HDFC Life · Life insurance · 2026
Renewal collections and inbound service for a regulated insurer, in eleven languages
Call policyholders whose premium is due and get them to pay. It sounds simple. Get the premium amount or the grace date wrong, though, and you lose the payment and create a compliance problem at the same time.
The problem
A human collections agent closes a call in roughly ninety different ways, and each one implies a different next action. A promise to pay with a date is not the same as one without. An active mandate and an inactive one sound identical on the call and behave completely differently afterwards. "Already paid" has to stop the calling at once. "Language barrier" is not a retry, it's a re-route.
What I did
- Spent most of my time with the collections and business teams, not IT. We went through every disposition and reduced each one to two questions: is this customer still callable, and when.
- Wrote the recycle rules that follow from that. Call date plus two, three, five or seven days depending on the reason. Promise-to-pay date plus one. Same-day recycle for calls closed before 1 pm. Mandate-driven waits before a re-debit.
- Specified the 100+ policy fields the agent personalises against: premium due date, grace end, revival quote, collection bucket, preferred language.
- Built the bulk lead-push and disposition write-back APIs so the client's system of record stayed the source of truth.
The second use case: inbound
Once renewals were running, the obvious next question was what happens when the customer calls back. We put an inbound agent behind HDFC Life's existing contact-centre routing. It picks up with the context of the outbound call already loaded, so nobody has to repeat their policy number or explain why they're calling.
- Handles the high-volume service queries directly: premium due and payment, fund value and NAV, policy status, surrender and revival.
- Hands off to a human with the transcript and intent attached when the query needs one.
- Every call is tagged by intent, which gave the service team its first clean view of why customers call.
How it works
Impact
Monthly interactions
Their business team now runs its own collections reporting off the agent: policies, premium collected and month-on-month persistency. The thing I took away is that the model was the easy part. The decision layer around it came from sitting with the people who had been making those calls by hand for years.
Stack