Every organization I've worked in tracks depreciation on physical assets. Servers, furniture, vehicles — book value, useful life, replacement cost, all of it. The judgment sitting inside your exception-handling process — the understanding of when a rule doesn't apply, when the standard procedure is about to produce a bad outcome — has no line item anywhere.
Let me be precise about what I mean, because I wrote about a neighboring problem in Issue 40: the way automation removes the routine work through which newcomers develop judgment, and hollows out the training ladder. That's people losing the chance to learn. This is the organization spending down something it never counted.
The pattern is familiar from ops, and it's about to show up everywhere agents touch. Your team handles exceptions, the cases where the standard process doesn't fit. Eventually someone notices a recurring one and writes a rule for it. The exception gets automated. That's good engineering. It's also a withdrawal from a balance nobody tracks, because the judgment that produced the rule was built out of context — specific conditions, specific failure modes, a specific regulatory and competitive environment — and the rule carries the conclusion forward while shedding the reasoning. Conditions drift. The fit degrades quietly, since the people who knew why the rule existed have moved on and the rule keeps firing anyway.
Sørensen and Stuart found a version of this in their research on organizational innovation: older firms produced innovations at higher rates while those outputs grew increasingly obsolete relative to current technological demands. The routines survive better than their fit does, and nobody put a review on the calendar.
So what would it look like to treat judgment capacity the way you treat capital equipment? You'd ask what it cost to develop — not the salary, but the exception volume, the review cycles, the mistakes that built the pattern recognition. You'd ask how long the conditions that generated it are likely to hold, which is your useful life. And then: what would regenerating it cost if the rule stopped working and nobody remembered why it was written?
Nobody does this. The international standard for intangible assets (IAS 38) requires training expenditure to be expensed immediately, on the grounds that organizations lack sufficient control over the future economic benefits of a skilled workforce. The IASB is reviewing IAS 38 now, and nothing in that review creates a judgment-capacity category. Researchers have measured organizational forgetting, knowledge depreciating year over year, and estimated replacement costs for organization capital using accumulated spending as a proxy. But the narrow thing — whether a particular automation rule drew down future judgment, whether anyone can reconstruct the reasoning when conditions move — has no instrument at all.
The rule carries forward the conclusion but sheds the reasoning. As conditions shift, the rule's fit degrades silently — and nobody scheduled a review.
As agents absorb more exceptions and compile more rules, the withdrawals speed up. I don't think anyone needs a literal depreciation schedule for institutional judgment. But not tracking it at all — not even knowing the balance exists — is how you end up running on rules nobody can explain, tuned for a world that stopped existing, with no budget to rebuild what got spent. I've watched runbooks accumulate for years until somebody triggered one that referenced a service decommissioned three migrations back. It fired exactly as written, nobody remembered why it existed, and we found out when production caught fire.
That's what I think is coming for organizations that automate their exceptions without tracking what it cost to understand them in the first place.