Boosty

CASE · INSURANCE

From plate to issued policy, without re-typing a single field.

From license plate to issued policy without re-typing data: AI document validation, issuance via API and exceptions routed to an analyst. We built the complete RCV policy activation flow for an insurer: 8-line quoters, AI document validation, and direct issuance against their core REST API.

8-line quotersAI doc validationInsurer APIMake webhooks

THE PROCESS BEFORE

Activating an RCV meant re-typing, waiting, and forwarding emails

100% manual issuance

Every RCV policy was activated by hand: look up the vehicle, re-type data in the core system, wait.

Documents that bounced

ID, invoice, and certificate of origin arrived with a field mismatch and went back by email.

No traceability

Nobody knew which step an activation was at or why it had stalled.

HOW WE UNDERSTOOD IT

The bottleneck wasn't the core: it was validating documents

We mapped the activation end-to-end and found that the insurer core already exposed a REST API capable of issuing. The real problem was upstream: data arrived with a field different from the supporting document, and every discrepancy meant an email back and forth.

The opportunity: place an AI validation layer between data capture and the API, and chain the entire process into a single traceable flow that starts with a license plate and ends with a policy number.

THE FULL FLOW

From license plate to issued policy, with an analyst for the exceptions

The entire activation of an RCV policy in nine chained stages. Each stage validates before letting the next one through: that is why issuance reaches the insurer core clean.

01

Vehicle license plate

The operator enters the plate

The flow starts with a single piece of data: the license plate. There are no long forms ahead yet — just the entry point that triggers the whole RCV policy activation process.

AI DOCUMENT VALIDATION

The document comes in, the AI reads it and cross-checks it against the form

Eight document types validate themselves. The AI extracts the fields, compares them with what was captured and, if something does not line up, returns a note in plain language — not an error code.

incoming_documentID document
Name· · ·
ID number· · ·
Date of birth· · ·

receiving document…

cross-check vs form
Name
—
ID number
—
Date of birth
—
Document accepted

DEVELOPMENT SCOPE

What we built

  • 01Multi-step quoters for 8 lines: Individual health, Group health, Auto, Travel, Home, Life, and Business.
  • 029-stage RCV activation flow: plate → factory inventory → serial → data → documents → AI validation → API → policy → webhook.
  • 03AI validation of 8 document types with human-language observations.
  • 04Public inventory upload by factory with column mapping, deduplication, and margin-based pricing.
  • 05Hard capture rules: emails ≤70 chars without accents, 4-digit economic activity code, company name without commas.
  • 06Admin panel: policies with CSV export, bulk download and reprocessing; catalogs synced with Monday.com; download audit trail.
  • 07Integration with the insurer REST API and webhooks to Make.com.

OUTCOMES

What changed day-to-day

Issuance without re-typing data

The RCV policy is issued end to end without anyone re-typing data in the insurer’s core; exceptions go to an analyst with the reason written down.

Automatic validation of 8 documents

ID, certificate of origin, invoice, title, license, tax ID, income tax certificate, and bank reference validate themselves.

Errors show up before issuance

Discrepancies are caught before issuance, with an observation that says what to fix; whatever does not match is reviewed by an analyst.

Traceable process

Every activation is visible step by step; reprocessing and download auditing from the admin panel.

YOUR PROCESS IS NEXT

Does your issuance still depend on re-typing and forwarding emails?

If you have a core with an API but the process is still manual, that's exactly what we rebuild. Schedule a 30-minute assessment.

✓
Assessment of your current processes
✓
What we would build: agents inside the system, with policy and rollback
✓
How it would be measured: adoption by role and outcome before and after

No spam. We reply within 24 business hours.