Financial Intelligence & Underwriting

Understand today.
Balance always.
Build tomorrow.

FinYang AI reads any bank statement from any bank in any country and turns it into a verified ledger, clear cash-flow analytics, credit underwriting models and forensic intelligence — in minutes, not days.

  • Any bank · Any country · Any currency
  • Reconciled to the statement’s own printed totals
  • Underwriting & fraud detection in minutes
  • Reconciliation Matches to the cent Opening → Closing · 12 months
  • Debt service cover 1.84× after proposed facility
  • Net cash flow ▲ 12.4% trailing 6 months

01

How it works

From a pile of PDFs to a decision-ready picture.

  1. 01

    Upload

    Drop in statements as PDF, scanned image or CSV — one account or twenty, one month or two years. Password-protected files, mixed banks and overlapping periods are handled.

  2. 02

    Extract & verify

    A deterministic layout parser reads digital statements; cloud OCR catches scans. Every ledger is reconciled against the statement’s own printed opening balance, closing balance and totals — a document that does not add up is never presented as if it did.

  3. 03

    Understand & detect

    Monthly income statement, net cash flow, recurring debt service, facility usage and conduct signals. Affordability, debt-service-cover ratios, and forensic balance continuity are computed once and shown the same way on every screen and export.

  4. 04

    Decide

    An AI underwriting memorandum and forensic audit grounded only in verified figures, a chat that answers from the ledger itself, and one-click exports to PDF, Word, Excel and Markdown.

02

The FinYang idea

Two forces. One clear picture.

Every financial life is a balance between what grows it and what protects it. FinYang AI reads both sides of your statements and shows them in harmony.

Yang Growth

Income, assets, investments, future. The inflows that build tomorrow, read from every credit on every page.

Yin Control

Expenses, liabilities, protection, today. Debt service, fees, fraud detection, conduct signals and the cash you must keep on hand.

Balance

Bringing every part of your financial life into harmony: verified numbers, honest ratios and decisions you can stand behind.

03

One engine, three questions

The same statements. A different question each time.

A bank statement is evidence, and what it proves depends on what you are asking. FinYang AI reads it once and answers in the register you actually need — with a different analyst, a different dashboard and a different report in each mode.

Every mode reads the same verified ledger. What changes is the question, the dashboard and the report.

Credit & underwriting

Lenders, credit desks

The question

Can they afford it, and will they repay?

What it computes

Affordability, debt-service cover, runway and conduct, against your own DSCR covenant and decline thresholds.

What you get

A credit memorandum with a defensible recommendation, naming the months it could not verify.

What it will not do

Apply our lending appetite instead of yours.

Business insights

Owners, finance teams

The question

Where did a year of money actually go?

What it computes

Category and counterparty breakdowns, income, runway and seasonality across the whole period.

What you get

An operating picture in plain language, with every figure traceable to its transactions.

What it will not do

Use one word of credit language — nobody here is applying for anything.

Forensic audit & compliance

Investigators, AML teams

The question

Is this document honest, and is this money clean?

What it computes

Balance continuity, structuring, layering, pass-through velocity and source-of-funds signals.

What you get

A forensic pack listing each signal with the exact lines behind it.

What it will not do

Reach a verdict about a person. Signals are reasons to look, not conclusions.

04

What it finds

The things that are easy to miss on page 40.

Every one of these is computed from the verified ledger and cites the transactions it came from. Nothing is inferred from a summary — they are reasons to look at named rows.

  • Altered statements. Every line must move the running balance by its own amount. Where it does not, a line is missing or the document has been edited.
  • Structuring. Cash deposits shaped to sit just under the reporting threshold — measured against the jurisdiction the statement was actually issued in.
  • Same-day splits. Separate sub-threshold deposits that only cross the line when added together.
  • Money-mule layering. Funds received from individuals and forwarded to individuals within days, and one reference series split across several payees.
  • Account-verification probes. A token payment from a payer, followed by a substantial one.
  • Rapid dissipation. Inflow that leaves again in the same period — the shape of a conduit, not a trading account.
  • Concentration and cash intensity. How much of the income comes from one payer, and how much of it arrives as cash.
  • Conduct and credit strain. Returned debit orders, overdraft breaches, payday and merchant-advance lenders, and every recurring commitment.

