The Lokta lending glossary

A Loan Management System (LMS) is the system of record for the post-disbursal life of a loan: schedules, repayments, interest accrual, charges, NPA classification, restructuring, write-offs, and the audit log behind every change.

Unlike a Loan Origination System, which underwrites and books the loan, an LMS owns it from disbursement to closure. A modern LMS exposes its data and rules to AI workflows through a deterministic core; a legacy LMS treats every change as a vendor configuration request, which is why most lenders eventually replace it.

Lokta's LMS product is Lokta Core →

An LMS RFP (Loan Management System request for proposal) is how a lender evaluates and selects an LMS vendor — scoping requirements, scoring responses, and stress-testing servicing, collections, accounting, audit, security, and AI governance before committing.

A real LMS RFP is run by an evaluation committee, and each member opens a different document: the technology evaluator works the functional requirements; risk and compliance work the security and AI-governance questions; finance and procurement work the scorecard and migration plan. The output of a good RFP is not a feature checklist but a stress test of how a platform behaves under the operational, regulatory, and integration realities of a real loan book.

Use the vendor-neutral LMS RFP toolkit →

A Loan Origination System (LOS) is the system that takes a borrower from application to first disbursement: lead capture, KYC, credit decisioning, document collection, agreement generation, and funding.

It hands the loan to the LMS at booking. The LOS is where speed and conversion matter; the LMS is where accuracy and audit matter. AI-ready origination treats every decision — eligibility, pricing, condition setting — as a workflow with policy guardrails and a full audit trail, not a black-box model output.

Lokta's LOS product is Lokta Originate →

A Lending Service Provider (LSP) is a regulated entity that operates loan products in partnership with one or more balance-sheet lenders — typically a bank or NBFC.

The LSP owns the borrower experience, distribution, and often the technology stack; the partner lender owns the balance sheet and the regulatory licence. LSP-led models include co-lending, BNPL, embedded finance, and the digital lending arms of fintechs. The architectural challenge is multi-partner from day one — tenant isolation, partner-aware reporting, and settlement workflows have to live in the data model, not bolted on later.

Co-lending is an arrangement where two or more lenders fund the same loan in a fixed ratio, sharing risk and return on a per-loan basis.

The model is most prominent in India, where Reserve Bank of India guidelines define how banks and NBFCs can split exposure (commonly 80:20). Operationally it requires per-loan ledger splits, partner-specific MIS, and settlement workflows that reconcile down to the rupee. Lenders running co-lending on a single-tenant LMS spend most of their reporting time in spreadsheets; multi-partner platforms model the partnership as a first-class entity in the data ontology.

Connected lending is the architectural pattern where origination, loan management, collections, and AI-powered servicing run on a single data ontology, a single ledger, and a single audit log — instead of being stitched together from separately purchased systems.

It is the answer to the integration tax: when each layer owns its own copy of the borrower record, every analysis becomes a reconciliation. Connected lending is what makes AI workflows possible at scale, because agents can read and write across the loan lifecycle with consistent state. The unconnected alternative is increasingly incompatible with agentic automation.

Read: What is connected lending? →

A deterministic core is the layer of a lending platform whose outputs are fully reproducible from inputs and policy — no probabilistic models, no LLM calls, no hidden randomness.

Interest calculations, repayment allocation, NPA classification, and the audit log all live here. A deterministic core gives the lender a system of record that auditors and regulators can interrogate. AI layers — workflow assistants, agents — sit above it and call into it; they never replace it. The architectural rule: the deterministic core is the truth, the AI layer is the help.

Read: AI in lending — the deterministic core →

Workflow AI assists a human inside an existing process; agentic AI runs the multistep process end-to-end. Both sit above a deterministic core; neither substitutes for it.

Workflow AI drafts a dunning letter, summarises a borrower file, suggests a hardship category, explains a calculation — the human approves and moves on. Agentic AI picks up a defaulted loan, contacts the borrower, negotiates a plan within policy, escalates exceptions, and logs every step for audit. Workflow AI is a productivity layer; agentic AI is a labour layer. They serve different hand-offs and require different governance.

Read: our AI philosophy in lending →

Audit-by-design is the principle that every decision a lending platform makes — human, workflow AI, or agent — produces a tamper-evident record that a regulator or internal auditor can replay.

It is a property of the data model, not a feature added later. Every state change is timestamped, attributed to an actor (user, workflow, agent), linked to the policy version it was evaluated against, and immutable downstream of approval. Audit-by-design is what allows AI to be deployed in regulated lending workflows; the alternative — bolt-on logging and screenshotted approvals — collapses the moment a regulator asks for a methodology trail.

An AI-ready LMS is a loan management system whose data model, APIs, and policy engine were designed for AI workflows and agents to read and write — not merely for screens and batch jobs.

Three properties qualify: a clean event stream every change is published to; a policy engine AI agents can call without breaching governance; and an audit log that records human, workflow, and agent actions identically. Most legacy LMS platforms fail all three, which is why bolt-on AI features rarely move beyond chatbot-on-FAQ.

Read: why a legacy LMS can't bolt on AI →

A strategic-liability LMS is a loan management system whose architecture, contract terms, and vendor relationship together prevent the lender from making the product, compliance, and exit decisions a board would otherwise expect to make.

The licence fee is rarely the issue. The cost surfaces as loan products that took 18 months to launch, portfolio insights that arrived a quarter late, regulatory exposure no one reconciled, and renewal years that arrived without options. Three or more diagnostic symptoms on the table is a board-level risk, not an IT one.

Read: Every Lender Replaces Their LMS Eventually →

Agentic loan servicing is the operating model where AI agents run multistep servicing workflows end-to-end — collections triage, dunning sequences, hardship handling, complaint resolution, KYC re-verification — under explicit policy guardrails, with every action logged for audit.

Agents pick up events (a missed EMI, a complaint ticket), execute the workflow within the lender's policy, escalate genuine exceptions to humans, and close the loop. It is distinct from a chatbot layer (which only answers questions) and from RPA (which only repeats fixed scripts). Done correctly, it sits on a deterministic core and produces an audit-by-design record.

Lokta's agentic servicing product is Lokta Delight →