A heart-attack drug, and the gap between knowing and doing
In the 1970s, the evidence that a drug could save heart-attack patients was already sitting in the medical literature. It took roughly a decade for practice to catch up, and in that gap, people died who might not have. The knowledge existed. The discipline to act on it did not.
Most organizations adopting AI today are making a version of that same mistake, sometimes in reverse. They rush to adopt tools they have never actually checked, or they cling to tools that quietly stopped working. Both come from one error: letting the feeling of knowing stand in for the fact of knowing. This book is about closing that gap.
"Having an AI tool that produces good outputs is one thing. Having an organization that makes better decisions because of those outputs is a completely different thing. The first is easy to buy. The second has to be built."
Not About How Models Work. About How Decisions Get Better.
This is not a technical book, and you do not need to know how a model works. It borrows the discipline medicine spent decades building, evidence-based practice, and applies it to AI in business.
Exposure vs. Readiness
Almost every organization has the tools, the data, and a working demo. Very few can show their actual decisions improved. The book names the difference, and from inside the meeting, the two feel identical.
Medicine as the Guide
Medicine faced this exact problem first, at the highest possible stakes, and built a formal discipline to stop smart experts from fooling themselves. The rest of us get to borrow those lessons.
Tools You Can Use
Real examples, named failures, and honest limits. Where the evidence is thin, the author says so. Reproducible worksheets in the appendices can be copied straight into your own organization.
Five Parts, Following the Path an Organization Actually Walks
The Trap
Why organizations feel ready before they are, through dashboards, successful pilots, and the confident voice of the tools themselves.
The Playbook
How medicine learned to trust evidence over authority, what counts as evidence, and how data debt quietly becomes decision debt.
A System You Trust
Start with the decision, not the model. How much evidence you need, how to measure honestly, and how to govern without paralysis.
Drawing the Line
Where machines should decide, where they should only assist, and where the honest answer is not to use AI at all.
Making It Last
Turning each decision into learning for the next, building the culture that lets evidence win, and treating AI as maintained infrastructure rather than a launched project.
Three Questions for Any AI Initiative
By the end, you should be able to look at any AI project in your organization and answer these. An organization that can answer them clearly has closed the gap that the streptokinase doctors couldn't.
What decision is this actually changing?
Not the report it generates, but the underlying choice a person makes. A tool can work perfectly and still not improve a single decision.
What evidence would show it improved?
Measured against how you were doing the decision before. Without a baseline, you can't tell improvement from noise, or from a good story.
Under what conditions would you stop?
A failure line drawn before the launch is honest. One drawn after is just a negotiation. Someone must own the result and have the authority to switch it off.
Written for Decision-Makers
Executives, managers, and team leads navigating AI adoption: anyone who has sat in the meeting where a clean dashboard and a good demo made a rollout feel inevitable. You will leave with more questions you can ask in your next meeting than slogans to repeat.
- Real, verifiable case studies, with failures named
- No engineering background required
- Aligned with responsible, evidence-based practice
The Evidence-Based AI Toolkit
Nine reproducible tools you can copy and bring to your next meeting. They are the practical core of the BRUKD methodology.
- The AI Readiness Scorecard
- The Decision Map
- The Evidence Ladder Worksheet
- The Pilot Charter
- The AI Decision Register
- The Governance Canvas
- The Data Debt Audit
- The Failure Post-Mortem
- The Meeting Checklist
Daniel S. Demoz, M.Sc.
Founder, BRUKD Consultancy · Adjunct Professor & Researcher
Daniel teaches big-data analytics and AI integration as a college professor while advising organizations across industry, government, and education. His work bridges evidence-based practice in medicine and evidence-based AI in business: the discipline of asking better questions, using better evidence, and making more careful decisions.
Bring the Discipline to Your Organization
Get Evidence-Based AI on Amazon today, available as an eBook for instant reading or as a paperback for your desk. Not sure yet? Start with the free sample, then buy the format you prefer.