Acceptance-tested on your own documents · #1 on NetArena MALT (third-party-run)

Banks and lenders

Underwrite from the right figure,
from the right period.

General AI tools sound confident. The work that carries risk is sourcing the right figure from the right period — and noticing when two documents disagree. The API returns every figure with its page and its calculation, and surfaces conflicts instead of resolving them silently.

Who it’s for

Credit, review and reporting teams.

Commercial and CRE lenders, credit and loan-review functions, regulatory reporting and research — at banks, and at non-bank lenders and debt funds.

Who buys

The people who own credit risk and the platform

  • The chief credit officer
  • The head of CRE lending
  • The head of commercial and corporate credit
  • The head of loan review or internal audit
  • The head of regulatory reporting
  • At large banks, also the chief data or AI officer and the owner of the internal AI platform

Who uses

The people who source every figure today

  • Underwriters and credit analysts
  • Portfolio managers
  • Loan reviewers
  • Reporting staff
  • Research analysts

Jobs we do for you

Every figure cited. Every conflict surfaced.

Each job starts from the files your team already receives and ends with figures that carry their page, their calculation and both sources when documents disagree.

Job 1 · Commercial real estate

CRE origination and annual review

Documents
Rent rolls, trailing-twelve statements, appraisals, leases, borrower tax returns, offering memoranda.
What you get back
  • Line items normalized to your template, each cited.
  • NOI, DSCR, debt yield and LTV with the calculation.
  • Conflicts surfaced with both sources: appraisal NOI vs. trailing-twelve; rent-roll total vs. T-12 rental income.
  • An evidence pack per file.
Measured by
  • Agreement with documented underwrites on closed deals.
  • Conflicts manual review missed.
  • Analyst hours per file.
Targets, measured on your documents
Start with
An in-perimeter pilot, then Enterprise.

Job 2 · Commercial credit

Spreading and covenant monitoring

Documents
Audited and interim statements, tax returns, annual and quarterly filings, credit agreements, compliance certificates, lender presentations.
What you get back
  • The spread to your chart of accounts, cited.
  • Each covenant recalculated from source using the agreement’s own definitions, with the definition cited.
  • Borrower-reported vs. recalculated differences flagged.
  • Headroom.
Measured by
  • Accuracy against a manual spread.
  • Hours per spread.
  • Certificate discrepancies found.
  • Days from package receipt to covenant test.
Targets, measured on your documents
Start with
An in-perimeter pilot, then Enterprise.

Job 3 · Research, M&A, capital markets

Defensible answers from public documents

Documents
Filings, prospectuses, offering memoranda, earnings and statistical releases.
What you get back
Candidate figures reconciled, which one was used and why, and the calculation.
Measured by
  • Time to a defensible answer.
  • Share of answers accepted without going back to the source.
Targets, measured on your documents
Start with
Early Adopter for teams on public documents.

Job 4 · Loan review, audit, regulatory reporting

Tie-outs with a re-runnable trail

Documents
Loan files, workpapers, source schedules behind regulatory reports.
What you get back
Figures linked to page and table, a re-runnable calculation, the version recorded, and report-line tie-outs with breaks flagged.
Measured by
  • Sampling error rate against your current process.
  • Prep time per file.
  • Exceptions found.
Targets, measured on your documents

Documentation for your model-risk and AI-governance review

Intended use, limitations, data sources, performance on your own documents, a monitoring plan and a per-answer audit trail. New versions ship only after passing your agreed test set.

Acceptance test

How we’d prove it on your documents.

We replay files your team has already underwritten, spread or reviewed, and score the output against what your file shows.

CRE

Closed loans with documented underwrites

NOI, DSCR, debt yield and LTV must land within tolerance. Conflicts are compared with what the file shows was resolved by hand. Plus a timed parallel run on new files.

Commercial credit

Your borrowers’ last four packages

A set of borrowers and their last four packages. Spreads are compared with the manual spreads, and covenants with your tested figures.

Loan review

Previously reviewed files

Exceptions found are compared with your reviewers’.

Pass criteria are agreed before the run, and we report the results as measured, misses included.

How acceptance testing works

Questions we hear

What credit and risk leaders ask first.

“Model risk.”

We provide documentation for your model-risk and AI-governance review: intended use, limitations, data sources, performance on your own documents, a monitoring plan and a per-answer audit trail. New versions ship only after passing your agreed test set.

“Third-party risk review.”

We complete your vendor security review: questionnaire, architecture review and support for your penetration testing. We don’t claim certifications we don’t hold — ask us for current status. See security and procurement.

“We already have a spreading or loan-origination platform.”

Keep it. The API is an endpoint that fills it — no new app for your staff.

“Who’s responsible for a wrong number?”

Your staff review outputs before relying on them, and the acceptance test measures the error rate before production.

“You’re a small vendor.”

Enterprise can be deployed in your cloud account or on-premises, and goes through your vendor, security and model-risk review before you sign. The data flow for each option, including any service providers, is documented up front.

Which license fits

Start inside your perimeter. Expand under Enterprise.

Borrower files that can’t leave the bank start with a fixed-fee pilot in your environment. Teams on public documents, and lenders that can use a hosted service under vendor management, can start with Early Adopter.

In-perimeter pilot · $40,000–$75,000 → Enterprise · $1,000,000

Borrower files that can’t leave the bank

An in-perimeter pilot on your own files, then Enterprise across credit, review and reporting. Pilot: acceptance-tested on your own documents before you pay the balance. Fixed fee, capped.

Ask about an in-perimeter pilot

Early Adopter · $20,000

Public documents, or a hosted service

Research teams on public documents; community and regional banks; non-bank CRE lenders and debt funds that can use a hosted service under vendor management.

Buy Early Adopter

ENTERPRISE

Contract-first

Enterprise license

Firm-wide, in your environment, or inside your platform.

Twelve-month license$1,000,000
  • Twelve months of development support with a named technical lead
  • Deployment in your cloud account, on-premises or hosted, scoped with you
  • Custom integrations with your systems
  • Documentation for your model-risk and AI-governance review
  • Acceptance test on your own documents, agreed in writing before you sign.
  • Paid by invoice, ACH or wire in milestones, never by card
Plan an Enterprise deployment

Keys are issued to verified businesses after a signed order. No free trial and no low-cost tier — here’s why.

How we prove it

Tested on your documents, not just ours.

Every license and engagement is measured on your own documents and on questions your team has already answered. We agree the pass criteria with you first — figures right within an agreed tolerance, every citation checked against its page, known conflicts flagged — and report the results as measured, misses included.

  • Pilots: Acceptance-tested on your own documents before you pay the balance. Fixed fee, capped.
  • Enterprise: Acceptance test on your own documents, agreed in writing before you sign.
  • Early Adopter: Your acceptance plan is agreed in week one, and six months of development support are aimed at passing it.

How acceptance testing works

A third-party-run result

#1NetArena MALT on AgentBeats
100% / 100%Correctness and safety over 2,500 queries
0.056 sFastest of the perfect runs

A network-policy benchmark, not a finance one. We show it because it’s how we work: define correct, test at volume, publish the run.

Standing checked October 6, 2026.

Talk to us

Plan a pilot on your own files.

Tell us which files you’d replay — closed CRE loans, borrower packages or reviewed files — and where they have to stay. We’ll come back with a pilot scope and a draft acceptance plan.

We reply within one business day. Prefer a call? Book 30 minutes.

We use this only to respond to your request. See our Privacy Policy.