Blast radius
Read the centre, the edges, and the system around both before declaring victory.
The book
Who should read it
It is written for leaders, transformation practitioners, HR professionals, consultants and managers who have lived through programmes that looked convincing in presentation and failed in practice.
Organisations trying to integrate AI, and whose CEOs demand the ever-elusive “use cases”, will find the methodology particularly handy.
10+
Real-life cases
2026
Forthcoming release
AI+
Socio-technical change
When the same transformation keeps returning
Why Now
The author was sitting in his office a few years ago, speaking to a consultant and recruiter advising the company on a new transformation. The narrative felt oddly familiar: a move from a centralised marketing model to a local-first, decentralised model.
The words rang familiar because the same company had done the same transformation once in the last decade, reversed it five years later, and was about to reverse it again.
An enormous waste of resources, talent and time. The conclusion was blunt: organisational transformations often fail for basic human reasons. Adaptation is incomplete, theatre is abundant, and fear is usually operating quietly underneath.
As a society, we have failed at failing. So have corporations.
AI Stress Test
AI transformations are the true stress tests
AI integration is used as the flagship running example throughout the book because it exposes the worst outcomes of an operating model not designed right.
If a workflow with humans can go wrong, with a hybrid AI plus human team, that wrong gets multiplied a thousand times.
Controlled Detonations™
The Method: A Sandboxed Experiment with a Detonator.
A pilot is often protected from the conditions that will decide whether it works.
A Controlled Detonation™ is a sandboxed experiment. It goes through full socio-technical design, measures KPIs, and is continuously tweaked as the model learns.
One of the most important KPIs it measures is people’s engagement with the change and the quality and quantity of adoption.
It has stakes. It has a hypothesis. It has a blast radius. It has evidence standards. And it has the discipline to let the result change the design.
Fragments from the doctrine
What the method forces into view
Telemetry
Evidence must be designed before enthusiasm starts rearranging the truth.
Kill criteria
Hope should enter after evidence improves, not before governance is written.
Learning yield
The question is whether the organisation learned enough to justify the cost.
Socio-technical wiring
Tools, workflows, judgement, incentives and trust have to be wired together.
Scale what earns it
A local success deserves wider life only after the operating model has been re-read.
Field notes
Essays before the book lands
The Controlled Detonations™ Substack carries the argument in public before publication: AI, operating models, failed pilots, adoption, evidence and the quiet ways organisations avoid learning what they already suspect.
For people who have sat in the room
Watched the slide deck land. Watched everyone nod. Then watched nothing change afterwards.
First edition list
Be notified when the book is ready
Join the first edition list to be notified when the book is published.