Use Cases
One customer, from first chat to claim. Measured 11 Nov 2026.
The track
- 1Sells the policyAcquisitionpolicybazaar.ai: Rohan buys Ebony Health Silver for his family, in one conversation.
- 2Checks the saleAdvisor call QAReads the advisor's call against the product sheet and flags mis-selling or wrong promises.
- 3Pays the claimClaims processingReads the hospital bill, discharge summary and payment proof, checks them, and decides.
Each step, on two stacks
| Step | Policybazaar models | Gemini 2.5 Flash |
|---|---|---|
| 1. Acquisition 31 test chats, on the Evals page | 30/31 right ₹0.13 a chat · 3.1 s | 22/31 right ₹0.99 a chat · 4.8 s 3.1 Pro on the hard turns |
| 2. Advisor call QA 2 scenarios | 2/2 right ₹0.01 a call · 2.0 s | 2/2 right ₹0.18 a call · 4.3 s |
| 3. Claims processing 2 scenarios | 2/2 right ₹0.02 a claim · 2.7 s | 2/2 right ₹0.09 a claim · 3.7 s |
Monthly model bill, at Policybazaar's volume
| Activity | A month | Today, on Gemini | Open models |
|---|---|---|---|
policybazaar.ai chats Today on 2.5 Flash and 3.1 Pro | 5.1 L chats | ₹5.1 L | ₹68,506 |
Advisor call QA Today on 2.5 Flash | 6.3 Cr calls | ₹1.2 Cr | ₹9.4 L |
Claims documents Today on 2.5 Flash | 2.1 Cr pages | ₹6.3 L | ₹1.5 L |
| Total, a month | ₹1.3 Cr | ₹11.5 L |
With open models
91% lower
₹11.5 L a month, against ₹1.3 Cr today, at the same quality on these tests.