Skip to content
Telesurety
Platform · 04 · Coverage comparisonPartly live

Quotes side by side, the difference in plain words.

The offers you select open in one table, coverage by coverage, and a difference summary is written in plain sentences to read to the customer. Common coverages meet on one row by rule today; a common model for the whole coverage language is on the roadmap.

Request a demo

Highlights

Insurers connected via web services
14
A quote request goes at once to every insurer connected for the chosen branch, and prices land on one screen as they arrive.
Side-by-side coverage table
Live
In the operations panel, one table per branch with a difference summary for the customer underneath.
Common coverage ontology
Roadmap
Will map motor own-damage, motor liability, home and DASK coverages to one model, with every new mapping approved by an expert.

The operations panel sends one quote request to every insurer connected over web services for the chosen branch, ranks the prices as they arrive and lines up the offers you pick, coverage by coverage. Every premium and coverage value comes from the insurer's own response. A model that maps all coverage wording to one common language is on our research agenda.

How it works

The side-by-side coverage table, the difference summary and rule-based matching of common coverages are live in the operations panel. An ontology that maps every coverage to one common Turkish model is on the roadmap and builds on the third problem in our research agenda.

  1. 01 / 05

    One request, live prices

    Customer details and the vehicle, home or people to insure are entered once. The top of the screen tracks how many insurers have responded, how many gave a price, how many declined and how many timed out.

  2. 02 / 05

    Ranking and highlights

    Rows hold their place while prices are still arriving. Once every insurer has answered, the list starts from the lowest price, and you can re-sort by response time or by insurer. The lowest price is marked “Best price”, and a separate “Recommended” flag arrives with its reason.

  3. 03 / 05

    Open the coverages side by side

    Tick offers in the list to open the comparison table. If you tick none, the cheapest offers appear automatically. Each branch gets its own table, laid out section by section from price and payment to deductibles and scope.

  4. 04 / 05

    Explain the difference

    The difference summary under the table compares each offer with the next cheaper one. It states how much more it costs, in amount and percent, and which coverages are better or weaker, in plain sentences you can read to the customer.

  5. 05 / 05

    Record the policy from the chosen offer

    Where the insurer's PDF service is connected, you preview, download or open its quote PDF in a new tab. The panel checks that the prices are still valid and lists any missing customer or vehicle details; an expired price cannot be recorded. The policy record is created from the chosen offer, along with the customer or vehicle record if one does not exist yet.

  6. Illustrative. Names and values shown are examples. Step by step. One request, live prices, Ranking and highlights, Open the coverages side by side, Explain the difference, Record the policy from the chosen offer.

Operations panel

Live

See what a cheaper offer leaves out.

The comparison table groups rows into sections and writes each offer's reported value on the same line. When an insurer does not report a coverage, the cell says “not reported by the insurer”; when no coverage details arrive at all, the table is limited to price and payment rows.

  • Sections: price and payment (cash premium, longest installment plan, installment surcharge), liability limits, sums insured, repair, extra services, health coverage, deductibles and scope
  • Favourable and unfavourable marks appear only where the meaning is clear: original parts and authorised service are favourable, equivalent parts unfavourable; no deductible is favourable, a deductible unfavourable; no replacement car is unfavourable
  • On differing numeric rows, such as the İMM limit, the highest value is marked “best value”
  • A “Show differences only” switch hides rows that are identical across all offers
  • In the price list, a distribution bar shows the lowest price, the average and the spread between lowest and highest; original-parts and long-installment filters narrow the list
  • Insurers that gave no price are listed separately as declined or timed out, with the reason and response time

Comparison PDF

Pilot

The comparison PDF reaches the customer unchanged.

For a quote request from the customer app, staff upload a price comparison PDF that lists insurers' gross premiums. The panel reads PDFs in the supported template, turns each row into an offer and accepts only the amounts it is sure of. The reader runs in the operations panel; the customer app this flow depends on is in pilot.

  • Extracted fields: product, coverage period, who prepared the PDF, when it was prepared and how long it is valid; for each row, the insurer and the gross premium
  • The ID and plate number are checked against the customer's request
  • A row with an ambiguous amount is left out and raised to staff as a warning; a row whose insurer cannot be matched blocks sending until staff pick the insurer from the list
  • Staff review every warning before anything is sent, and cannot edit the prices that were read
  • Offers are published to the customer app with the same PDF attached; when a customer phones or e-mails their choice, staff record it

Common coverage language

Partly live

Common coverages already share a row.

Insurers give the same coverage different names. Today a fixed rule list converts common coverage types to one shared name so their rows line up across insurers. A complete, expert-approved coverage ontology is on our research agenda.

