automated system.
Why this book matters
The most dangerous automated system may not look broken. Its internal metrics may improve precisely because it is optimizing against evidence it helped create. The book gives readers a practical way to distinguish genuine learning from self-confirmation before customers, workers, regulators, or outside events become the first reliable correction.
Why Rad Stephens can tell this story
Rad Stephens has spent more than three decades inside operational and technology systems, from warehouse floors and quality programs to large-scale planning and executive decision environments. His focus is not AI in the abstract. It is what happens when system output becomes operational reality and the organization stops asking where an independent correction would come from.
When the numbers look best, the system may be most wrong.
The Loop is preparing for release. Explore the book now, join the release list, or compare all three titles before deciding where to begin.