Underwriting Agent

The Problem

Every credit analyst is reinventing the same memo from scratch.

In SME lending, a credit analyst spends significant time on tasks that are fundamentally repetitive and data-driven: pulling bureau reports, reading bank statements, summarising financials, mapping red flags against policy rules, writing a credit memo. In many mid-size NBFCs and fintechs, this process takes 2–4 hours per case.

The output quality depends heavily on the individual analyst. A senior analyst brings pattern recognition built from hundreds of cases. A junior analyst misses nuances. There is no structural way to enforce consistency — every memo reflects the individual, not the institution's credit policy.

Inconsistency creates downstream problems: credit committee reviews take longer because memos aren't structured the same way. Approvals get delayed. Borderline cases get resolved by gut feel rather than structured analysis tied to the policy document.

The irony is that the underlying rules are usually well-defined — the credit policy exists — but there is no mechanism to apply it uniformly at the memo-writing stage, case after case, at scale.

The Solution

Structured analysis on every case. Automatically.

An autonomous underwriting agent that takes structured inputs — applicant financials, bureau data, bank statement summary — and produces a standardised credit memo with risk flags, scoring summary, and a recommendation (approve / refer / reject) with rationale. The agent follows a defined credit policy ruleset and applies LLM reasoning to interpret edge cases that don't fit cleanly into the rules.

The output is designed for the underwriter to review and override, not to replace them. The goal is to cut memo prep time from hours to minutes, and to ensure every case gets the same baseline of structured analysis regardless of who handles it — so the analyst's judgment goes into the right place: the edge cases, not the boilerplate.

Key design decision: the agent surfaces its reasoning explicitly. Every risk flag is tied to a specific data point. Every recommendation includes the policy clause it maps to. This makes the output auditable and helps the credit committee trust the memo rather than re-verifying it from scratch.

Product Screenshots

Interface