What the work taught me, and what it keeps asking.

Cadence should be installed early, but early is exactly when you have no credibility to install it.

I believe reporting and cadence are operating infrastructure rather than administration, and that waiting makes the eventual intervention more expensive. That belief is in direct tension with how influence actually works when nobody reports to you.

On day one nobody wants your operating model. By the time they want it, the problems you would have prevented have already happened. What has worked for me is solving one visible thing first and spending the credit immediately. The open question is whether that tension can ever be removed, or only sequenced around. I lean toward the latter, which makes timing the sequence a discipline in its own right.

From · AMC Networks

What is the honest price of the status quo?

Every proposed change gets measured against its implementation cost. Almost nothing gets measured against the cost of continuing as-is, because that cost is spread thinly across everyone's week and appears on no line item.

I moved senior leadership by making that comparison explicit. The open question is whether the cost of the status quo can ever be measured with the same rigor as an implementation budget, or whether making it visible at all is the entire available win.

From · AMC Networks

When does adding another quality gate stop paying for itself?

Every automated gate in my system exists because something failed. That is a defensible way to grow a quality layer. It is also a process with no natural stopping point.

Gates cost time on every run forever, in exchange for preventing an error that happened once. Nobody has a clean way to price that trade. The harder discipline, and the more interesting one, is not adding the next gate but knowing when to retire one.

From · Executive Office

Integrate the layer, not the company.

Fourteen acquisitions in a consolidating portfolio, and the framing everyone reached for was a speed question. Absorb quickly and you destroy what you paid for. Leave them alone and you own a holding company with a shared logo.

I decided that was the wrong axis. The question was not how much to integrate but which layer. Names and products stayed, because that was what customers valued and what the acquisition actually bought. Systems, metrics, reporting, planning cycle and operating procedures became common underneath.

What it cost: the integration never finished, and could not be reported as finished. That is uncomfortable in a company that likes closing things. What it bought was a portfolio that could absorb the next acquisition without anyone designing a new process for it.

Kaseya · 2018–2020

Treat every fabrication as a bug.

The comfortable framing is that language models hallucinate, that this is inherent, and that the mitigation is a careful human. That framing locates the problem in the technology rather than in the system built around it, which conveniently means nothing ever has to be fixed.

I decided to treat each one as a specific defect with a cause, a reproduction case, and an encodable fix. That single reframing forced everything downstream: documented failures, a testing layer, approval gates, and separating the thing that produces work from the thing that judges it.

What it cost: a permanent tax on every run, and a quality layer that only grows. That is the trade the open question above is trying to price.

Executive Office · Current