The AI loan review platform

AI-assisted loan review, audit-grade by design.

Klerum reads credit files, recommends your review sample, normalizes cash flow, and drafts the memo. Every AI output cites its source document and page, and your reviewers make every rating decision.

Built for independent loan-review firms, community banks, and credit unions.

Engagement overview: progress, milestones, and team
An engagement at a glance — workpaper progress, milestones, and the assigned review team.
Page + excerpt
Citation behind every AI-extracted claim
Cents per file
Measured AI cost to read a credit document
100% human
Rating decisions stay with your reviewers
What the AI does

The mechanical work, handled. The judgment, yours.

Klerum's AI reads the file, spreads the numbers, reasons like a senior reviewer, and drafts the memo. It works against your methodology, and it shows its source for everything it surfaces.

Reads the credit file

Point the AI at a credit presentation and it extracts approval terms, DSCR, LTV, sources of repayment, weaknesses, related entities, and covenants. Financials, personal financial statements, and collateral valuations come back spread and structured.

Reasons like a senior underwriter

Key Considerations reads the whole relationship — loans, collateral, guarantors, cash flow — and surfaces the strengths, watch items, and risks a senior reviewer would raise, each with the questions to chase and the signals that triggered it.

Normalizes cash flow

Adjusted cash flow, DSCR, and surplus compute live as the reviewer enters add-backs, and global cash-flow groups combine DSCR across related entities and guarantors in one view.

Tests against your loan policy

The AI turns the bank's loan-policy manual into testable underwriting standards, then checks documents and collateral against them — drafting the documents that should be on file and raising the exceptions when they are not.

Builds and refines your sample

Upload the loan trial and the AI proposes a review population against your scope and profile, with a per-borrower rationale. Refine it in plain English instead of a thirty-checkbox modal.

Drafts and revises the memo

The AI drafts the review memo from the work already in the relationship, with cited evidence behind each point, and revises it when you tell it what to change. An independent AI peer review challenges the rating. A senior reviewer signs off.

How our AI is different

AI you can put in front of an examiner.

Generic AI gives you an answer. Klerum gives you an answer you can defend. The AI never makes a credit decision. It drafts, extracts, and recommends, and every figure it produces points back to the document and page it came from.

  • Every extracted figure shows its document and page number
  • A verbatim excerpt sits beside each claim for verification
  • An independent AI peer review challenges the primary rating
  • Each rating change is recorded with actor, time, and diff
Key Considerations

The AI reasons like your most senior reviewer.

Klerum reads the whole relationship — loans, collateral, guarantors, cash flow — and returns the strengths, watch items, and risks a senior reviewer would raise, each written up with the questions to chase and the exact signals that triggered it. That is the difference between AI that summarizes and AI that thinks like a reviewer.

AI Key Considerations for a commercial relationship
A 662% aggregate LTV and thin guarantor liquidity drive a written risk consideration, with the questions a reviewer should chase and the signals behind it.
Engagement profile editor
Your methodology, applied by the AI

The AI works to your standard scope, not a generic one.

Your sampling approach lives in an editable engagement profile: coverage targets, exposure tiers, classified and criticized handling, document-review depth, and weighting. The AI applies it the same way across every bank, every engagement.

Private by construction

The borrower's PII never reaches the model.

Before any document reaches the AI, an in-perimeter service redacts it: names, SSNs, and tax IDs are replaced with tokens like [PERSON_1] and [SSN_1], and the result is rehydrated on our own servers. The model only ever sees anonymized text, and if redaction is unavailable, the AI simply does not run. The math — DSCR, LTV, coverage — is deterministic code, never AI.

  • PII tokenized before anything is sent to the model
  • Model output rehydrated server-side, inside the perimeter
  • Redaction fails closed — no redaction, no AI
  • Every ratio and coverage figure is computed, not generated
Who it's for

One AI platform, two buyers.

Independent review firms

Put your methodology behind the AI and cover more banks without scaling the review team. Multi-FI structure is first-class: one firm, many institutions, per-bank overrides.

Community banks & credit unions

Let the AI do the heavy lifting on extraction and coverage while a small review team keeps control. FFIEC-aligned, with citable evidence and regulator-ready documentation.

AI portfolio considerations for an M&A target
New · M&A due diligence

Diligence a target's whole loan book with the same engine.

Point Klerum at an acquisition target's loan tape and it re-grades the portfolio, marks the incremental credit reserve, and drafts the diligence takeaways — concentration, credit quality, maturity walls, and sector exposure — from anonymized figures only.

See Klerum on your own portfolio.

A 30-minute demo, no deck. Upload a sample trial and watch the AI propose a review population against your methodology.