  • Live today: rules map common coverages such as the different spellings of voluntary third-party liability (İMM), replacement car, glass, legal protection, earthquake, flood and assistance; anything the rules do not recognise keeps the insurer's own wording
  • Roadmap: a Turkish coverage ontology for motor own-damage (kasko), motor liability (trafik), home and compulsory earthquake (DASK) insurance, built with vector similarity and language-model re-ranking
  • New or unclear wording does not enter the model without expert approval: the model suggests, an expert decides
  • Proposed success measures are mapping accuracy and agreement with experts (Cohen's kappa). Results will be benchmarked against today's rule-based readers and an untuned general model, with an insurer never seen in training measured separately
  • In the planned version, every difference will link to the relevant line in the insurer's document, and the same model will feed source-cited search over policy documents

Use cases

From an intermediary's day.

  • 01

    “Why not the cheapest?”

    At a motor own-damage renewal, the customer asks why the recommended offer is not the cheapest one. Staff open the recommended and the cheapest offer side by side and show only the differences. The table shows that the cheaper offer repairs with equivalent parts and has no replacement car, and the difference summary already puts this into sentences staff can read out.

  • 02

    An insurer that did not quote

    Some insurers return no price on a motor own-damage request. The panel lists them separately as declined or timed out, with each reason and response time. Staff can tell the customer which insurer did not quote and why, without guessing.

  • 03

    A quote request from the app

    A quote request arrives from the customer app, which is in pilot. Staff upload the price comparison PDF; the reader extracts the rows, checks the ID and plate against the request and leaves out an amount it is unsure of, with a warning. After reviewing the warning, staff publish the offers, the customer picks one in the app and the policy record is created from that choice.

What it does

  • Live

    Side-by-side coverage table

    Selected offers per line of business, section by section in one table; “Show differences only” hides rows that are the same.

  • Live

    A difference summary for the customer

    Plain sentences that compare each offer with the next cheaper one, ready to read to the customer.

  • Live

    Rule-based coverage matching

    Common coverages land on the same row by rule; wording the rules do not recognise keeps the insurer's own name.

  • Roadmap

    A common coverage language

    Will map motor own-damage, motor liability, home and earthquake coverages to one model; a new or ambiguous phrase will not enter the model without expert approval.

  • Roadmap

    Differences with sources

    Every difference will link to its line in the insurer's document.

What AI does · What you decide

What AI does

Today matching and the difference summary run on rules. In the planned model, AI will suggest matches for new coverage phrases, and an expert will approve them.

What you decide

Which quote to recommend and every premium decision are yours.

Details

Technical details

Data sources
Insurers' web service responses; for quote requests, the price comparison PDF uploaded by staff
Premiums and coverage values
Taken from each insurer's response; the platform does not calculate premiums and AI does not decide these values
Table sections
Price and payment, liability limits, sums insured, repair, extra services, health coverage, deductibles and scope
Coverage alignment
Rule list for common coverage types is live; the ontology for motor own-damage, motor liability, home and DASK is on the roadmap
Where it runs
Quote production in the operations panel (live); comparison PDF reading for quote requests, and the comparison screens in Quick Quote and the customer app (pilot)
Quote PDF
Preview, download or open in a new tab where the insurer's PDF service is connected

Questions

Frequently asked questions

Does the platform calculate premiums or coverage values?

No. Premiums, installments and coverage values come from the insurer's own response, and the screen states this. The panel only calculates the price and installment differences between these values and converts a foreign-currency premium to lira at the rate the insurer reports. Our research plan keeps premium, sum insured and commission out of AI decisions.

How is the “Recommended” flag chosen?

It goes to an offer priced within a set range of the cheapest that has no unfavourable coverage and the most favourable ones. The flag explains why it was chosen. When the cheapest offer already meets these conditions, no separate flag appears. Staff decide which offer to recommend to the customer.

How do rows line up when insurers name coverages differently?

Today fixed rules convert common coverage types to one shared name. A coverage the rules do not recognise appears on its own row in the insurer's wording. The common coverage model, which will use a language model and expert approval, is on the roadmap and is not live.

Can I give the customer the comparison as a document?

In the operations panel you preview, download or open each insurer's own quote PDF where its PDF service is connected; the panel's comparison screen does not produce a PDF of its own. A printable comparison that can be saved as PDF exists today in the Quick Quote pilot, and it leaves out internal data such as ID numbers. The side-by-side comparison screen in the customer app is also in pilot.

What happens when the PDF reader is unsure of an amount?

It drops that row and shows staff a warning. A misread decimal separator can change a price many times over, so the reader accepts only amounts it is sure of. If no offer table can be read from the PDF, sending is blocked.

Other parts

Back to platform