Hand someone the policy library and ask them to run operations from it for a week. They will come back and tell you they cannot.
The policies may be clear and the obligations may be correct. The problem is that policies are written to say what must be true. They are rarely written to say what evidence would prove it is true. Those are different things, and in most institutions only the first one exists.
People carry the missing layer
So the translation happens in someone's head. A person who has read enough of the policy and seen enough of the systems looks at a case and decides whether what happened was allowed. They know which record matters and which field to trust. They know which exception is real and which one can be ignored. They know that this policy means checking one system against another before anything can move.
That judgment is real work. It is also almost completely invisible, because it sits inside salary expense, tribal knowledge and the accumulated experience of the people who keep the operation running. The institution has the rule. What it does not have is a machine-readable test for whether the rule held.
The missing artifact is the translation
Take a simple requirement: "Settlement must occur within two business days." That sounds operational until you try to make a system check it. Now you need to answer:
- Which event starts the clock?
- Which system records it, and which timestamp matters?
- What counts as settlement, and which system proves it happened?
- Which calendar defines a business day?
- What exceptions apply?
- What happens when one of those records is missing?
The policy did not answer those questions because it was never written to. A person answered them every time the work came up.
That is the gap. The rules are not missing. The translation from rule to evidence is missing, and in most institutions nobody owns the job of writing it down.
Cordant is built to hold that translation
Cordant turns an institution's policies, agreements and operating rules into checks that run against the systems themselves. Each rule stays tied to the document it came from, and each answer carries the evidence used to reach it. If a rule has to be inferred rather than read directly from the source, Cordant says so, and the institution approves it before it runs.
The policy stays exactly as strict as it was written. What changes is that the institution defines once what evidence proves the rule held, and a system checks it continuously instead of a person interpreting the same obligation again and again. People still own the judgment. They stop having to rebuild the test every time.
Write one rule down this week
Pick one rule from one policy. Then write one line that explains how a system would know the rule was satisfied:
- Which record?
- Which field?
- Compared with what?
- Under which condition?
If that line is hard to write, that is the point. The institution may have documented the obligation, but it has not documented how to prove it. Until that translation exists, a surprising amount of compliance is still running on memory.
