AORIE

A third-year associate is drafting and needs the firm's position on a clause type it has taken a view on before. That view exists—in closed matters, in the precedent bank, in a memo from 2023, and in the head of a partner who is currently in court.

They search the document store, open the results that look closest and reuse language from a matter they were not on and cannot fully assess. Or they ask whoever is nearest and free. What it costs is rarely dramatic and always expensive: advice that drifts between teams, work redone at partner rates and no way to reconstruct what the associate relied on when they drafted it.

Every answer is recorded with its sources. If advice is questioned later, what the drafter was told—and what it was based on—is still there.

A first-line colleague takes a customer question close to a boundary: what a product covers, whether a fee applies, what can and cannot be said. The answer lives across product terms, a policy library and last quarter's briefing note.

They ask the person next to them. That person is confident and usually right. Two desks away, someone answers the same question differently, also confidently. This inconsistency is invisible until it becomes the subject of a review. When the question comes—what were staff told, and when?—the answer is usually reconstructed from training decks and memory.

The record is not a byproduct here. It is the thing that makes the answer defensible.

A treatment question comes up mid-engagement. The answer depends on the standard, the firm's methodology and how the firm handled something comparable two years ago.

The junior asks the senior. The senior asks the manager. The manager asks the technical partner, who has answered this exact question dozens of times and will answer it again next month. The reasoning reaches the file as a conclusion, not as a chain. Then file review asks what the basis was for this judgement—and that basis has to be rebuilt from memory and email.

The reasoning reaches the file with the conclusion, ready for review.

An employee asks about family leave, a redundancy consultation step or how a grievance proceeds. It is a reasonable question with a definite answer.

The answer sits in a policy updated eleven months ago, an intranet page that was not updated with it and a manager's recollection of how it went last time. HR answers in good faith. The employee acts on the answer. The serious part is the variance: two employees in similar circumstances receive different answers, and later nobody can establish what either of them was told.

The answer trail preserves what was asked, what was answered and which policy version supported it.

A claims handler decides coverage. The policy has a base wording, endorsements and a version history—and the version that matters is the one in force on the date of loss, not the one currently on the intranet.

Handlers develop instincts and those instincts are mostly right. But wordings change, endorsements accumulate and the handler is working at pace. When a complaint arrives later, the question is which wording was applied and what the handler was working from at the time. That is often unanswerable.

When a complaint arrives, the file already contains what the handler was told.

An operator or QA reviewer needs the current procedure. Procedures are controlled, versioned and changed under a formal process—which is exactly why the wrong version is so easy to be holding.

In practice, people work from what is to hand: a printed copy, a local file or an experienced colleague's account. Usually it matches the controlled version. Occasionally it does not. When a deviation is investigated, the question is what procedure was in force and what the person was actually working to. The gap between those two is where the investigation lives.

The answer and its source record are ready when the procedure is later examined.