U.S. adults who can't be scored; the system has no story to read. (Federal Reserve)
AI governance & credit-decision infrastructure
Tell a recovering borrower from a synthetic identity.
Kenshiki supplies banks, credit unions, and lenders with independent, inspectable evidence of continuity, context, and fit, added to the file rather than hidden in another score, so thin-file applicants get a fair read and fabricated identities get caught.
in estimated 2023 U.S. synthetic-identity fraud losses — credit files with no real life behind them; an industry estimate, not a measured loss. (Boston Fed)
student-loan borrowers whose scores dropped more than 100 points in one quarter of 2025 when delinquency reporting resumed. (New York Fed)
The problem
The file can look complete while the life behind it is missing.
A credit score is a snapshot. A person is a journey.
Scores, tradelines, eCBSV name-and-number checks, and uploaded documents each show a fragment. The underwriting question is whether those fragments add up to a real, continuous life that fits the loan in front of the lender.
The adversary has industrialized too: synthetic-identity rings now combine manufactured identities, residential network paths, real human operators, and AI-generated documents. Against that tier, no single signal is enough; what survives is cross-signal coherence and a replayable record of why the decision was reached. Read the working paper: The Apex Adversary.
Approvals rescued
Good borrowers denied because the file cannot show the life.
Thin-file and post-shock applicants can be profitable customers when continuity, context, and fit are visible enough to defend.
$49B in modeled annual interest income at stake from credit-invisible approvals traditional files miss.
Losses avoided
Synthetic identities that look clean enough to pass on the file.
A synthetic bust-out is principal committed to a fabricated life, not a small pricing mistake.
Measure as principal not lent to applications flagged for synthetic-identity risk.
The $49B figure is a Kenshiki estimate based on credit-invisible counts, credit application rejection rates, and alternative-data approval-lift research. Impact is portfolio-specific and should be measured on completed decisions, not pending recommendations. Kenshiki supplies evidence; the lender keeps final decision authority. The public sources behind these gaps are collected on The Evidence.
The decision moment
The file shows the shock. Kenshiki shows what happened after.
Step through three applicants whose bureau-visible records need context, and watch a decision-time view resolve into evidence a lender can open and defend.
Shock in the file
2023
Caregiving, income interruption, and a cluster of late payments enter the record.
Pattern restored
2024–2026
Obligations current. Payment rhythm restored. The bounded shock no longer describes the life.
Defensible record
Now
Continuity, context, and fit are packaged as evidence the lender can open and defend.
Composite caseRepresentative
A moment is not a life.
- Bounded shock
- Eight-month disruption, now closed.
- Recovery signal
- Two years of restored payment rhythm.
- Privacy line
- Continuity only — not merchant, purchase, or neighborhood.
A lender should be able to see the bounded shock, the resumed pattern, and the defensible reason to add context to the file.
Method
How Kenshiki works
Kenshiki adds evidence to the file by asking two narrow questions, then checking the source pattern behind the claim.
Is there a real life here?
Presence and continuity show whether the claimed person, place, and timeline hold together over time.
Does that life fit the loan?
Life context and fit show whether the requested loan matches the evidence a lender can inspect.
Kenshiki sits between lenders, credit bureaus, and approved source environments. It asks narrow questions, keeps raw data where it belongs, and returns evidence the lender can inspect. In mortgage and auto, where AI can forge the documents, Kenshiki tests the pattern at source.
Who buys it
Built for the people who carry the decision.
Every consequential credit decision lands on a person: the analyst who clears the alert, the underwriter who signs the approval, the reviewer who defends it later. Approve a fabricated life or decline a real one — both are losses, both have victims, and the same underwriter answers for both. Kenshiki exists to make sure they are holding evidence, not guesswork.
Credit risk teams
See when a thin file or recent shock hides a recoverable borrower.
Declining a recoverable borrower is not a safe default. In FinRegLab’s 2025 consumer-underwriting study, models combining bureau and cash-flow data increased approval rates by as much as 4% at mainstream risk thresholds without increasing default risk — roughly two million card accounts and 152,000 mortgages a year.
Source: FinRegLab, 2025.
Fraud and identity teams
Refer applications where the file looks plausible but the life pattern does not hold together.
Every dollar of fraud loss costs more than the dollar itself. LexisNexis found U.S. financial-services fraud costs averaged $5.75 per $1 of fraud loss, while 44% of institutions relied mostly or entirely on manual processes.
Source: LexisNexis True Cost of Fraud, 2025.
Compliance and model-risk teams
Open the evidence behind a recommendation without relying on another unexplained score.
When the examiner asks why, “the model said so” is not an answer. Kenshiki leaves a replayable record: what was asked, what returned, and what the reviewer decided before the decision had to be defended.
Related: SR 26-2 and model validation and CFPB adverse-action guidance.
What’s different
Most tools validate the file. Kenshiki tests the life behind it.
The difference is not a sharper prediction. It is a record a reviewer can open.
Most checks
Validate fragments of an application.
Scores, tradelines, eCBSV, and documents can each look right while the life behind them does not cohere.
Kenshiki
Corroborates the pattern at source.
It asks whether the person, place, obligation, and continuity hold together where the signal originates.
Most tools
Return a score you have to trust.
A number you can’t open is a number you can’t defend to an examiner.
Kenshiki
Returns evidence you can replay.
Every check is built to be opened, explained, and traced back to what was known before the decision.
Most data feeds
Add more raw data to the lender’s audit surface.
More copies can mean more permissions, retention questions, and downstream review burden.
Kenshiki
Returns bounded proof, not another data lake.
Raw data stays in approved source environments; the file gets the narrow evidence needed to defend the decision.
The decision-quality evidence behind this — reviewer variance, disclosure experiments, the auditability gap — is on our approach.
Trust
Built to be defended
Every check is designed to leave behind what a lender, reviewer, or examiner needs later.
Open the evidence.
Show what was asked, what returned, and why it mattered.
Keep raw data bounded.
Bring back proof, not a broader private-data burden.
Define proof first.
Gate evidence before it becomes decision-grade.
Standards
Built to the standards our customers are held to
Designed to align with the frameworks our customers are examined against, including FFIEC cybersecurity guidance and SR 26-2 model-risk guidance, which supersedes SR 11-7.
SOC 2 (readiness)
ISO/IEC 42001 (aligned)
AIGP-certified staff (credentialed)
NIST AI RMF (aligned)
GDPR / CCPA (aligned)
Next step
See whether this fits your portfolio
A 30-minute briefing on where Kenshiki fits.
- Portfolio fit
- Fraud exposure
- Governance posture