Computed on every forensic run Balance continuity · Structuring · Same-day splits · Reference series · Peer layering · Test deposits · Pass-through · Round-sum velocity · Payer concentration · Cash intensity · Returned debits · Overdraft strain · Payday lenders · Duplicate statements · and more

Signals for review, never a verdict about a person.

05

Built for anywhere

Any bank. Any country. Any currency.

FinYang AI does not assume where a statement comes from. It reads the letterhead, the regulator line, the currency and the date format and places the document in the right region automatically — then formats every figure the way you expect.

  • Currency, locale and date order detected from the statement itself
  • Comma-decimal, dot-decimal, space-grouped and lakh-grouped amounts
  • Dedicated layout readers for banks with a corpus behind them, plus a generic ledger parser for everything else
  • A region registry of 60+ countries: banks, reference base rates and vocabulary — extended with data, not code

Currencies & regions

  • South Africa
  • United Kingdom
  • United States
  • Germany
  • France
  • Australia
  • Canada
  • India
  • Nigeria
  • Kenya
  • UAE
  • Singapore
  • Japan
  • Brazil
  • Mexico
  • + 50 more

Statement text is the only input. No manual country selection required — set one when you want to pin it.

06

What you get

Everything between the statement and the decision.

Eighteen capabilities, grouped by the three things the system actually does: read the document, compute the numbers, and answer for them.

I

Reading the document

01

Deterministic extraction

Column-aware parsing that reads the ledger the way it is printed — newest-first, column-major, split-page and overlaid layouts included.

02

Printed-summary verification

Every period is checked against the bank’s own opening, closing and turnover figures. Partial or unverifiable documents are labelled, never hidden.

03

Scanned, photographed & locked files

Phone photographs and faxed scans go through cloud OCR with spatial layout reconstruction. Password-protected and encrypted PDFs are read where the structure allows, and clearly reported where it does not.

04

Duplicate & overlap detection

The same statement uploaded twice, overlapping months, and two accounts printed on one page are all detected and separated — so a period is never counted twice and a total is never quietly inflated.

05

No page or document limit

Documents are read page by page and resumed automatically where a run stops, so length is not a ceiling — a hundred process at once and the queue simply carries on. Long statements, whole years and multi-account files go in together and come back as one reconciled ledger.

06

Any language, any script

Descriptions are read as tokens and counterparty names, not against an English word list. Written month names, comma-decimals, lakh grouping and non-Latin scripts are all handled — a new language is a table entry, never a code change.

II

Computing the numbers

07

Debt-service cover & DSCR

Monthly performance, net movement, debt service, runway and debt-service cover across your accounts — computed once, shown everywhere.

08

Fraud & forensic detection

Detect altered balance chains, sudden fund dissipation, round-sum velocity, split transactions, and suspicious sweeps with mathematical continuity checks.

09

AML & structuring alerts

Automated source-of-funds verification, KYC provenance tracing, pass-through velocity analysis, and regulatory structuring flags for compliance teams.

10

Conduct & risk signals

Overdraft strain, bounced debit orders, loan repayments, high-velocity cash sweeps and policy thresholds highlighted automatically.

11

Executive dashboards

A dashboard per mode, not one dashboard with things hidden. Monthly performance, ratios and signals, every figure computed once and identical on screen and in the export.

12

Your risk tolerance, your thresholds

DSCR covenant, minimum runway, debt-to-income ceiling, margin over base rate, industry focus and written lending criteria — set once and applied to every assessment. Non-lending modes hide them without discarding them.

