Writing Books with Agentic AI, and Why the Australian Medicare Agent Breach Should Worry Every Executive

Expert Corner

For those of you following along at home, I have been a professionally published author for more than 30 years. Over that time, I’ve helped with, contributed to, and ultimately written several books from scratch. My first novel, There Are Days You Just Want to Go Home, is long out of print. My second, Danny and the Corporate Ladder, is still floating around on Amazon for anyone looking for a good story about friendship and surviving corporate life. But in the last three years, my focus has shifted entirely to technical writing.

Three years ago, I started and finished Infrastructure: Designing AI Networks. Navigating that process—alongside my day-to-day life inside a large corporation, sparked a deep dive into the physical backbone of artificial intelligence. That research culminated in my recent release, Designing AI Data Centers, which hit shelves earlier this year.

Behind the scenes, my workflow has completely transformed. Over the past two years, I’ve integrated Agentic AI as my primary research tool. To put that efficiency into perspective: what used to take me 12 months to research and write now takes around six. That velocity is why my next book, The Four Debts: Or Why Your AI Project is Going to Fail, is already in layout with the publisher. That book lays out the structural and operational pitfalls that compound and ultimately derail enterprise AI initiatives.

I am also thrilled to share that my newest book was just optioned by my publisher, acting on the speculation that I can finish the manuscript by November (fingers crossed). Titled Authority in the Age of AI, it tackles the emerging concept of physical AI, what happens when AI models move off the server and into embodied systems with real-world agency. By the way, if anyone wants a signed physical copy of any of my books, just buy one and let me know; I will be happy to figure out how to get it signed for you.

The reason for this more personal front-end to the newsletter is, frankly, I don’t want to talk in depth about the reality of the news this week. The recent agentic situation in Australia made me genuinely nervous. On June 18, 2026, an autonomous OpenAI agent assigned to research public medicine spending encountered blocks on the Australian Medicare Statistics Reporting Service portal. Instead of stopping, the agent actively found workarounds to access both public and unreleased, non-public files. It was a direct, autonomous infiltration of Australia’s equivalent to Medicare, and it stands as the first known instance of a rogue agent independently hacking a government network. Watching an agent refuse to take “no” for an answer when faced with government security barriers is terrifying. It turns the theoretical risks I write about into immediate reality.

I will leave you with a quick reminder. My father always had a unique perspective on the word “expert.” He used to say an expert is interesting, but if you break the word down: an Ex is a person with a past, and a Spurt is just a drip under pressure. Keep that in mind the next time someone tells you they have it all figured out.

How are you all handling the balance between AI’s massive productivity boosts and these escalating real-world risks in your own work?