Every growing company hits the same wall. In the early days, the hard calls run through one or two people whose judgment is good - founders, usually - and the quality of that judgment is a large part of why the company is working at all. Then the company grows, the volume of decisions outruns those few people, and the judgment has to be distributed to people who do not have it yet. This is where most of the original quality leaks out.
The leak has a standard cause. Faced with the need to scale judgment, organizations reach for the nearest available mechanism: they write rules. The founder’s intuition gets compressed into a policy, a checklist, a decision tree. And something essential is lost in the compression, because the rule captures what the founder usually decided without capturing why - and the why is the part that knew when the usual decision was wrong.
Why rules dilute judgment
Judgment is not a lookup table. When a founder makes a good call, they are not retrieving a stored answer; they are weighing a situation against a deep, mostly tacit model of what matters, what the tradeoffs really are, and which considerations dominate in this particular case. The output looks like a decision. The asset is the model underneath it.
A rule captures the outputs and discards the model. It says “in situation X, do Y,” which is correct exactly as long as the situations stay inside the range the rule-writer imagined. The moment a genuinely novel case arrives - and in any interesting business, novel cases are most of the consequential ones - the rule either gives the wrong answer confidently or gives no answer at all. The person applying it has no model to fall back on, because the model was never transferred. Only its shadow was.
A rule tells you what to do. Judgment tells you what matters. You can scale the first easily and it will fail you exactly when it counts. The second is hard to scale and is the only thing worth scaling.
This is the dilution. Each time judgment is compressed into a rule for the sake of transfer, the tacit model is stripped out and what remains is a brittle approximation. Stack enough of these approximations and you get an organization that follows its own playbook faithfully off a cliff, because the playbook encoded the founder’s answers without the founder’s reasons.
Transferring the model, not the answers
The alternative is harder and it is the actual work of leadership at scale: transferring the model rather than the outputs. This means making the reasoning legible, not just the conclusion. When a hard call is made, the valuable artifact is not “we decided X” - it is “here is what we were weighing, here is which considerations we let dominate and why, here is the case where we would have decided differently.” That is the model, externalized, and it is what a person can actually learn from.
Done consistently, this builds something more useful than a rulebook: a body of worked examples with their reasoning exposed, from which people absorb the model the way an apprentice absorbs a craft - not by memorizing rules but by seeing enough reasoned cases that they internalize the weighting. The goal is people who decide the way the founder would, including in situations the founder never saw, because they carry the model rather than the rules.
This is the heart of what I think about as leadership at scale: the leader’s job stops being to make the decisions and becomes to make the reasoning behind the decisions transferable. The founder who hoards the judgment becomes a bottleneck. The founder who flattens it into rules dilutes it. The founder who externalizes the model multiplies it.
The discipline of preserving nuance
There is a real tension here, and pretending it away helps no one. Reasoning that is fully nuanced is also slow and hard to apply, and at some point an organization needs decisions made quickly by people who cannot reconstruct the founder’s entire model every time. Some compression is unavoidable.
The discipline is to compress consciously and to keep the uncompressed reasoning available behind it. A heuristic is fine - useful, even - as long as everyone knows it is a heuristic, knows the model it approximates, and knows to escalate to the full reasoning when the heuristic feels wrong. The failure is not having shortcuts. The failure is forgetting they are shortcuts, treating the approximation as the truth, and losing the path back to the model that would have caught the exception.
The closing thought
Scaling judgment is not a matter of writing better rules. It is a matter of making the underlying model - the sense of what matters and why - legible and transferable, so that more people can apply it without it thinning out into a checklist that misses the point exactly when the point is hardest to see.
The measure of a leader who has done this well is not that the organization follows their rules. It is that the organization makes decisions they would be proud of in situations they never anticipated, because what was transferred was not the answer but the way of seeing that produces good answers. That is judgment scaled rather than diluted, and it is one of the few things in an organization that genuinely compounds.