III

Answering and issuing

13

AI underwriting memo

A structured credit memorandum written only from verified figures, assessing debt-service capacity, runway, and risk thresholds — cross-audited by an independent AI reviewer.

14

AI chat, grounded in the rows

Ask anything — “largest recurring debit?”, “debt service cover with the proposed loan?”, “show every round-sum cash deposit” — and the assistant queries the ledger itself and answers with the transactions that prove it.

15

PDF, Word, Excel, CSV & Markdown

One click. The credit memorandum or forensic pack as a formatted PDF or Word file, the full structured ledger and every metric as a real Excel workbook — money as numbers, dates as dates.

16

Institutional exports & audit trail

Executive summaries, credit memos and transcripts to PDF, Word, Excel, CSV and Markdown. An admin vault keeps every AI interaction for compliance review.

17

Custom client profiling

Record what the numbers cannot say — the trade, the season, the one-off contract, the reason last March looks wrong. It is carried into every memorandum and marked as your context, never mistaken for measured evidence.

18

Always on current models

Running on the current generation of frontier AI models, named in one place and verified against the live API — so a provider deprecation is caught before it reaches a customer, not after.

It cannot invent a figure it did not read.

07

Who it is for

A bank statement is read for more than one reason.

The same document decides a loan, exposes a fraud, settles a dispute and explains a year of trading. FinYang AI is built for all four, and knows which one you are doing.

Lending

Lenders & credit desks

Affordability and debt-service cover against your own covenants, with a memorandum that says why — and names the months it could not verify.

DSCR 1.84× · 12/12 months verified
Forensics

Forensic accountants & investigators

Statement integrity, laundering typologies and asset dissipation, every finding traceable to the lines it came from — which is what survives being questioned.

every finding cites its rows
Compliance

Compliance & AML teams

Source-of-funds review, structuring and pass-through checks measured against the reporting threshold of the country the statement was actually issued in.

threshold read from the issuing jurisdiction
Insights

Business owners & finance teams

Where a year of money actually went, by category and counterparty, with runway and seasonality — no credit language, because nobody is being assessed.

12 months · by category and counterparty
Practice

Accountants & bookkeepers

A structured, reconciled ledger out of a client’s pile of PDFs, categorised and exportable to Excel in one pass instead of an afternoon of retyping.

reconciled ledger → Excel, one pass
Legal

Legal, insolvency & disputes

Disclosure that runs to hundreds of pages, read in full and reconciled to each statement’s own printed totals — with the arithmetic shown rather than asserted.

no page limit · arithmetic shown

08

Security

Your statements stay yours.

Built for lenders, finance teams and operators who answer to regulators.

  • Encrypted in transit and at rest. Files live in a private bucket; access is scoped to your workspace.
  • Row-level security. Every table is guarded by database policies tied to the signed-in user.
  • Zero persistent OCR storage. Cloud OCR is used to read, not to keep. Extracted text lives only in your database.
  • No keys in code. Every credential is injected at runtime; nothing is committed.
  • GDPR, POPIA & global privacy-law aligned. Designed for data-protection regimes on every continent.
  • Admin vault and audit trail. Every AI interaction is retained for compliance review, exportable on demand.

09

Transparent plans

Simple, predictable pricing.

Choose the plan that fits your statement analysis volume. Start free, upgrade when you need to.

Free

Try it properly with weekly evaluation.

$0 / month
  • 1 seat (single user)
  • Weekly statement quota
  • A full statement, read end to end
  • Insights, metrics and exports
  • Interactive ledger chat
Get started free

Max

For high-volume teams and enterprises.

$100 / month
  • Unlimited team seats
  • Shared company data & quota pool
  • 5× the weekly capacity of Pro
  • Large multi-statement batches
  • Highest burst allowance
  • Everything in Pro
Choose Max

Billed monthly in USD or EUR. Cancel anytime with 1 click. Backed by our 14-day money-back guarantee.

Bring your financial picture into balance.

Upload a statement and see the whole story — growth, control and the balance between them.

Open FinYang AI