A mortgage file runs from the loan application and income records to the appraisal, title commitment and closing package. Each value is read once and traced to its source page.
LandingAI transforms documents into highly accurate, verifiable, structured data so teams can reliably automate document-intensive workflows.
Loan files arrive as stacked PDFs. A self-employed borrower's file can run past 80 pages, with pay stubs, W-2s, tax returns and bank statements interleaved.3
Pay stub layouts change with every employer and payroll provider, and self-employed income spreads across a full return and its schedules. Income and appraisal issues cause more than half of GSE loan buybacks.4
The loan amount, property value and borrower address appear on the application, appraisal, purchase agreement, title commitment and Closing Disclosure. Every copy has to agree before the loan closes.
The same Parse call reads a uniform residential loan application, an appraisal report and a Closing Disclosure. Every document type uses one set of APIs, with zero templates to build and zero models to train or fine-tune.
A new form, issuer or document type works on its first call, with nothing to configure.
If an issuer redesigns a form or a new version arrives, no pipeline changes are needed.
Upload a sample in the Playground; agentic reasoning will immediately understand it.
Parse reads the whole stacked PDF into structured Markdown and keeps dense tables intact, from the appraisal grid to the Closing Disclosure cost sections.
Parse API →Split takes the Parse output and separates the file into its pay stubs, returns, statements and disclosures, with the page range of each.
Split API →Extract returns the fields in your schema as JSON for each document. Build Extract Schema drafts a starting schema from a sample.
Extract API →Ground maps each extracted value to the block, page and bounding box it came from, so a reviewer checks a flagged figure on the original page.
Ground docs →Trusted, high-volume processing of the documents involved in mortgage lending requires handling diverse document types and layouts. A pay stub is very different from an appraisal report, which is different from a Closing Disclosure. The pipeline must provide high accuracy and consistent outputs across all of them.
Our agentic approach is able to reason over each document type and return structured hierarchical Markdown and JSON outputs that downstream systems can rely on.
Each document below links to a worked example. The documents and schemas are available in full on GitHub and you are welcome to use them for your own testing.
| Field | Source document | Used for |
|---|---|---|
| Loan amount, loan purpose and occupancy | Uniform Residential Loan Application | Program eligibility and pricing |
| Gross pay and year-to-date gross pay | Pay Stub | Qualifying income |
| Wages, tips and other compensation | W-2 Form | Income history and trend |
| Adjusted gross income | Form 1040 | Income consistency across years |
| Net profit or loss | Schedule C | Self-employed qualifying income |
| Employer, start date and position | Employment Verification Letter | Employment stability |
| Ending balance and total deposits | Bank Statement | Funds to close, reserves and deposit sourcing |
| Credit score and total monthly payments | Credit Report | Eligibility and debt-to-income ratio |
| Purchase price and close of escrow date | Purchase and Sale Agreement | Loan-to-value basis and closing timeline |
| Opinion of market value | Appraisal Report | Loan-to-value ratio |
| Overall condition and quality ratings | Appraisal Report | Collateral eligibility and repair conditions |
| Flood zone and special flood hazard area | Flood Certification | Flood insurance requirement |
| Vested owners, requirements and exceptions | Commitment for Title Insurance | Clear title and closing conditions |
| Interest rate and cash to close | Closing Disclosure | Final terms checked against the approved loan |
| Principal amount, interest rate and maturity date | Promissory Note | Note terms match the Closing Disclosure |
| Grantee and legal description | Deed | Vesting matches the title commitment |
Mortgage operations leaders measure document work in three places: how many applications close, what each loan costs to produce and how many loans come back as defects.
Borrowers and agents reward the lender that delivers a solid preapproval first. When income documents are read in minutes, a file reaches the underwriter the same day, and fewer applications stall while processors rekey pay stubs and returns.
Processors, underwriters and closers read the same figures from the same pages. Extracting each value once, with its source attached, removes repeat data entry at every stage of the file and lets each team handle more loans.
Income and appraisal issues caused 57 percent of GSE loan buybacks in an industry study of April 2023 to October 2024. Values traced to their source give QC teams a faster check of the figures most likely to be challenged.4
Autyn runs a mortgage operations platform for independent mortgage brokers, and income reconstruction is the slowest step in each file. Its income engine now extracts a full borrower package in one to three minutes, work that takes a person one to two hours, with 94 to 98 percent field-level accuracy on borrower documents.
Read the Autyn case study“Very often a loan officer who gets a borrower a solid preapproval fastest earns the deal and the real estate agent’s referrals. Reconstructing income is the hardest part, and Agentic Document Extraction is the cornerstone to getting it right and traceable. Accurate data upfront means we can get a cleaner loan file to the underwriter and get to CTC faster. Get it wrong and everything downstream gets affected.”
Myra D'SouzaCEO, Autyn
The best accuracy on the difficult documents other tools get wrong, such as messy scans, complicated tables, checkboxes and handwriting.
Every answer cites its sources. Values come back with an audit trail (page, bounding box, character range) so a human can review them.
Applies multi-step reasoning, and plans before it acts. Handles documents it has never seen before and adapts to variations.
Composable APIs that drop into any tech stack and can be called from anywhere. Easy to build on and embed in your pipelines and products.
Cost-competitive rates with complete price transparency. Spend controls before you run, and every job reports its exact cost.
Built for production and scale from day one. Fully observable and monitored, running securely in our cloud or yours.
Available on Team and Enterprise plans. VPC and on-prem deployment require an Enterprise plan.
