Lama AI

The removal test leaves the legacy systems the company diagnoses precisely: static forms, rigid workflows and services-heavy implementations that cannot account for the thousands of borrower variations, document exceptions, policy nuances and edge cases defining small business lending.

Agents run the whole chain from intake through spreading, underwriting and decisioning to closing and monitoring, turning an unstructured borrower package into a decision-ready credit memo in about five minutes. That diagnosis is the clearest explanation in this index of why this work resisted automation until models could handle exceptions.

The containment principle is stated well and is the right one for this buyer: the platform operates within each bank's existing policies, credit standards, approval processes and compliance requirements, giving teams scale without replacing human judgement, and it deploys alongside incumbent systems rather than displacing the institution's controls.

Held at B because the described workflow runs through decisioning, approval and closing, and no threshold, referral rule or mandatory review point is published for where automation stops and a credit officer begins.

No accuracy, extraction error rate, validation result or evaluation method is published for agents that spread financials and produce credit memos at dozens of regulated banks. The published figures are speed rather than correctness, and to the company's credit one is qualified honestly, describing decisioning reduced from weeks to minutes in targeted workflows rather than universally. That qualification is not a substitute for measurement.

Six banks are named in production, including one with roughly 45 billion dollars in assets whose president of institutional banking is quoted saying that if a bank is going to employ artificial intelligence, credit underwriting is one of the most impactful places to start.

The company states dozens of institutions are live and that billions of dollars of loan volume have been processed, with revenue growing threefold year on year and total funding above 20 million dollars from a fintech-focused venture arm and established backers. Two analyst houses have recognised it, one naming it a major player in loan origination systems.

No boundary statement was located. Dozens of community and regional banks lending into overlapping local markets route their borrower packages and credit outcomes through one platform, and an embedded finance network connects platforms to a wider set of lenders, so the company observes demand and performance across competitors. Nothing states whether models learn across the customer base or what an institution contributes by joining.

No data protection agreement, retention schedule, subprocessor list or deletion commitment was located. The platform ingests complete borrower packages including financial statements and tax records across dozens of institutions, and its embedded finance arm extends that to end customers of third-party platforms, with no handling terms published for either path.

No attestation, certification, trust centre or enumerated framework was located. Dozens of banks including a large regional institution have completed supplier assessment before routing borrower packages through the platform, and a core provider has admitted it to its own ecosystem, so assurance exists privately and none of it is published.

No regulator, statute or programme rule is named. Government-guaranteed lending is a core product line with its own eligibility and documentation requirements, and compliance is described as built in without indicating which requirements are covered or how the platform tracks rule changes that determine whether a guarantee survives.

The access argument is economic and clearly stated: underwriting a small-dollar loan can cost as much as a larger one, so banks limit their focus on the segment even where borrower demand and community impact are significant, and removing that cost lets them approve more small businesses. The company frames the outcome as ensuring access to fair capital and describes its embedded programmes as fair and scalable. Held at B because no fairness testing, approval analysis by borrower group or measured access outcome is published, and fairness is asserted as an intention rather than evidenced.

No guarantee, indemnity or correction process was located. The small business applicant is the affected party and is unaddressed: nothing describes whether a borrower learns that agents assembled their credit assessment, how a misread financial statement is corrected, or what recourse exists where an automated workflow produces a decline in minutes rather than weeks.

No base model, provider, hosting arrangement or subprocessor is identified for the agents. One data partner is named for small business credit analytics, which is more than nothing, and the bureau, banking and tax data sources feeding spreading and underwriting are otherwise unnamed, so a bank cannot document the third-party model and data dependencies its own supervisors expect it to understand.

Six distribution and integration partners are named, and the significant ones reach institutions at platform level rather than individually: a lending-as-a-service listing on a major customer platform's financial services cloud, selection by one of the largest core banking and payments providers to bring these capabilities to its own client base, a card network's commercial lending partner programme, a consortium of more than 75 member banks, a commercial lending distribution partner and a small business credit analytics provider. The platform deploys alongside a lender's existing systems rather than forcing replacement, which is what makes that reach practical.

No hosting provider, region selection, residency commitment or private deployment option was located. The company operates a research centre in a second country while serving United States banks, which makes processing and access location a question a community bank would raise during vendor review, and it is not addressed.

No pricing, packaging or basis of charge was located. The company's central argument is that underwriting a small loan costs as much as a large one, which makes its own cost per origination the number that determines whether the economics it describes actually improve, and none of it is published.

Product coverage is genuinely broad within commercial lending, spanning small business and government-guaranteed loans alongside commercial and industrial, commercial real estate, construction and specialty products, with a separate embedded finance arm letting platforms launch credit products against a bank network. Buyers are community and regional banks nationwide. Held at B because coverage is a single country and a single institution class, with no international or large-bank presence evidenced.