Kotak Bank Voice AI: Scaling a Loan Disbursement Campaign from Zero
Built and iterated a voice AI agent for new-to-credit loan customers through structured call audits and phased optimization.
Bhargava T
AI Product Manager at SquadStack.ai
From their time as
AI Product Manager
SquadStack.ai • 2026
Overview
Bhargava took on the Kotak Bank campaign at the POC stage, with a lead base of four lakh new-to-credit customers, a telephony ceiling of 25,000 calls per day, and a conversion target of 1% on the base.
The Story
Bhargava took on the Kotak Bank campaign at the POC stage, with a lead base of four lakh new-to-credit customers, a telephony ceiling of 25,000 calls per day, and a conversion target of 1% on the base.
The first challenge was prioritization. With a channel constraint, he had to identify the leads most likely to convert before the voice agent could do its work. Once targeting was set, he audited the early calls and found the agent was unprepared for the questions users actually asked: processing fees, interest rates, and why Kotak was worth choosing.
He fixed those gaps in the zero-to-one phase, then moved into the one-to-ten phase, where a new problem emerged. After the offer was generated, users were dropping off at high rates. He audited those calls too, identified the hesitation patterns, and fed that feedback back into the agent.
The methodology was consistent across every phase: audit the calls, find the missing piece, fix it, and measure. Within one month of live operation, the campaign reached 0.6% conversion on the base, with the target of 1% in sight as channel capacity scaled.