A lender cannot defend a black box, a loose dashboard, or a privacy promise that only exists in a policy. Kenshiki is designed to turn each governed credit question into evidence a lender can inspect, explain, and replay.
Governance flow
Prepare. Propose. Prove.
01
Prepare
Bind the lender’s question to an approved proof rule, policy version, and provider boundary before any check runs.
Boundary
Allowed question
Proof
Defined in advance
02
Propose
Ask for a narrow bounded result from the approved source environment instead of collecting a raw private-data feed.
Boundary
Provider-side check
Output
Bounded answer
03
Prove
Seal the bounded result, timestamp, policy version, and explanation into a reviewable record for underwriting, compliance, and examiner review.
Record
Inspectable
Use
Replayable later
The model, provider, and lender policy do not collapse into one black box. Each boundary is visible enough to govern.
Governance posture
Governance has to happen at the decision, not after it.
The record has to be produced at decision time, not reconstructed later. For credit, that means the proof boundary, the allowed question, the check result, and the lender-facing explanation all have to be fixed while the decision is being made — written to the same replayable record the platform uses for every governed decision.
Policy theater
Write a standard and hope the system followed it.
Static policies are necessary, but they do not prove what happened in a specific credit decision.
Kenshiki
Bind the proof rule before the check runs.
The question, allowed evidence type, evaluation rule, and Boundary Gate decision are part of the decision record itself.
Dashboard governance
Monitor after the answer already left.
Post-hoc review can find problems, but it cannot make an unsupported answer defensible retroactively.
Kenshiki
Gate what can become decision-grade evidence.
If a check cannot be tied to approved proof, the Boundary Gate, applied to credit evidence, keeps it from masquerading as verified support for a credit decision.
Each governed decision shows what was checked, what the bounded result was, which proof rule applied, and how the claim ledger preserves the replayable record.
Data boundaries
Designed to keep private source data controlled.
Kenshiki brings a narrow question to approved data environments and returns bounded proof, not a new raw-data warehouse for lenders to govern.
Predefined proof
Defined in advance.
What counts as valid proof must be written down, dated, and version-controlled before a decision uses it. No moving the goalposts after the fact.
Decision record
What a governed credit record should preserve
The record does not need to expose private source data to be useful. It needs to preserve enough structure for a lender, reviewer, or examiner to understand what was asked, what boundary applied, and why the output was appropriate to use.
Decision record · sample structure
What every governed credit decision preserves.
Question boundary
The narrow continuity, context, or identity question Kenshiki was allowed to answer.
Evidence boundary
The approved class of proof and provider boundary, without disclosing raw private data.
Replay path
The returned proof, policy version, timestamp, and explanation a lender can review later.
Assurance posture
Show the status honestly.
No badge we have not earned. No claim the architecture does not support.
Kenshiki operates against five governance commitments. Each is something the platform delivers at runtime, not a future plan.
Decision records
Every governed credit decision writes an inspectable, bounded-result record at decision time.
Provider boundaries
Providers compute on their side and return bounded results. Kenshiki does not ingest raw private data.
Fair lending
Signals are policy-bound, monitored for impact, and explainable. A signal that cannot be governed does not become decision-grade evidence.
SOC 2
Controls operate against the SOC 2 Trust Services Criteria. We display the report once a Type II opinion is issued.
NIST AI RMF
Governance is structured against the NIST AI Risk Management Framework — Map, Measure, Manage, Govern.
Programs like TRUSTe/TrustArc, VeraSafe, and PrivacyTrust are commonly held by B2B data providers in our space. The processor-scoped program (BBB PRP / CPRA) is the one aligned to our role.
Fairness
Built to be examined, not to avoid examination
Using outside evidence in lending carries a real obligation: make sure it does not quietly disadvantage protected groups or become a hidden second score.
We treat that as runtime discipline, not a checkbox. Signals must be tested, controlled, limited by lender policy, and explained in plain language. A signal that cannot be governed should not become decision-grade evidence.
Kenshiki supplies evidence a lender can inspect. It does not replace the lender’s credit policy or final decision authority.
Not a raw-data sharing scheme.
The product direction is bounded proof, not broader exposure of sensitive source data.
Not a certification page.
This page describes design posture and governance commitments. Certifications, legal opinions, and data-processing terms require their own evidence and agreements.