One model of the loan. Shared by every system.

A lending ontology is one governed model of the loan — borrower, application, repayment, and risk, plus how they relate — that your whole stack reads and writes through. Instead of many drifting copies, there is one canonical loan object, and every system and every agent acts on it.

A deterministic core means the same inputs always produce the same outputs. Run a decision twice, get the same answer twice — no hidden state, no drift.

The entities the model resolves to

Governed AI doesn’t just act safely on the canonical object — it acts better. One resolved truth makes the agents faster, lets them reason across the whole loan, and keeps every move inside policy.

Because every application, repayment, and recovery writes back to the same loan object, agents can test hundreds of pricing and underwriting rules against real outcomes — in the time a risk team usually ships one.

The outcome this is built to produce is profit per rupee disbursed — lower delinquency and better-priced risk as the book learns from itself. Lokta is pre-customer and co-designing with its first partners: this is the design intent, not a result being claimed.

One governed model, one audit trail — and capabilities a fragmented stack can only do in pieces. Each pairs with the control that makes it safe.

We did not arrive at lending architecture recently — we built the rails the industry runs on.

The canonical loan object sits under what you already run — LOS, LMS, and AI loan servicing — and gives them one loan to agree on. It boards from your existing systems; you adopt modules one at a time, not a rip-and-replace.

What is a lending ontology?

A lending ontology is one canonical model of the loan — borrower, application, repayment, and risk, plus the relationships between them — that the whole lending stack reads and writes through. Instead of a separate copy of the loan in every system, there is a single canonical loan object that governed AI and people act on, with every change logged.

Why do LOS, LMS, and core banking systems disagree on loan data?

They disagree because each keeps its own copy of the loan and updates it on its own schedule. The loan origination system, loan management system, core or general ledger, and collections tool all model the same loan slightly differently, so the records drift and have to be reconciled — which is slow and, in regulated lending, an audit liability.

What is a deterministic core in lending?

A deterministic core means the same inputs always produce the same outputs — run a decision twice and you get the same answer, with no hidden state. It is reproducible, idempotent, and event-logged, which is what makes it safe to move regulated money on and to certify after the fact.

What does “audit-by-design” mean for a lending platform?

Audit-by-design means the audit trail is produced by the system as it runs, not assembled afterward to satisfy a regulator. Every state change lands on one append-only log in order, so the record a compliance officer certifies is the same record the system acted on.

What is “profit per rupee disbursed”?

Profit per rupee disbursed is the outcome the architecture is designed to produce — better-priced risk and lower delinquency as the book learns from its own lending. Because every repayment and recovery writes back to one canonical model, agents can test and promote better policy under maker-checker, and the book sharpens itself the more it lends. Lokta is pre-customer, so this is the goal the design targets, not a result being claimed.