What is Let the AI Be Smart about?

Let the AI Be Smart by Jason Santiago is a book about getting real, trustworthy work out of AI. Its central discipline is the Council Method: running multiple AI models independently on the same question, forcing every claim to prove its source, and keeping the final decision human. Around that core it teaches the whole working loop, from briefing AI the way you would brief a brilliant stranger to knowing which decisions should never be delegated at all.

By Jason Santiago, author of Let the AI Be Smart. As of 2026-07-05.

The arc of the book

The book opens with the problem it exists to solve: the gap between seeing and shipping. Most people who never built the thing they envisioned did not lack the vision. They lost it in translation friction, the work of turning a whole picture into executable steps. AI is the first tool that removes that friction, if you can trust what it gives you back.

The first third teaches the input side: prompt engineering as plain clear communication, real constraints instead of invented ones ("boundaries, not blinders"), and the four-word question the author calls the most powerful sentence he has ever typed into an AI: "What did I miss."

The middle teaches verification: why one model is one witness, how to design a Council with roles instead of personas, why independence comes before conversation, why provenance comes before consensus, and how to synthesize a verdict you would sign your name to.

The last third teaches the jobsite controls: RFIs and submittals ported from construction to AI work, the superintendent role that survives an AI's memory loss, building for one person by name, and the hard rule for when not to delegate: if being wrong cannot be undone, it stays human.

The tools

Nine appendices turn the method into copy-paste tools, from the Council Protocol to the When-Not-to-Delegate Checklist. All nine are free on this site.

The book is 166 pages, hardcover and Kindle, published July 2026.