Skip to content

Platform

An operations platform built from the ground up with AI.

In today's agency software, staff enter the data and the software stores it. In Telesurety, software reads the document, reaches the insurers and prepares the comparison; staff decide and approve.

Highlights

Operations panel
Live
A real insurance intermediary runs its daily work on it.
Insurers connected by web service
14
Quote production calls every connected insurer that offers the line at the same time.
Field accuracy on real registrations
85.7%
Measured on real documents for the registration-reading model in production.

Telesurety is designed as one operations platform for insurance agencies and brokers: a shared foundation, core services, an AI core and workspace modules. The goal is to replace manual entry with approval in a staff member's day. Work the platform is unsure about is left to a person, on purpose.

Parts

Live: in daily operational use. Pilot: being tried in real work. Partly live: part of it is in daily use, the rest is in development. Prototype: a research experiment. Roadmap: in development, not yet in use.

Architecture

Four layers. Each one carries the next.

  1. 01 / 04

    Shared foundation

    The layer everything else stands on. It decides who can reach which data, which insurer account the work runs under, and what gets logged.

    • Tenant isolationPrototype
    • Users and rolesPrototype
    • Tenant settings and feature flagsPrototype
    • Encrypted insurer credential vaultPrototype
    • Usage meteringPrototype
    • Audit logPrototype
  2. 02 / 04

    Core services

    Every operation that goes to an insurer, plus the reference data the work depends on. Insurer calls follow rules, and AI never chooses what is sent where.

    • Parallel quotes to 14 insurersLive
    • Insurer quote PDF previewLive
    • Policy record from the chosen quoteLive
    • Read-only policy transferLive
    • Vehicle make/model catalogueLive
    • UAVT address lookupLive
    • Insurer-neutral contractPrototype
  3. 03 / 04

    AI core

    The layer for models that read documents, answer from contracts and put coverages into one language. By design, each output carries its source and a confidence score, and personal data stays with in-house models.

    • Registration readingLive
    • Q&A over PDFsPrototype
    • Document intelligenceRoadmap
    • Contract intelligenceRoadmap
    • Turkish coverage modelRoadmap
  4. 04 / 04

    Workspace modules

    The modules that carry the agency's daily work. In the design, each module defines the job it takes on together with the role AI plays in it.

    • Quotes and policiesLive
    • Endorsements, renewals and new businessLive
    • Claims managementLive
    • Customer requestsPilot
    • ReportsLive
    • Team, calendar and HRLive
    • Collections, commission and statement reconciliationRoadmap
    • Central document managementRoadmap
    • Compliance trackingRoadmap
    • Unified inboxRoadmap
    • Broker modulesRoadmap
    • Renewal-loss flaggingRoadmap
  5. Illustrative. Names and values shown are examples. Four layers. Shared foundation, Core services, AI core, Workspace modules.

How it works

Not a chat box bolted on later

AI is the part of each module that does the work. Insurer connections, premiums and permissions follow rules; they are never left to AI.

How it works

It reads. It understands. It prepares. You approve.

  1. 01 / 04

    Reads

    Reads the document and the insurer's response, and scores every field.

  2. 02 / 04

    Understands

    Matches common coverage names with rules and ties each field read to its place in the document; the full common coverage language is on the roadmap.

  3. 03 / 04

    Prepares

    Gets the record, the comparison or the action ready for approval.

  4. 04 / 04

    You approve

    Uncertain fields and every irreversible action wait for your approval.

  5. Illustrative. Names and values shown are examples. Four-step flow: the platform reads the document, matches common coverage names by rule, prepares the work for approval, and the user makes the final call.

How it works

Step by step.

  1. 01 / 05

    A request is opened

    Staff open the request in the web workspace. Quote requests from the customer app also collect in the workspace Requests menu.

  2. 02 / 05

    The platform reads the document

    From a registration photo it reads the plate, chassis number, document serial, make, model, model year and owner. A field it is unsure about, or one that fails a format rule, is flagged for review next to the cropped image from the document.

  3. 03 / 05

    The request goes to the insurers

    The request goes in parallel to the web services of the connected insurers that offer the line. Rules decide what goes to which insurer, and the premium comes only from the insurer's response.

  4. 04 / 05

    Quotes arrive side by side

    Results are compared on one screen. When the insurer supplies a quote PDF, it opens as a preview. For health insurance for foreign residents, a bulk quote can be run from an Excel list.

  5. 05 / 05

    You approve

    The quote you choose becomes a policy record in one step, and the customer, vehicle or property record is created if it does not exist yet. A person always chooses and approves the quote.

