Insurance marketplace, IndiaLive
A private document engine turns policy PDFs into checked fields on the client's servers. Field accuracy is measured on a hand-verified test set.
Delivered by our founding team at a previous firm. Clients anonymised under NDA.
Every quote and every policy was retyped from PDFs by hand, twice. A policy arrived on WhatsApp, email or chatbot; ops retyped every field, priced it across four or five insurers, then retyped again to issue. Rule-based tools caught ~12 fields; generic AI hit 83% on ~40 of 180+. Regulators want the data kept in India.
A private document AI engine on the client's own servers. Six steps on every document: receive, identify the policy type and load its rulebook, mask names and IDs before any model reads it, extract 180+ fields, verify every field against the rules, route failed checks to a person. An open-source model replaced per-call public AI fees.
About 98% of fields read correctly on a hand-verified test set. Turnaround 80% faster. Ten to twenty percent more quotes became issued policies. AI cost per document fell 78%. No PII went to public AI. No new hires: the same team moved from typing to reviewing exceptions. Started with motor, now extended to life, health and home.
Delivered by our founding team at a previous firm. Clients anonymised under NDA.
Every quote and every policy was retyped from PDFs by hand, twice.
Receive, identify and mask personal data.
Extract fields and apply policy rules.
Failed checks go to a person.
Live. Delivered by our founding team at a previous firm. Clients anonymised under NDA.
WhatsApp, email, chatbot
Identify policy type and load its rulebook
Mask names, phones and IDs before any AI reads it
Extract 180+ fields
Verify every field against format and policy rules
Route failed checks to a person
Failed checks go to a person.
Clean records land in your system.
Tell us about one costly process. A founder replies within two working days.
Start with the problem. Agree the right next step together.
For example: