The liability nobody priced

When a firm deploys a decision it cannot account for — an approval it cannot explain, a denial it cannot reconstruct, a determination it could not defend if pressed — it tends to believe it has done something clean. It has automated a judgment, removed a cost, moved faster. What it has actually done is quieter and heavier than that. It has taken on a liability, and then declined to write it down. The decision may be right; it may be wrong. What is certain is that if it is wrong, someone will eventually present the cost of its being wrong, and the firm will have no reserve set aside against that day, because it never admitted the day could come. The saving was booked immediately. The obligation was left off the books.

This is not a metaphor borrowed from finance to make a moral point sound rigorous. It is the literal structure of a contingent liability, and it behaves exactly as one. A firm that guarantees a subsidiary's debt, or faces a lawsuit of uncertain outcome, or has promised pensions it has not fully funded, carries an obligation that has not yet come due and may never come due in full — but is real now, is measurable in principle, and does not disappear because it went unrecorded. Accounting has spent a century learning to force these obligations onto the page precisely because the instinct to leave them off is so strong and so expensive. An unaccountable decision is the same kind of object. The absence of a record does not mean the absence of a debt. It means the absence of an entry.

The debt off the books

Consider what a firm holds after it makes a consequential decision it cannot examine. It holds the upside — the speed, the throughput, the labor it did not spend on deliberation. It also holds, though it rarely names it, the full exposure to that decision turning out to have been wrong: the reversal owed to the person it wrongly denied, the penalty owed to the regulator whose rule it unknowingly broke, the damages owed to the class whose members were all treated the same wrong way. Those exposures exist from the instant the decision is made. They are contingent — they crystallize only on a triggering event — but contingency is not nonexistence. An earthquake clause in an insurance policy is a liability before the earthquake. The firm that cannot account for its decisions is carrying the earthquake clause and has decided not to mention the fault line.

The failure here is precise, and it is worth naming exactly. It is not a failure to owe. The firm owes whatever it owes regardless of what it wrote down; liability tracks the world, not the ledger. The failure is a failure to measure. By declining to keep a record of what it decided, on what evidence, under which rules, the firm gives up the one thing that would let it know the size of what it is carrying. It cannot count the decisions that would fail examination because it cannot examine any of them. It cannot separate the exposure it should fear from the exposure it can dismiss. It has a liability of unknown magnitude and no instrument for reading the dial — which is a worse position than a large known liability, because a large known liability can at least be managed.

You cannot reserve against what you cannot see

The whole discipline of provisioning rests on a single capability: the ability to bound an exposure. A firm that can estimate its liability — even roughly, even as a range — can set aside a reserve against it, can price it into what it charges, can decide rationally whether to carry it or insure it away. A bounded liability is a line item. You can look at it, argue about it, fund it, and sleep. The thing that makes a liability governable is not that it is small. It is that it is known well enough to be sized. An unbounded liability admits none of this. It cannot be reserved against, because there is no figure to reserve. It can only be feared, or — more commonly — ignored, which is fear with the lights off.

Skipping the record does not retire the liability; it only guarantees you will meet it for the first time as a number you cannot dispute.

This is where the record does its economic work, and it is a different kind of work than the one usually claimed for it. We tend to defend accountability on the grounds of fairness to the person the decision lands on — and that defense is right. But there is a colder argument that ought to move even a firm that cares only about its own exposure. A body of examinable decisions is the instrument that bounds the liability. It lets the firm prove which decisions were sound and confine its exposure to the ones that were not. Without it, every decision is equally suspect, because none can be distinguished from any other; the good and the bad sit in one undifferentiated mass, and the firm defending itself has no way to say this one was correct that the correctness of the decision, rather than the fluency of the defense, actually backs. A Decision Receipt — a record carrying the evidence actually consulted and the rules actually in force, built to be contested rather than merely believed — is what converts the mass into a portfolio. Some entries hold. Some do not. The firm that can tell them apart pays only for the ones that do not. The firm that cannot pays for all of them, because it cannot prove otherwise.

The bill arrives named

Contingent liabilities have a habit the unrecorded ones share: they do not accrue on a schedule you set. They crystallize on an event you do not choose — a regulatory finding, a filed class action, a wave of reversals, a single public failure vivid enough to make everyone look at the ten thousand decisions behind it. On that day the liability stops being contingent and becomes a bill, and the two firms that made the same decisions meet the bill in two completely different postures. The firm that kept examinable records arrives able to apportion. It can show which decisions followed the rules that bound them, settle only the ones that did not, and defend the rest with something a tribunal will actually weigh. Its exposure was large but bounded, and it pays the bounded part.

The firm that kept nothing arrives with no such instrument. It faces the whole undifferentiated tail — every decision presumed suspect, because it made every decision unexaminable and cannot now un-make that choice retroactively. It cannot prove the good decisions were good, so it is made to answer for them alongside the bad. It pays for its errors, and then it pays again for its blindness, and the second charge is often the larger of the two — the settlement that is really a settlement about the missing evidence, the penalty inflated by the inability to show good faith, the reversal granted wholesale because the firm cannot contest it one case at a time. This is what it costs to discover the size of a liability only when it is handed to you as a total. You lose the ability to argue about the number. You are simply told it.

So this is why accountability belongs on the balance sheet and not in the compliance binder, why it is a matter of what the firm is worth and not merely of what the firm is permitted. The choice to deploy decisions one cannot account for is not the elimination of a risk. It is the assumption of an unpriced one — an exposure that is real the moment it is taken, deferred until an event outside the firm's control names it, and unbounded for exactly as long as there is no record to bound it. The record is not a nicety laid over a decision that would have been fine without it. It is the instrument that turns an unmeasured, unreservable, indefensible liability into a measured, provisioned, insurable one. A firm can carry the first kind for years and feel unencumbered. It is unencumbered the way an unfunded pension is unencumbered — right up until the reckoning, which sends its bill to the party that declined to keep the account, and never to the party that saved by skipping it.

— Dispatches · Summit Cognitive