Operator debt is the backlog of decisions an operator has made in their own head and not yet written down in a reusable form. Every company has some. Most have more than anyone tracks.
Example. A VP decides not to hire a certain role this quarter. They don’t write down the reason. Six months later three different people separately raise the same hiring idea. Each time the VP works through the same tradeoff and explains the same call in person. The decision was made once and the deliberation is paid three times. Multiply that across the dozens of decisions a senior operator is carrying at any given moment and you have operator debt.
It doesn’t appear as a line item. It shows up as the feeling that the business is slow, that the same questions keep coming back, that senior time gets eaten by things that should’ve settled already, that meetings keep ending with the question still alive. The interest on operator debt is paid in meetings and Slack threads. It’s also paid in the hour a senior person meant to spend on the next thing. Most senior people are paying it constantly and not calling it what it is.
The debt gets retired by writing decisions down in a reusable form. Not a diary, because a diary doesn’t get retrieved. A decision log: what was decided, what fed the call, what conditions would force a revisit, and who owned the call. You still pay the deliberation once. You stop paying it every time the question returns.
Operator debt is also one reason AI bogs down in senior roles. A model can’t retrieve what was never captured. An executive who wants a model to help think through a decision first has to pull the context out of their head and onto a page, which is work they’ve often been avoiding for years and would rather keep avoiding. AI makes the debt visible and makes the cost harder to dodge, rather than creating it.
Better to retire the debt on purpose than to have a model pressure it out of you sideways. The operator who has a model that helps them think has usually done the boring work of writing things down first. The operator who says the models don’t help often hasn’t.