K1 and K11
Compliance and Controls
Every decision has an outcome, a reason and a name against it. The pipeline is policy- driven: a versioned graph of controls bound to the points in a lifecycle where they run. Each returns a recorded outcome, and the engine keeps what it was evaluated against so the decision can be reconstructed later.
Seven registered controls.
Named as the registry names them, because these are the strings that will appear in your evaluation records and in the answer you give a supervisor two years from now.
A policy is a versioned graph bound to the points in a lifecycle where it runs, so adding a control to onboarding is a published policy version rather than a release.
kyb.structure, assigned to the compliance queue. One case per subject per control, so a second evaluation adds a sighting rather than a duplicate, and it lapses on its own when the control stops objecting.Six outcomes, and the two that matter most are the awkward ones.
And how it got the answer.
An examiner asking whether a customer was screened is asking two questions. The engine records both.
Called now, against a hard deadline.
A prior decision satisfied it, within its TTL and reached against the same reference data. Change the list and the cache is no longer valid.
Not called, because a case had already cleared this exact subject against this exact data. Recorded as its own mode, because ‘did you screen them’ and ‘did somebody once decide you need not’ are different questions with different answers.
Ownership, because the subject is rarely the person.
A control runs against a subject, and for a company the subjects that matter are the people behind it. The graph is traversed and the arithmetic composed before screening runs, so the thing being screened is the real beneficiary rather than a holding company.
Twenty-five per cent is the threshold here, and it is configuration, not code. Cross it and the person becomes a party in their own right, with their own screening and their own record.
Where it sits in onboarding.
Controls are not a stage. They run at the points the policy binds them to, and the record each one writes is what an acceptance is later defended with.
- 1Application submittedX2, client portal
- 2Identity verifiedK5, idv.verify
- 3Ownership resolvedC3, through every layer
- 4Screening clearedK2, sanctions.screen
- 5Accepted by officerX3, named decision
- 6Account openedC2, wallet per currency
Bring your own policy and we will run it.
Your thresholds, your jurisdictions, your appetite. The interesting part of the demo is the case that opens, not the five that pass.