  6. Illustrative. Names and values shown are examples. Step by step. A request is opened, The platform reads the document, The request goes to the insurers, Quotes arrive side by side, You approve.

Workspace

Live

From the quote desk to the claim file, one workspace.

Menus sit under General, Insurance, Finance, Corporate and Administration, and each user sees only the menus they have rights to. Policy records cover a wide range of lines, from motor and traffic to life and machinery breakdown.

  • Quote production, quote list, policies, endorsements, renewal tracking and new-business tracking
  • Transfer: the agency's approved policies at a chosen insurer are listed read-only for a date range, and no record is created
  • Claims: an expected-document checklist with progress, SLA and performance tracking, reserve splits, currency conversion, Excel import and export
  • Requests: quote requests from the customer app (at pilot stage), and PDF offers that staff publish back to the customer
  • Calendar, chat, announcements, staff and HR, departments, company credit cards and assigned company vehicles

AI core

Partly live

Document intelligence that starts with registrations.

The AI running in production today is registration reading: an OCR model fine-tuned for vehicle registrations and run in-house. It measured 85.7% field accuracy on real registrations. Document intelligence, contract intelligence and the Turkish coverage model are on the roadmap.

  • Fill from registration: live in quote and customer forms, and nothing is written to disk
  • Values read appear in an editable table and reach the form only when the user transfers them
  • Q&A over policy and quote PDFs: an internal prototype on a locally hosted language model, local embeddings and a vector database, with no data leaving the premises
  • Roadmap: document intelligence for licences, IDs, policies, quotes, invoices and claim documents, with a confidence score and location per field and an approval queue for uncertain fields
  • Roadmap: contract intelligence that cites the relevant clause and says “no provision” when the documents do not support an answer, plus a Turkish coverage model that maps insurers' coverages to one vocabulary

Shared foundation

Prototype

Each agency in its own tenant.

The platform's new foundation is built as a working first-version skeleton. It runs only on sample and synthetic data, with a Connector Hub wired to a simulation provider. It holds no real agency data and connects to no real insurer.

  • Database-level tenant isolation with PostgreSQL row-level security
  • Insurer credentials are stored with envelope encryption; an encrypted block moved to another tenant's record cannot be decrypted
  • Insurer authentication errors are never retried, so the agency's insurer account is not locked out
  • Agency admin, operations staff and viewer roles, with tenant settings and feature flags
  • Usage metering and an audit log for every quote, transfer and document read

Access

One platform, several doors.

  • Live

    Web workspace

    The agency's main screen: quotes, policies, claims, reports and team work run in the browser.

  • Partly live

    Mobile

    The workspace also opens on a phone, with bottom navigation and search in the mobile view. A separate staff app is in development: it uses the workspace API, covers today's summary, registration reading, quotes and policy records, team and notifications, and asks for biometric verification or the device lock before a policy record is created. An assistant used by voice or text is on the roadmap.

  • Pilot

    Customer app

    An app that keeps the customer's policies and documents together: policy wallet, emergency card, step-by-step claim notification and document vault. Its mobile API is in production and the app is at pilot stage. Offering it under the agency's own brand is on the roadmap.

  • Partly live

    API and channels

    Live today: an ERP-connected quote and policy API for a corporate partner's dealer network, and a dealer lead hub that passes leads to a call centre. A general API for agencies and software vendors is on the roadmap.

Use cases

From an intermediary's day.

  • 01

    A motor quote that starts with a registration photo

    A client sends a photo of the vehicle registration. Staff fill the quote form from the registration, and the platform shows any uncertain field with its image from the document. After the check, the request goes to the connected insurers at once, quotes come back side by side, and the chosen quote becomes a policy record.

  • 02

    Reviewing policies approved at an insurer

    Staff pick an insurer and a date range on the Transfer screen. The agency's approved policies at that insurer come back in a read-only list, and each one can be opened for detail. The list opens no record and is cleared when staff leave the page. Staff compare it with the workspace records and enter any missing policy from the Policies screen.

  • 03

    A claim file waiting on documents

    Staff add the expected documents to the claim file; when several insurers share the claim, the list is split by insurer. As documents arrive and are ticked off, progress and the SLA stay in view. At month end, the manager exports the files to Excel and reviews them with drill-down analytics.

A closer look

Seven parts, one operation. The current status of each is stated plainly.

  1. Connector Hub, Partly live
    Partly live

    One connection to 14 insurers.

