Rarely anything exotic. Who changed this price, when did this close date move, who approved that discount, and who could see this record.
Each is a five-minute question with a trail and an afternoon of archaeology without one. The cost is not the audit; it is that the answers are reconstructed rather than read.
Every write records the actor, the timestamp, the field, and the value before it changed — including writes made by integrations and automations, attributed to the token and the user behind it.
The trail cannot be edited by anyone, administrators included. That single property is what makes it evidence rather than a log.
It does not prevent misuse by somebody with legitimate access. It makes it visible afterwards, which is a different and necessary thing.
And it does not interpret your obligations. It records thoroughly; what you must retain and for how long is a decision for you and your advisers.
Field-level permissions and the trail are read together: permissions stop access by people who should not have it, and the trail covers everyone who legitimately does.
So the question "who saw this" has an answer, which is usually the one nobody can produce from a CRM.
No. Nobody can, which is what makes it usable as evidence.
Yes, attributed to the token and the user behind it.
Yes — who looked is often the question.
We will run it end to end on your own numbers in half an hour, and tell you honestly which parts Treepie does not improve.