LendAPI

The platform carries the work without its AI: origination, a no code decision engine, contract generation, pricing and servicing run on configured flows and versioned rulesets. AI sits across it in four places. ARC AI drafts flows from prompts and pre reads applications, Doc AI extracts document data, Underwrite AI answers questions against a lender's guidelines and documents, and Model Runtime runs a lender's own credit and fraud models in under 100 milliseconds, returning a probability of default and risk tier in under 50.

ARC AI reads each application, applies the lender's rules and drafts the next action, and exceptions wait for a person's sign off while the rest is completed before staff log in. Which applications count as exceptions is set by the lender's rules rather than described as a default.

Model Runtime can send high risk applications to manual review automatically, Doc AI attaches a confidence score to every document, and staff in the agent dashboard approve or decline documents with every action written to an audit trail.

Accuracy figures for Doc AI and ARC AI are not published, but Model Runtime gives a lender's risk team working controls. Each model file is fingerprinted with a cryptographic hash and stored in an encrypted registry, new versions can be shadow tested on 1 percent of traffic and compared on lift before promotion or rollback, and underwriting strategies can be backtested and A/B tested on the platform, with latency monitors and audit logs throughout. Doc AI scores every document for interpretability and completeness, and Underwrite AI cites the source behind each answer.

1st Franklin Financial, a family owned consumer finance company founded in 1941 with more than 400 branches, reports a 41 percent reduction in manual overrides since migrating from MeridianLink, with credit policy deployment cut from weeks to hours. United Consumer Financial Services, a Berkshire Hathaway company serving more than 4 million consumers, and JG Wentworth are named in customer stories, and the chief executives of Relli, RAI Partners and Enervee are quoted. The outcomes describe the platform rather than its AI features.

Whether client data reaches the models behind ARC AI, Doc AI or Underwrite AI is not stated, and the terms allow LendAPI to process client data to improve the service. Isolation is described in architecture terms instead. Tenants and sub tenants are separated below the application code, and each model a lender loads into Model Runtime runs in its own hardware isolated enclave inside LendAPI's private network, with no shared memory and no public address.

Client data belongs to the client under LendAPI's terms, which let LendAPI process it to provide and improve the service, commit to commercially reasonable safeguards and make data available for export for 30 days after termination. A data processing agreement, subprocessor list or retention schedule for borrower data is not published, and the privacy notice says personal information may be shared with service providers and business partners. Each tenant can configure its own legal disclaimers, consent prompts and data access rules.

LendAPI has completed a SOC 2 Type 2 audit, with Johanson LLP as auditor and SecureFrame monitoring its infrastructure continuously, and Model Runtime is described as running inside that SOC 2 environment. The report period is not given and no trust portal is published. Data is encrypted at rest under the Advanced Encryption Standard, model files are verified before first run, and every user action in the servicing system is logged and downloadable.

LendAPI names no regulator, statute or rule for its platform, and its terms put compliance with applicable law on the client. Know your customer and know your business modules cover identity, entity, beneficial owner and watchlist screening, and each tenant can configure its own disclaimers and consent prompts. Disputes go to binding arbitration under California law.

The decision engine and Model Runtime approve, decline and price credit with lender rules and models, and ARC AI flags risk on every application. LendAPI publishes nothing on fair lending or disparate impact for any of them, or on how its engine turns a decline into adverse action reasons, though its credit bureau service passes the bureau's adverse reason codes through with each file.

Its mission speaks of broader financial inclusion, and its tools let a lender see the factors behind an outcome, through ARC AI's explanations and the dashboard's record of every call and data point used.

LendAPI's terms put the risk on the client: the service is provided as is, other warranties are disclaimed, the client's remedy is limited and total liability is capped at the fees paid in the prior 12 months. What a lender gets instead is traceability, with every ARC AI flag linked to supporting data, every Underwrite AI answer cited and a full audit trail of calls and data for each application.

The models behind ARC AI, Doc AI and Underwrite AI are not named, nor is their provider or host. Lenders bring their own models to Model Runtime as Docker images, Pickle files or Predictive Model Markup Language files, which makes them the supplier for that layer, and data suppliers are named individually in the marketplace, from Equifax and Plaid to Middesk.

Connections to a core banking system are not named, but the data layer is. A marketplace of more than 270 integrations connects providers including Equifax, Plaid, Flinks, Middesk, Baselayer and Spinwheel, and every flow can be triggered through an interface with custom input and output schemas, backed by webhooks and a sandbox for developers. United Consumer Financial Services pairs LendAPI's origination with LoanPro for servicing, and 1st Franklin Financial migrated to it from MeridianLink.

The hosting provider, region and any residency commitment are not stated, and no private or on premises option is described. LendAPI runs as a multi tenant cloud service, models in Model Runtime execute inside its private network with encryption at rest, and lenders typically go live in about a month without writing code.

Pricing is public in structure though not in figures. The origination system is billed as a platform subscription plus a fee per application, and the Embarc loan management system as a monthly fee per loan, with amounts given on contact. A free trial and a sandbox tenant need no procurement process or credit card, so a lender can build a working flow before it negotiates.

Buyers span consumer finance companies, indirect and retail installment lenders, auto and marine lenders, point of sale and buy now pay later providers, fix and flip lenders and credit unions, plus fintechs that lend through bank partners. Embedded finance, point of sale, buy now pay later, retail and in store, auto and fix and flip lending each have a solutions page, while consumer finance companies such as 1st Franklin Financial, marine lenders and credit unions have none of their own.

The evidenced market is the United States. Dealer groups, store banners, referral partners and credit union branches can each run as a sub tenant with their own products, pricing, users and reporting under the lender's policy.