The Collections Cost Problem

A mid-size NBFC with ₹500 Cr AUM and 50,000 active loans spends ₹1–1.5 Cr monthly on collections — field officers, call centre agents, digital contact platforms, and legal recovery. This cost runs regardless of recovery performance, making collections both the largest variable cost and the most critical determinant of NIM (net interest margin).

The AI Collections Architecture

Predictive Default Scoring

AI models trained on payment history, loan utilisation patterns, bureau updates, and behavioural signals identify accounts that are 30–60 days away from default — before they show any delinquency. Early intervention on predicted defaults (relationship call, temporary EMI restructuring, payment flexibility for documented hardship) prevents defaults that cure poorly even with collection effort.

NBFCs using predictive default scoring consistently reduce NPA formation rates by 20–30% — not by collecting better after default, but by preventing defaults that data signals are coming.

Automated Early-Stage Collections

Accounts that miss a single EMI — Stage 1 delinquency — respond well to automated digital contact. WhatsApp payment reminders with convenient payment links, IVR payment options, and scheduled callback requests handle 60–70% of early-stage overdue resolutions without field officer involvement. Cost per resolution drops from ₹300–500 (field visit) to ₹15–30 (digital contact).

Prioritised Field Collections

AI collections prioritisation — ranking field officer portfolios by recovery probability, optimal contact timing, and resolution value — improves field officer productivity 30–40%. Officers spend time on accounts where their intervention adds value, rather than working through portfolios sequentially regardless of recovery probability.

RBI Compliance in Collections

RBI's fair practices code and collections guidelines require documented contact timing, contact frequency limits, and mandatory customer grievance handling. Automated collections workflows enforce these requirements systematically — reducing the regulatory risk that manual collections creates.