TRUE AI
TRUE sells AI mortgage operations software to lenders, wholesalers and correspondents, mortgage insurers and loan buyers. Ari Gross and Alison Wasserman founded the New York company as SoftWorks AI, and it has traded as TRUE since 2022. Its Mortgage Operations Service handles more than one million loans a year, and the logos on its homepage include Fairway, MGIC, Arch MI, Arvest, First Continental, VIP Mortgage, HomeLend, Candor and Tavant.
MOS covers three stages. Loan Setup and Data Quality turns uploaded borrower documents into named, validated files, classifying each document, extracting income, asset and other data and cross checking values across documents and against a loan origination system such as Encompass. Underwriting Automation prepares decision ready borrower data, producing an income analysis in under three minutes without manual calculation. Processing, Closing and Post Close matches and clears conditions, splits closing packages into indexed documents and checks a file before it ships or goes to audit.
The models combine computer vision, character recognition, classification and language models, with deterministic rules on top. Every extracted value links back to its source page, and each check, edit and approval is logged through to the loan origination system. TRUE reports straight through processing above 90%, and First Continental Mortgage cut processing from almost an hour per loan to four loans in under 20 minutes.
Capability Axes
Capability grades
15 of 15 axes rated · 11 graded A or B
Reading borrower documents is the product's core job, and models do it. Computer vision, character recognition, classification and language models turn unstructured files into source linked values that the rest of the workflow can check, compare with the loan origination system and pass to underwriting. Deterministic rules run on top, but without the model layer there would be no data for them to check at loan volume.
Documents are classified, data extracted and compared across sources, and deterministic checks run before a lender relies on any value. Low confidence results, discrepancies and out of policy cases go to a reviewer with their source evidence, and the lender's own rules drive decisions. Every user interaction, change and approval is recorded up to entry in the loan origination system, and the work runs onshore rather than through offshore review teams.
Each borrower value carries a chain from its source document through transformations, validation, cross document comparisons and any human review, retained up to its entry in the loan origination system. TRUE maps that chain to Fannie Mae Lender Letter LL-2026-04 on AI and data traceability, so a lender can show how a value was produced rather than lean on a confidence score.
First Continental Mortgage reports that processing fell from almost an hour per loan to four loans in under 20 minutes with TRUE. Across its base, TRUE cites straight through processing above 90%, income analysis in under three minutes, up to an 80% cut in manual document handling and up to two days saved per loan, all vendor figures. Fairway, MGIC, Arch MI, Arvest and VIP Mortgage are among the lenders and insurers shown as customers, and the service handles more than one million loans a year.
Extracted values stay tied to their source evidence, and human edits are logged. TRUE has described its document technology adapting as it receives a client's documents and users make corrections, but whether that learning stays within one lender or improves a model shared across clients is not stated, and neither is how long any outside language model provider keeps requests. Customer data is deleted when a client leaves, according to the trust center's controls.
The service processes full borrower files, including income, asset and identity documents. Review screens keep the source page, field and audit history with every approval and edit. The trust center lists data retention procedures, data classification and deletion of customer data when a client leaves among its monitored controls. A full subprocessor list and a deletion schedule for AI inputs are not public.
TRUE's trust center lists a 2025 SOC 2 Type 2 report and a 2026 bridge letter, both released on request, beside a live view of monitored controls covering encryption, access restriction, penetration testing and disaster recovery testing. An earlier Insight Assurance examination of the TRUE Data Verification System, dated 14 July 2023, is freely available but covers the system that preceded MOS.
MOS prepares and checks the information lenders, insurers and loan buyers use, and the institution keeps the credit decision and its regulatory duties. The evidence chain back to borrower source documents can be retained for investor or regulator review, and the controls are framed against Fannie Mae's lender guidance on AI and traceability.
The decision design keeps validated facts and rule outcomes open for a lender's review and ties each adverse action reason to the lending rule that produced it, built with fair lending examinations in mind. Source fields, cross document checks and review history stay available when a file is reopened. Results by applicant group or a disparate impact measurement are not offered.
A lender can reconstruct where any borrower value came from, which checks ran and who changed or approved it, and adverse action reasons trace to the triggering rule. That evidence supports correcting a loan record, which stays the institution's job. A compensation commitment for a wrong extraction is not part of the offer.
TRUE built its own document recognition technology to read and structure mortgage data, and MOS combines several recognition engines, comparison rules and language models. Microsoft Azure hosted the earlier data verification system. The language models in the current service and their providers are not named.
Encompass, from ICE Mortgage Technology, is the named loan origination link: TRUE Data Intelligence connects to it so the lender processes the loan with validated data. MOS compares document data with loan origination records and keeps the source path when data is entered, and post close automation splits packages and files documents into electronic folders. Condition matching and clearing run against the same records.
MOS runs as a prebuilt cloud deployment with out of the box integration to a lender's loan origination system, and the work runs onshore. The 2023 examination of the earlier system names Microsoft Azure as host. A region commitment for current loan data, backups and AI processing is not stated.
Sales run through a demonstration and a scoping of the institution's loan operations, configuring file setup, underwriting data and post close review around the work the lender does. An online document test lets a prospect try selected extraction types. A price per loan or a schedule of software charges is not shown.
The service fits mortgage originators and wholesale and correspondent lenders that need complete borrower files for underwriting, mortgage insurers such as MGIC and Arch MI that assess loans for coverage, and servicers and loan buyers that review portfolios for missing documents and discrepancies after closing. Each group uses the same document evidence under the rules of its own workflow.
Alternatives to TRUE AI
Other products that do the same kind of work as TRUE AI, and what each one does.
DocVu.AI is Visionet Systems' document processing platform for mortgage lenders and servicers, launched in June 2021 and sold under its own brand.
Cotality, formerly CoreLogic, sells mortgage software for borrower verification, fraud review, and property valuation, along with property intelligence for insurers.
Copperlane automates mortgage intake, the stage where most of the roughly $11,800 it costs a lender to originate a loan is spent.
Land Gorilla sells construction loan software to banks, credit unions, independent mortgage banks and private lenders.
Tabrasa One is an AI powered growth platform for mortgage loan officers, brokers and lenders, combining a mortgage CRM, marketing, video and borrower follow up from Tabrasa, LLC in Red Bank, New Jersey.
Blend sells digital origination software to banks, credit unions, independent mortgage banks and mortgage servicers. Borrowers apply through its app, and lenders work the file in tools behind it, across mortgage, home equity, consumer loans and deposit accounts.
Pricing
Vendor-published figures are labeled as such. Figures labeled “Estimated” are derived from third-party sources and have not been confirmed by the vendor.
| Entry Price | Pricing Basis | Data Protection Terms | Implementation | Source |
|---|---|---|---|---|
|
Quote for the lender’s workflow
|
TRUE quotes mortgage workflow automation configured for each customer's operations, across loan setup, underwriting data and post close review. A per loan rate is not shown. | Borrower files are processed with source links, logged edits and approvals. TRUE's trust center lists data retention procedures and deletion of customer data when a client leaves. | Implementation configures document types, validation rules and the connection to a loan origination system such as Encompass, on a prebuilt cloud deployment. A standard setup charge is not listed. | Vendor Published |
TRUE is sold by quote after a demonstration, scoped to loan setup, underwriting preparation and post close auditing. An online test lets prospects try selected document extraction tasks.