Aloan

Spreading is the structured output of a model that has read the whole tax return, and tracing multiple entity K 1 schedules runs as a reasoning task rather than a tab an analyst opens. Remove the models and only the manual process it replaces is left.

Aloan prepares and people decide. The lender's credit policy is compiled into rules that the credit team reviews, approves and activates, every deal is checked against them, and each exception cites the policy section and page for the underwriter, who makes the final call. Multi tier approval authority and decline recording are on the roadmap for the fourth quarter of 2026 rather than shipped, so approval routing above the underwriter still runs on the lender's own process.

Every calculated figure links to its source document by citation, the evidence supervisors now expect where models touch credit. Extraction accuracy and error rates go unpublished for a system reading tax returns and reconciling related entities. Aloan argues that a single accuracy percentage means little without a stated reference and a definition of an error, and it leans instead on tracing every number to the exact source page so a reviewer can check it.

Buckeye State Bank, Alliance Catholic Credit Union and West Central Bank are named as customers, though no figure is tied to any of them. The outcome accounts come from reviewers on software marketplaces, who describe going from application to credit memo within 30 minutes of upload and turning days of manual spreading into minutes.

Aloan's own figures are of the same kind, 77 percent less time per loan file from intake through credit memo and a complete memo in under 30 minutes, tied to no named customer, and those reviewers rate it five of five.

Customer inputs and outputs are never used to train, improve or fine tune any model, a promise Aloan puts in writing in a data processing agreement with zero training language. Aloan removes data on request or at offboarding under a retention policy whose period is not given.

Google Cloud and AWS, which run its inference, keep no copy after a request, inference runs in isolated environments, each institution's data is kept logically separate, and Aloan says it does not monetize, share or repurpose customer data.

Complete tax returns, personal financial statements and guarantor schedules arrive, some through a branded borrower portal that collects directly from applicants. Aloan offers buyers a data processing agreement with zero training language and lists its subprocessors in its trust center. How the portal tells applicants their data is handled is not described beyond the privacy policy. The platform and its data stay in the United States.

Aloan has completed a SOC 2 Type II audit covering security, availability and confidentiality, and its trust center shows the status as compliant, with the final report, penetration test results and an AI data isolation policy released on request under a nondisclosure agreement. Its inference providers, Google Cloud and AWS, are carved into that audit boundary.

Data is encrypted with Transport Layer Security (TLS) 1.2 or higher in transit and 256 bit Advanced Encryption Standard (AES) at rest, and a certificate of cyber and professional liability insurance for 2026 to 2027 is available to buyers.

Its controls are mapped to the Interagency Statement on Model Risk Management (SR 11-7), Office of the Comptroller of the Currency Bulletin 2023-17 on third party risk and Federal Financial Institutions Examination Council guidance, and Aloan publishes an examiner readiness guide for lenders. Its own writing tells lenders that 2026 guidance replaced SR 11-7, so documentation built on the older statement needs updating. Government guaranteed lending is handled at form level, with SBA forms stored in the platform, and Aloan claims no license of its own.

The agents spread the file, check it against the lender's credit policy and draft the memo a credit decision rests on, though the lender's people make the call. Aloan publishes no disparate impact testing of its models and no method for producing adverse action reasons. Its fair lending material is about consistency: it argues that policy thresholds belong in a controlled library so every application meets the same rules, and each exception it raises cites the policy section behind it.

Once documents go through the portal, nothing published tells a borrower that a model spread their financials or how a misread figure gets corrected. Source citation serves the credit committee and examiner, not an applicant facing a decline.

AI processing runs through Google Cloud's Gemini Enterprise and Amazon Web Services' Bedrock under enterprise terms that bar training on customer data, and Aloan describes a multi model pipeline that classifies, extracts and verifies each document. Which model handles which step is not named. Its data suppliers are the bureaus and account data services named in its integrations.

Aloan works beside the systems a commercial lender already runs. Lenders on nCino, Abrigo, Baker Hill, MeridianLink or Finastra keep it as the system that books the loan and run the analysis in Aloan, with web interfaces and webhooks carrying results across, and booking stays with the core. Credit and account data arrives from Equifax, TransUnion, Experian, Plaid and Stripe. Sanctions checks, lien filings, government IDs and SBA forms are stored at no per document cost.

The platform and its data are hosted in the United States, and AI inference runs in isolated environments on Google Cloud and AWS enterprise services. A lender can run Aloan as its commercial loan origination system or alongside the one it has, with booking left in the core either way. Most lenders go live in two to four weeks, with Aloan handling configuration, credit policy ingestion and integrations.

Aloan prices by the documents it processes rather than by seat: each contract sets a monthly minimum with per document overage above it, sized to the lender's portfolio mix and volume, with no multiyear lock in. Contracts typically run 12 months, with month to month terms for pilots and smaller engagements and a one time implementation fee. Three plans, Starter, Professional and Enterprise, differ by features rather than volume gates. No rate is published, so a figure takes a quote, which Aloan promises within one business day. The Starter and Professional plans come with a one week free trial.

Buyers are named precisely: community banks of $500 million to $10 billion running commercial real estate, Small Business Administration (SBA), commercial and industrial and equipment books, regional banks of $10 billion and up, credit unions doing member business lending, credit union service organizations and specialty and equipment lenders. A typical borrower is a business with $500,000 to $10 million in revenue seeking $50,000 to $2 million. Consumer lending is outside its scope.