    Instead of logging into each insurer's portal, the platform carries quote and policy operations to the insurers' web services through one common contract and writes the result to the record.

  2. Document intelligence, Partly live
    Partly live

    The platform reads the document and asks you when it isn't sure.

    Every field comes with a confidence score and its location in the document. A field below the threshold is never written to the record on its own; it goes to human review.

  3. Contract intelligence, Roadmap
    Roadmap

    Ask your contracts. Get the clause with the answer.

    Reads your corporate documents, summarizes them and answers your questions by pointing to the clause. When the documents don't support an answer, it doesn't invent one; it says there is no such provision.

  4. Coverage comparison, Partly live
    Partly 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.

  5. Workspace, Partly live
    Partly live

    All of an intermediary's work, in one workspace.

    The web workspace where an agency runs its day. AI works inside each module as the part that prepares the work, not as a separate window.

  6. AI Assistant and Agents, Roadmap
    Roadmap

    An assistant that will know insurance, understand you and work with your approval.

    The AI Assistant will run on an in-house language model of at least 12 billion parameters. It will understand the requests you type in Turkish and answer from the policy and the agency's own records, with the source. Its agents will use the panel's own endpoints with your permissions to do the work or prepare it for your approval. Every irreversible action will wait on an approval card.

  7. API, Partly live
    Partly live

    The same capabilities, inside your own software.

    Document reading, coverage comparison and insurer connectivity, available through an API without leaving the software you use.

Live: in daily operational use. Pilot: being tried in real work. Partly live: part of it is in daily use, the rest is in development. Prototype: a research experiment. Roadmap: in development, not yet in use.

Illustrative. Names and values shown are examples.

Details

Technical details

Lines in policy records
Comprehensive motor, compulsory traffic, short-term traffic, traffic add-on, compulsory seat accident, green card, excess liability (İMM), legal protection, home, earthquake (DASK), complementary health, private health, life, travel, personal accident, liability, business, home package, electronic devices, machinery breakdown
Insurer connections
Web services of 14 insurers. Parallel quotes, quote PDF preview, read-only policy transfer and bulk quotes for foreign-resident health are live. Connector Hub, which brings every operation under one insurer-neutral contract, runs only with a simulation provider in the new foundation.
Reference data
The insurance sector's shared vehicle make/model catalogue and official address-code (UAVT) lookup, used in quoting and in customer and asset records.
Interface and access
Turkish and English interface. Users, roles and menu permissions are managed from the Administration menu.
Where AI runs
Registration reading runs in-house on CPU, needs no GPU and writes nothing to disk. By platform principle, personal data is processed only by in-house models on servers in Türkiye, and cloud models are used only with anonymised or synthetic data.
Status
Workspace live. Customer app at pilot stage; staff mobile app in development. Shared foundation at prototype. In the AI core, registration reading is live, PDF Q&A is a prototype and the rest is on the roadmap.

Questions

Frequently asked questions

Which parts of the platform are live today?

The operations workspace is live: quote production with parallel quotes from connected insurers, policies and transfer, claims, reports and team work. On the AI side, registration reading is live. The customer app is at pilot stage, and the staff mobile app is still being built. The new shared foundation and Connector Hub are at prototype stage.

Does manual entry disappear completely?

No, and we do not promise that. Low-confidence fields and work reach a staff member by design. The goal is for staff to spend their time checking and approving work that has already been prepared.

Which decisions does AI never make?

Premiums, permissions and insurer routing are rule-based, and the premium comes only from the insurer's response. By design, issuing, cancelling, payments and sending to customers always need human approval. In registration reading, model output also passes rules that run independently of the model: plate format, chassis-number length, document-serial format, the digit count of ID or tax numbers and model-year range. A field that breaks a rule waits for approval.

Can another agency or platform staff see our data?

The workspace that is live today is an installation dedicated to a single insurance intermediary. In the new shared foundation, each agency is a separate tenant and its data is isolated at database level. By design, platform admins cannot see tenant data, and any cross-tenant job runs under a separate database role and is written to the audit log. This foundation is a prototype today and runs only on sample data. As a principle, staff corrections will become training data only with the tenant's explicit consent and after anonymisation.

In what order is the roadmap moving?

The planned order is design first, then the shared foundation and the move of the 14 insurer connections into Connector Hub, then the AI core. The next phase brings the first workspace modules and a first mobile assistant that answers questions, summarises and takes approvals. The last phase adds collections, compliance, communications, broker modules, assistant actions and the partner gateway. Each phase opens only after the previous phase's modules are in real use at an agency.