Inside PolicyBazaar’s AI overhaul to reduce disclosure errors, maintain cost discipline
PolicyBazaar began investing in AI as early as 2019, training its own automated speech recognition engine instead of relying on cloud vendors.
PolicyBazaar is pushing deeper into proprietary artificial intelligence as the online insurance marketplace tries to automate underwriting, boost claim approvals, and strive to meet rising customer expectations in India’s fast-digitising protection market.
The Gurugram-based firm began investing in AI as early as 2019, training its own automated speech recognition engine instead of relying on cloud vendors, CTO Saurabh Tiwari said at YourStory’s flagship startup-tech summit, TechSparks 2025, in Bengaluru.
In a fireside chat, Tiwari explained that transcribing every call became essential once the insurtech company began handling millions of inbound and outbound conversations. With an in-house automatic speech recognition (ASR) software, PolicyBazaar can analyse sales pitches, disclosures, objections, medical questionnaire responses and complaint patterns at scale.
That data feeds quality audits, coaching models for agents and AI systems that flag risky disclosures or incomplete information during the purchase process. It also lets the company track whether agents are sticking to compliance scripts in products where misselling is a regulatory flashpoint.
“It transcribes all the calls, and with the transcriptions, we have been able to build many important assets within the company, which helps us in raising the productivity and efficiency of our contact centre, as well as the customer journey. It has been a core of our DNA right now,” Tiwari added.
The ASR is also integrated with real-time assistance tools that prompt agents with next-best questions, accurately capture disclosures, and reduce back-and-forth for policy issuance. Across underwriting-heavy categories like term insurance and health, a reduction in disclosure errors directly improves issuance rates and future claim acceptance.
The other reason the PB Fintech subsidiary built the ASR was cost discipline. With 12 crore recorded users, millions of policies and more than 6 crore historical interactions, outsourcing ASR would have been financially absurd, the CTO noted. “At a scale where we operate, the economies of scale will not make sense for us not to build our own”, he said.
Tiwari said the company’s technology overhaul began more than a decade ago, when it dismantled a single monolithic codebase and replaced it with a modular architecture that allowed product lines such as health, life, and motor insurance to operate semi-independently.
The shift accelerated execution as each category could modernise its own stack without relying on a central team.
Despite the automation push, Tiwari said human support remains indispensable during moments of distress. “People want to speak to someone whom they can trust, and a robot talking at that spot will not make sense,” he said. PolicyBazaar uses AI in the background for summarisation and workflow automation, but claims servicing still requires a blend of technology and live agents.
The company expects AI to become more embedded across teams as generative tools mature. “In a year or so, we’ll have AI engineers plugged into every board… most engineers will not be software engineers, they will be AI engineers,” the executive said.
PolicyBazaar says its next phase of investment will focus on service credibility, an Achilles’ heel for India’s insurance market. “If I am disclosing a disease, then I must get a claim. And that trust we have to build,” Tiwari said.
Edited by Kanishk Singh