Most finance functions are good at documenting some things and not others. The pattern is consistent enough across organisations that it is worth naming.

Compliance procedures, audit trails, approval matrices, schedules and reconciliations all get documented thoroughly. The reasoning behind why they were designed the way they were, the judgment calls that produced the rules, the exceptions everyone knows but nobody captured: these often do not.

Anyone who has been tasked to write minutes will recognise the rule. Focus on actions, decisions, and outcomes, not a transcript of what was said. This is the structural reality of working in any organisation. The explicit record carries weight. Anything written down can be read by someone who was not in the room when the decision was made. An auditor reviewing a procedure five years from now does not see the trade-offs that were weighed at the time. They see only what was written.

This is not unreasonable. It is a sensible response to the fact that documentation, in a regulated environment, is also evidence. The team writing the document knows it might be read in contexts the writers cannot foresee. So the document gets cleaner than the thinking behind it.

This is why an SOP typically contains the procedure but not the reasoning behind it. The reasoning is the kind of knowledge that is hard to capture in writing because it lives in experience rather than rules. And the loss compounds when people leave. New joiners learn the procedure but not the why. The procedure gets followed long after the original context has changed, because nobody remembers what the context was. The next time a similar judgment is needed, the team starts again from first principles.

This is the part of finance work that AI cannot yet help with. Models can search what was written. They cannot search what was deliberately left unwritten.

— YS Lim