Skip to content
Book a consultation

Medical insurance quotes, compared

A product example, not a client story: how Compare handles quotes for UAE medical insurance brokers. It reads insurer quote PDFs, learns each layout once with AI, then extracts later quotes in that layout at no model cost, and builds the side-by-side comparison and a proposal.

Client
Product example: UAE medical insurance brokers
Industry
Insurance brokers
Product
Compare
CompareDemo data
Compare list of group medical quote comparisons by client with draft, sent and won status, demo data

The situation

A broker placing a group medical policy asks several insurers and third-party administrators to quote. Each sends a PDF in its own layout, some of them scans. The broker's team retypes premiums and dozens of benefit lines into a spreadsheet, lines them up, and turns the result into a recommendation the client can read.

What was broken

  • Every quote is retyped by hand, and every insurer lays its PDF out differently.
  • Per-page AI extraction tools charge for every document, including the hundredth quote in a layout they have already seen.
  • Renewals are compared against a fresh spreadsheet, not against the plan the client bought last year.
  • Whether a premium is good is answered from memory, not from the brokerage's own book.

What we built

  • Learn each layout once

    Compare fingerprints every quote layout. The first time it sees a layout, an AI model learns a set of extraction rules for it. Every later quote in that layout is read by those rules, with no model call and no model cost.

  • Checks before a number reaches a client

    Scanned quotes are read with OCR. An optional mode only accepts a value when two independent AI reads agree, under a spend ceiling per batch. Anything that disagrees goes to a review queue for a person.

  • Rules that notice when they drift

    When a saved rule disagrees with a live read, Compare records it. A nightly sweep flags layouts that have drifted and offers reviewers a rebuild.

  • The comparison and the proposal

    Side-by-side comparisons in Excel, HTML and PDF, with the best value per benefit highlighted and a weighted ranking. A shareable proposal goes through maker-checker approval before the client sees it.

  • The brokerage workflow around it

    A client directory, request-for-quote intake linked to the comparison it produces, a renewals engine, renewal comparisons against the plan the client bought, premium benchmarking against the brokerage's own quotes, and win and loss analysis by insurer.

How it runs day to day

An account manager opens an upload for a named client and drops in the insurers' quotes. A quote already seen in that layout is read straight away by its saved rules; a new layout is learned once. Anything uncertain waits in the review queue. The manager checks the comparison, edits the narrative and sends a proposal link for approval. At renewal the new quotes are set against last year's plan, benefit by benefit.

Results

fields read from each quote, across 11 sections from premiums to claims serviceThe product's field schema
80
model cost to read a quote in a layout Compare has already learnedLearned layouts are read by saved rules, with no AI model call
0

What's next

Multi-currency pricing for brokers outside the dirham, and a sign-off record a client can accept as audit evidence.

Related

Product

Solutions

Industry

Start with one job

Opens an email to info@elitestack.ae with your sentence in it. Or WhatsApp us on +971 56 121 1011.