Exact enough to preserve role boundaries, independent signal, evidence obligations, and human accountability.
About re:compound
Built from two careers inside consequential work.
Every installation is delivered personally by the two founders and designed to run without them.
11organizations started
9installations completed
9 of 9led to the next installation
2currently in flight
I’ve spent twenty-five years on how serious work gets done. Four places taught me the pieces.
McKinsey taught me how the work is structured, and what an apprenticeship is. I was the associate whose chart nobody wanted to see, and later the manager asking an associate whether the exception could be reconciled with the story. That was how judgment got made: by contact with the material, under someone who had read a thousand pages like it.
Bridgewater taught me that a decision system can be written down. I ran a hundred-person department inside a set of explicit principles, and built the instrument every manager at the firm used to make and record management decisions. Judgment could be made legible enough to operate, if you were willing to write down what you actually believed and let it be tested.
Vega Factor and Talentism taught me what drives people, and how to train a thousand of them at a time without flattening them. I did diligence on leadership teams for private equity, and learned what a partner reads in a room.
Deloitte taught me what happens when AI arrives at a firm of a hundred and fifty thousand people. On a small team reporting to the CEO and the Board, I built the firm’s approach to generative AI: where the models get used, how their output is governed, what human capacity gets redeployed rather than cut. What I saw was an industry heading toward faster and cheaper at the layer below the judgment, and a much bigger opportunity above it: to lift what senior people are capable of, not just how quickly the work beneath them moves. So I left to build for that.
The condition I kept seeing had no name, so in July 2026 I gave it one: unreviewable fluency. The method came from asking what the old way of working had been doing for free: leaving marks on the page a reviewer could read, forming judgment through repetition, keeping the understanding alive in the people who did the work. Then designing a replacement for each, now that the work is cheap and the marks are gone. What was never designed can be replaced by something that is.
I spent more than a decade as a nurse: ICU, emergency, recovery, rural and remote. Australia, where I trained, then the UK, then the US. Every one of those units runs on one distinction: a patient goes back to the ward when they meet the discharge criteria, not when they look fine.
Coming back to a patient after a break was always harder than it looked. The chart told me what had happened. It didn’t tell me what was true right then. Rebuilding the picture was its own work, separate from the care I was about to give. I got fast at it: under five minutes if I knew what I was doing, twenty if I didn’t.
The hardest thing I learned in emergency wasn’t treating the sickest patient. It was deciding who didn’t need me right now. After enough shifts you walk into a waiting room and you’ve already placed them: one can wait, one will deteriorate in the next hour, one needs a bed before they finish telling you what’s wrong. The instinct to be useful everywhere is the thing you have to put down.
When Bud started describing what he was seeing in professional firms, I recognized all of it. Work that looks fine. Senior people rebuilding the picture from scratch every time they come back to a matter. Attention going where the noise is loudest instead of where it’s needed. And a bar I had held for years: a junior is signed off only when they can run it without supervision. Week ten is not a new standard to me.
My side of the firm is operations and practice. Every installation is scoped through my desk and held to that bar.
I didn’t expect to end up here. AI gave me capability in a field I had no training in, and a voice in rooms where I hadn’t known how to speak. What I found is that this is the work I’ve always done: how you make good decisions when things are moving fast, how you keep hold of what matters. Just at a different desk, and for more people than one unit could hold.
The combined practice
A method is not installed until people can run it under pressure.
We run the discipline on our own work. Each of us is alternately the reviewer and the producer of the other’s.
The work requires both a precise architecture and a disciplined transfer into human behavior. The founders deliver every installation together and design it to run without them.
Light enough to fit live professional work, real deadlines, limited attention, and existing AI tools.
Installed to independent team-level operation rather than continuing advisory dependence.
Working principles
What the practice will not trade away.
The external object governs.
The point is not compliance with a framework. The decision, evidence, profession, and live work remain in charge.
Human accountability stays legible.
Models can expand, structure, test, and compress. Humans own what crosses, what is relied upon, and what happens next.
Useful difference is preserved.
The human and model, and the producer and reviewer, should not collapse into one agreeable signal before the work has earned it.
The capability leaves with the client.
Success is independent use by the working pair, not an accumulating dependency on re:compound.
A practical first step
Thirty minutes about the next piece of work you’ll have to sign.
Who’s bringing it, when, and whether that’s the team to start with. No deck, no demo, nothing to prepare.