Skip to content

Evidence network

How NRD works

NRD is the layer of organised, verified evidence between a claim and a decision. Every step can be checked. Nothing is guessed without a source.

The evidence network, step by step

Seven stages, all logged. Each stage either produces something you can check, or hands the file to a person before moving on.

  1. 01

    It all lands in one place

    Records arrive via WhatsApp, a photo from the field, email or an insurer feed. Each file is timestamped, given a unique reference, and saved to write-once storage write-once storage — records can’t be changed once saved in the AWS Mumbai region. Once it's saved, it can't be changed.

  2. 02

    Read the text — with an honest gate

    Printed and typed documents go through automatic text recognition automatic text recognition , which gives each page a confidence score — how sure it is that it read the text correctly. Handwriting and regional scripts (Hindi, Marathi and others) often score low. Anything below the pass mark is sent to a person to check before it goes any further. We never quietly pass low-confidence text downstream.

  3. 03

    Pull out the facts

    From the confirmed text, Netra (NRD's document-reading engine) pulls out clean, typed fields: diagnoses, ICD-10 the standard medical diagnosis codes codes, the treating doctor, hospital and dates, itemised charges, witness names and statements, and policy / FIR references. Every field points back to the page it came from.

  4. 04

    Quality check

    Thirteen fixed checks run on every set of records — fixed rules, not an AI guess: are all pages present, is each page legible, are there duplicates, do the dates run in order, does the bill total match the line items, and are PAN / Aadhaar Indian ID numbers (sensitive data we detect and protect) numbers detected and protected.

  5. 05

    Find what's missing

    Netra compares each claim against a checklist of documents expected for its type (head injury, fracture, soft tissue, death) and returns a plain list of gaps — missing medicals, no FIR, no discharge summary — that a surveyor or panel lawyer can act on.

  6. 06

    Summary

    Netra puts together a clear summary: the treatment timeline, billed vs paid totals, a suggested reserve based on the extracted figures, and a short account of how the injury links to the accident — every line traceable to a document.

  7. 07

    Seal it with a certificate

    Every document in the final claim file gets a digital fingerprint (SHA-256) a digital fingerprint of a file , a timestamp in IST, and is locked in write-once storage write-once storage — records can’t be changed once saved so it can never be altered. The resulting certificate is court-ready under the Bharatiya Sakshya Adhiniyam, Section 63.

How much to trust each output

We sort every output by how it was produced and how much weight you can give it. Honest labelling is the whole point.

Tier How produced Examples Weight
Calculated Worked out by fixed rules from the numbers on the page. No AI involved. MACT compensation (Pranay Sethi formula), court fees, limitation period, date-order checks, file-fingerprint checks. Full trust — you can audit the formula and get the same answer every time.
Read from the document Read straight from a document, with a score for how sure the software is. If it isn't sure enough, a person checks the source before the field is used. ICD-10 codes, treatment dates, itemised charges, provider names, regional-script text, handwritten content. Trust, with a source — every field points back to its page. Regional-script and handwritten text always gets a human check.
AI-estimated A signal from an AI model that helps a person decide — never the decision itself. Not used for anything binding. Always labelled on screen. Reserve-adequacy range, fraud-indicator score, likely-to-defend signal, settlement ceiling range. A hint, not a decision. A claims officer reviews and confirms before acting.

The human check on low-confidence text isn't a disclaimer — it's how the pipeline is built. Low-confidence documents never move forward without a person reviewing them.

Glossary

MACT
Motor Accident Claims Tribunal — the Indian forum that adjudicates compensation for motor-accident injury and death claims.
FMF / FMR-1
IRDAI Fraud Monitoring Framework and its FMR-1 return — mandatory fraud-monitoring reporting for Indian insurers since 1 April 2026.
DPDPA 2023
India's Digital Personal Data Protection Act — governs processing and residency of personal data.
CERT-In
India's Computer Emergency Response Team directions — incident reporting and log-retention requirements.
BSA S.63
Bharatiya Sakshya Adhiniyam, Section 63 — admissibility of electronic records as evidence.
Pranay Sethi
Supreme Court precedent setting the standard formula tribunals use to calculate MACT compensation.

See the evidence network on your claims.

We'll scope a pilot on a slice of your motor/MACT book.

Request a pilot
Held in India Walled-off per insurer Court-admissible record