LendFoundry

The platform originates, decides and services loans without its models: the decision engine runs lender configured rules in sequence, and origination, servicing, collections and bureau reporting are workflow and data. AI sits on top as language model agents that summarize and extract, machine learning analytics that predict default, and external scoring models a lender can feed into the rules.

The language model agents return a summary, reminder, action item or answer to the lender's origination and servicing screens or to an endpoint for staff to use. In the decision engine, lenders set thresholds that route borderline credit scores, missing documents and policy exceptions to underwriters for manual review, and multilevel approval tiers can follow loan amount, risk score or internal policy. Instant approvals sold as AI driven run through the same lender rules, with no separate control described for the model's part in them.

LendFoundry publishes no accuracy or validation figure for the loan defaulter model or the language model summaries. On the decision side, every application carries a summary of the rules triggered, the data used and why the outcome was reached, and rules, conditions and outcomes are versioned so changes can be tracked and performance reviewed over time. Rule changes can be simulated against historical data in a sandbox before they are published.

BriteCap, a California business lender, is named, and its chief executive says the managed origination and servicing systems replaced a homegrown system that was expensive to maintain and slow to change, and let the lender scale with little technology of its own to run.

The measured outcomes sit with unnamed clients: a New Jersey business lender cut abandoned loan applications by a third, and a merchant cash advance servicing program for a South Korean ecommerce platform went live in 90 days and grew active loans 14 percent month over month. None of these figures is tied to the AI features.

Whether lender data trains any model, and what a model provider retains, are not stated. The agents are built around tenant boundaries instead: each one fetches only data scoped to a single lender and runs independently of the others, and a lender can bring its own large language model rather than use LendFoundry's, which keeps the choice of provider with the lender.

Support staff see a lender's customer data only when the lender grants access for troubleshooting, and that data is not downloaded to or processed on LendFoundry's internal systems. Retention terms, a subprocessor list and Gramm Leach Bliley Act alignment for borrower data held in the platform are not set out. The agents are presented as compliant with the General Data Protection Regulation, the California Consumer Privacy Act and the Fair Credit Reporting Act.

TÜV certifies LendFoundry to ISO 27001:2022 for information security and ISO 9001:2015 for quality management, and it holds System and Organization Controls (SOC) 1 Type II and SOC 2 Type II reports, attestations from an auditor it does not name. The reports and certificates are not offered through a trust portal. Inside the platform, role based access controls and detailed audit trails cover users and decisions.

LendFoundry states no license or examination of its own. The compliance work sits in the product for the lender: Metro 2 files in the Consumer Data Industry Association format go to Equifax, Experian, TransUnion and Innovis, with reporting built around Fair Credit Reporting Act accuracy, and the platform keeps audit trails and runs compliance workflows.

The decision engine approves and declines credit applications, and LendFoundry sells AI driven decisioning for instant approvals and AI powered credit scoring alongside it. Nothing addresses fair lending, disparate impact or adverse action notices for those decisions, and no bias evaluation or governance process is named for the scoring, the loan defaulter model or the agents.

The decision summary records which rules fired and what data drove each outcome, material a lender could draw on for adverse action reasons, though nothing connects it to that use or to testing approval rates across protected groups.

LendFoundry publishes no accuracy commitment or indemnity for decisions or AI output. Each decision's summary of triggered rules, input data and outcome is recorded for audit, compliance and appeals or reviews, which gives a lender, and through it an applicant, a record to challenge a decision against.

Neither the agents nor the loan defaulter model names the model or provider behind it, and lenders can swap in their own large language model. Amazon Web Services appears as a partnership and in a client deployment, which shows where the platform can run rather than which models it uses. On the scoring side the external models are named where they plug in: Fair Isaac's Small Business Scoring Service, fed with Experian data, can supply a score to the lender's rules.

More than 90 prebuilt integrations connect the platform to credit bureaus, bank account and transaction data, identity verification, fraud prevention, payment processors, customer relationship management platforms and eSignature providers. Named ones include Giact, DataMerch, BirdEye and the Better Business Bureau, and Metro 2 files go to all four consumer bureaus.

LendFoundry is itself the origination and servicing system for its clients, and for one merchant cash advance client the servicing system creates journal vouchers in SAP automatically. Other named connections include Salesforce, kept in real time sync with loan data and automated workflows, Fair Isaac's Small Business Scoring Service with Experian data for small business credit, LexisNexis SmartLinx for compliance checks and Persona for document based identity verification.

LendFoundry runs as a fully managed software as a service platform on a cloud native microservices architecture, and a lender can deploy the whole platform or single modules. No region or data residency commitment is stated for the managed service.

A second option appears in a South Korean merchant cash advance program, where the servicing system was deployed inside the client's own Amazon Web Services virtual private cloud on Elastic Kubernetes Service clusters, running active in more than one availability zone.

LendFoundry sells by demo request and publishes no price, tier or unit of charge. The pitch is lower upfront cost from a fully managed software as a service model and a launch in weeks with modular rollout, so a lender can start with a single module and add the rest later.

Nonbank lenders, fintech lenders and specialty finance companies are the core market, with banks named as buyers too. Each asset class has its own page, from personal and working capital loans to merchant cash advance, hard money, healthcare, home improvement and commercial lending, and embedded programs get dedicated dealer, contractor, merchant and patient financing portals. Case studies cover a nonprofit community development financial institution, a Texas hard money lender, a Florida home improvement lender and a merchant cash advance program in South Korea.