A Maple Crew product
Maple Insight turns a borrower's documents into an explained score out of 100, a comparison of financing products and a draft credit note. Built for credit analysts in retail and SME banking.
Demonstration data. All customers and companies shown are fictional.
Why Maple Insight
A credit file is mostly reading: statements, payslips, financial filings, a credit bureau report. Then comes arithmetic, then a note for the committee. The judgement that only an experienced analyst can bring gets whatever time is left.
Maple Insight does the reading and the arithmetic, shows its work, and prepares the note. The analyst verifies, corrects where needed, and spends their time on the questions that matter. The lifecycle is the same for every file: draft, to complete, ready for assessment, assessed, decided, closed.
Two modules, one product
The analyst uploads the customer's standard documents. Maple Insight reads six months of bank statements and payslips, the employment letter, identity and tax documents and the credit bureau report.
One case per financing request. Maple Insight reads three fiscal years of financial statements, twelve months of business account activity, the company credit bureau report and the legal documents.
How it works
A classifier sorts each uploaded file. One extraction agent per document type copies the values as printed, with a confidence level.
Coded controls compare the documents with each other: same identity, balances that chain, salary found on the statement. Problems are flagged before any scoring.
Six specialist agents each assess one criterion and explain it. Every figure comes from tested code, and the rule always wins over the model.
Three financing products are ranked by suitability, with the monthly commitment and its effect on the borrower's ratios.
The AI drafts a note grounded on the computed figures: summary, favourable and adverse factors, reservations. The analyst edits and validates it.
An authorised person records the decision with a mandatory reason, separately from the recommendation. Every step is kept in the case history.
Inside the product
The interface is shown here in French.
Each control, module by module, with what was compared. Blocking controls can be switched off only with a written reason.
Three financing products ranked by suitability, with the calculation shown line by line.
A draft built only from computed figures and validated data, ready for the analyst to edit.
Guardrails
Credit decisions affect people's lives. These rules are built into the product and are not configurable.
Every result is labelled for human review. Nothing is granted automatically, whatever the score.
Maple Insight shows what it can justify: a transparent scorecard, not a statistical promise.
They are not inputs to any rule, and they are not extracted from identity documents.
Scores and ratios are computed by deterministic code. The language model explains and drafts; it does not invent figures.
When essential data is absent, the result is 'insufficient data' with the list of what to request. The score is withheld.
The screens distinguish what the documents say from what the system concludes, so a reviewer can always check one against the other.
We will walk you through a complete file, from document upload to the recorded decision, and discuss how it could fit your credit policy.
