Case study · Insurance · Agentic AI
How India's largest general insurer built an agentic AI advisor that answers in 5 seconds.
General insurance · Advisor intelligence
3,000+ advisors were searching for answers mid-call. We embedded an agentic assistant that acts like a real-time sales supervisor-live data, live reasoning, zero context-switching.
5 sec
Avg query response time, down from 3 days
20%
Reduction in support ticket volume
30%
Increase in customer satisfaction
3,000+
Advisors on one consistent knowledge standard
The opportunity
Advisory accuracy that depended on every advisor knowing everything.
One of India's largest general insurers runs a distributed network of 3,000+ advisors across Health, Motor, Travel and SME products. Sales performance is determined in the moment of the conversation-which meant advisory quality was entirely dependent on what each individual advisor could recall under pressure.
There was no centralised support system. Advisors relied on PDFs, emails and peer networks during live customer interactions. As product complexity grew, query resolution time grew with it-delaying responses, breaking conversations and creating inconsistency at scale.
No centralised advisory system
3,000+ advisors operated independently, with no single source of real-time product or policy information.
High TAT during customer engagement
Query resolution took days-a direct cost to conversion at the most critical moment of the sales call.
Inconsistent objection handling
Sales quality varied entirely by individual capability, with no standard messaging or escalation framework.
Growing product complexity
Health, Motor, Travel and SME portfolios kept expanding, and static tools could not keep up with product nuance.
The impact
Every advisor now has a live AI supervisor in the workflow.
5-second query resolution
Down from 3 days-the agentic assistant dynamically retrieves live policy data, renewal notices and incident status, returning a structured, sales-ready response in real-time.
20% fewer support tickets
Advisors resolve routine queries at the point of conversation, freeing support teams for complex cases rather than volume.
30% higher customer satisfaction
Faster, consistent and accurate responses during live interactions created a measurably better customer experience across the network.
One knowledge standard, 3,000+ advisors
Uniform messaging, objection-handling and upsell positioning now operate across every advisor, not just the experienced ones.
The stack
What it runs on.
- Agentic reasoning layer
- Claude (Anthropic), orchestrating multi-step reasoning across live data sources to compose sales-ready responses
- Real-time data orchestration
- Dynamic API calls to policy PDFs, renewal notices, policy issuance status and incident management systems on every query
- RAG knowledge system
- Retrieval-augmented generation grounded in current product documentation, eliminating hallucination and ensuring compliance
- Embedded advisor workflow
- Integrated directly into the existing advisor interface-no tool switching, no external dependency, no disruption to the live conversation
The road ahead
The assistant is live. Next is deepening what it knows.
With real-time orchestration, agentic reasoning and embedded delivery in production across 3,000+ advisors, the focus moves to expanding coverage and closing the loop on revenue outcomes.
- 01
Proactive upsell recommendations from customer policy history
- 02
Deeper integration across renewal, claims and onboarding workflows
- 03
Advisor-level performance measurement tied to conversion outcomes
Talk to us
Bring one process. Leave with an outcome architecture.
30 minutes on one function, mapped to the architecture behind it. No slides.