FundMore.ai

The document processor is built on natural language processing and machine learning to recognise, sort, digitise, label, extract and analyse borrower documents, described as the first of its kind in its home market, and an agentic assistant applies rules to assess eligibility and affordability and recommend structures. Scoring uses pattern recognition alongside logic-based decision making to grade mortgages. Remove the models and the spreadsheet-heavy manual process the company measures itself against is what remains.

In the chief executive's plain words, underwriters are not removed from the process, but their decision is enhanced by returning a recommendation to approve, decline or review further, with very clear narratives and all the reasons attached. The scoring widget shows which individual factors passed or failed as a visual cue to where the file needs attention. The stated design goal is consistency between branch and centralized underwriting rather than dependence on individual talent. One path bypasses that model entirely. Applications are automatically declined where the applicant appears on a blocklist or the property is in an unsupported region, with no review step described.

Explainability is delivered at the level a credit committee can use, with the score decomposed so that individual passing and failing factors are visible, recommendations accompanied by clear narratives and full reasons, and audit-ready reporting throughout. Rules are lender-defined rather than vendor-set, so the institution owns the logic being applied.

No accuracy, extraction error rate or validation result is published for the document processing and scoring models, and an independent assessment questions how complete the decision engine actually is.

The only outcome figure is self-reported, with lenders said to report over 50 percent gains in underwriting efficiency, particularly in broker-based models, and no customer is named anywhere. Awards are referenced generically. An independent review offers a useful counterweight, assessing the platform as strong in document automation but focused more on extraction and automated conditioning than on being a complete underwriting decision engine, and suggesting it pairs best with separate tools for full guideline checks.

No boundary statement is published. The platform processes complete borrower document sets across lenders and brokers competing for the same applicants in one national market, and document extraction models improve with exposure to varied real world files. The architecture therefore raises the question of whether customer material informs them. Nothing states whether data is isolated per lender, whether models learn across the customer base, or what happens to borrower documents after a file closes.

The applicable privacy statute for its jurisdiction is named directly and treated as an automated compliance requirement rather than a policy statement, alongside financial crime and identity checks. Secure data handling is claimed, with audit ready reporting. No data processing agreement, retention schedule or subprocessor list is published. The platform ingests complete borrower document sets, including income and identity records.

There is no public attestation, certification, trust center or enumerated control set. Secure data handling is claimed, along with fraud detection built into the platform, and institutional banks are named as a target segment. A prospective buyer at that end of the market would require control documentation that is not published.

Three regimes are named for its home jurisdiction and described as automated within the workflow: financial crime and identity verification, the prudential supervisor's requirements, and the federal privacy statute. They are supported by fraud detection built into the platform and audit ready reporting. The domestic prudential regulator is named specifically. No individual rule, guideline or supervisory expectation within those regimes is mapped to a product control.

No fairness testing or bias disclosure is published, and one published mechanism stands out. Applications are automatically declined where the subject property is in an unsupported region. That is a geographic exclusion applied without human review, and in mortgage lending geographic exclusions are the precise mechanism fair lending supervision exists to scrutinize. The lender defines the regions rather than the vendor, and the platform executes the rule at speed and scale with no described monitoring of its distributional effect.

Nothing published offers a guarantee, an indemnity or a correction process. The borrower's position is uneven. An applicant who reaches an underwriter benefits from recommendations carrying explicit reasons, which supports a meaningful explanation of a decline. An applicant automatically declined on a blocklist or geographic rule never reaches a person at all and has no described route to see why or to contest it.

No base model, provider or hosting arrangement is identified behind the document processing or the agentic assistant. External data dependencies are described only by category as credit bureaus and insurers, and in mortgage those providers determine what the underwriting sees, so a lender cannot document either its model or its data dependencies from published material.

Integration is described by category across the systems a mortgage lender actually runs, covering credit bureaus, insurers, point-of-sale platforms, customer systems and internal databases through open interfaces, with electronic signature and digital legal portal support for the closing end. Independent assessment credits good interface connectivity. Not one bureau, insurer or platform is named individually, so a lender cannot confirm its own stack is supported.

Hosting is undisclosed: there is no published provider, region, residency commitment or private deployment option. Residency is a live question in this market specifically, since the privacy statute the platform automates compliance with carries expectations about cross border transfer of personal information. The company does not address where borrower documents are processed or stored.

The company publishes no pricing. An independent review discloses the structure, describing per-document pricing and noting it can add up on paper-heavy files, which is a meaningful warning for a mortgage lender whose files routinely run to hundreds of pages, and no rate accompanies it.

The range is stated as spanning institutional banks down to private lenders, with brokers and broker-assisted channels covered explicitly and intake supported through web, mobile and broker routes, which matters in a market where brokered origination is a large share of volume. Coverage is one product line in one country, and no institution count or market share is evidenced.