Welcome

We blend strategy, design and engineering to build intelligent systems that help businesses in Saudi Arabia and the GCC grow and stand out.

Hakam

AI claims platform for motor and medical insurance, from first notice to payout

  • A Shfrah product
  • Insurance
  • AI
  • SaaS
  • At a glance

Sector

Insurance

Platforms

Web · iOS · Android · WhatsApp

Modules

8

Integrations

7

Screens

4

Total duration

22 weeks

Hakam takes the first notice of loss on WhatsApp or the app, reads Najm reports and medical invoices automatically, estimates damage from photos and flags fraud patterns before payout. Adjusters work from one desk, medical claims flow to NPHIES, and the customer follows every step.

Hakam: screen 1
Control tower: open claims, time to decision and fraud-flagged claims on one screen.
  • Challenge

  1. 01

    Claims arrive through scattered channels — WhatsApp, email, branches — and staff re-key the data by hand.

  2. 02

    Medical claims are rejected for missing codes or documents, and the reason surfaces days later.

  3. 03

    Fraud hides across many small claims, with no view linking parties, vehicles and workshops.

  4. 04

    There is no single SLA view, so managers learn about delays from customer complaints.

Who it is built for

  • Policyholder

    Reports in two minutes and always knows where the claim stands

  • Claims adjuster

    Documents, extracted data and the decision on one screen

  • Medical reviewer

    Correct coding and pre-authorisation before sending to NPHIES

  • Fraud analyst

    Early alerts and a network of links between parties

  • Finance

    Documented payouts and daily bank reconciliation

  • Modules

What the system does

Connected modules working on one set of data.

  • 01

    First notice of loss

    Report via WhatsApp or app with photos, location and Najm report number

  • 02

    Document AI

    Extracts fields from Najm reports, invoices and Arabic medical reports

  • 03

    Damage estimation

    Identifies damaged parts from photos and estimates repair cost

  • 04

    Fraud engine

    A risk score per claim and a network linking parties and workshops

  • 05

    Adjuster workbench

    Review, decide and request documents with a full trail

  • 06

    NPHIES gateway

    Eligibility, pre-authorisation and medical claims in FHIR

  • 07

    Payouts & settlement

    Payouts to customers or workshops with daily bank reconciliation

  • 08

    SLA control tower

    Tracks target time per stage and alerts before a breach

Where the AI works

  • AI · 01

    An Arabic document model extracts the report number, fault ratio and vehicles from Najm reports, and line items and codes from medical invoices.

  • AI · 02

    A vision model finds damaged parts in vehicle photos and proposes repair lines from the approved workshop price list.

  • AI · 03

    Fraud scoring combines expert rules with network analysis across people, vehicles and bank accounts.

  • AI · 04

    A private in-Kingdom language model summarises each file for the adjuster and suggests ICD-10 and CPT codes for the medical reviewer.

  • Screens

The system as the client's team and customers use it.

Hakam: screen 2
Adjuster workbench: the Najm report read by AI, extracted fields, damage estimate, fraud score, then the decision.
Hakam: screen 3
Fraud analysis: a network linking claims to people, vehicles, workshops and accounts reveals organised rings.
Hakam: screen 4
Customer app: report in under two minutes with photos and location, then follow every stage.
  • Architecture

Independent services, in-Kingdom data, and integrations over documented APIs.

Channels

  • Customer app: iOS & Android
  • WhatsApp bot: FNOL & updates
  • Adjuster desk: Internal web app

Edge

  • API gateway: Rate limits & WAF
  • Identity: Nafath & roles

Services

  • Claims service: Claim lifecycle
  • Documents service: Files & photos
  • Workflow engine: Stages & SLAs
  • Payouts service: Payouts & reconciliation

AI

  • Document AI: Arabic OCR & extraction
  • Damage vision: Damage from photos
  • Fraud scoring: Rules & graph

Data

  • PostgreSQL: System of record
  • Kafka: Claim events
  • Object storage: In-Kingdom

Integrations

  • Najm: Accident reports
  • NPHIES: HL7 FHIR
  • Yakeen: Identity checks
  • Bank API: Payouts & IBAN

Connections

  • Customer app to API gateway
  • WhatsApp bot to API gateway
  • Adjuster desk to API gateway
  • API gateway to Identity
  • API gateway to Claims service: REST
  • API gateway to Documents service
  • Claims service to Workflow engine
  • Claims service to Fraud scoring
  • Documents service to Document AI
  • Documents service to Damage vision
  • Claims service to PostgreSQL
  • Claims service to Kafka: events
  • Documents service to Object storage
  • Workflow engine to Kafka
  • Payouts service to PostgreSQL
  • Claims service to Najm
  • Claims service to NPHIES: FHIR
  • Claims service to Yakeen
  • Payouts service to Bank API

Stack

  • Next.js
  • NestJS
  • Flutter
  • PostgreSQL
  • Kafka
  • Temporal
  • Python · FastAPI
  • Kubernetes
  • Workflow

Claim lifecycle

  1. 01

    First notice

    CustomerWhatsApp · App

  2. 02

    Policy & identity check

    AutomaticCore policy · Yakeen

  3. 03

    Read documents

    AutomaticDocument AI · Najm

  4. 04

    Estimate damage

    AI, then adjusterDamage vision

  5. 05

    Fraud check

    AutomaticFraud scoring

  6. 06

    Adjuster decision

    AdjusterWorkbench

  7. 07

    NPHIES approval (medical)

    Medical reviewerNPHIES

  8. 08

    Payout & notify

    FinanceBank API · WhatsApp

  • Data & API

The core entities and relations, and a live example from the API.

Policy

  • iduuidPK
  • numbertextunique
  • holder_iduuidFK
  • productenum
  • coveragejsonb
  • valid_fromdate
  • valid_todate

Claim

  • iduuidPK
  • codetextunique
  • policy_iduuidFK
  • typeenum(motor,medical)
  • statusenum
  • amount_claimednumeric
  • sla_due_attimestamptz
  • fraud_scoresmallint

Party

  • iduuidPK
  • national_idtext
  • nametext
  • phonetext
  • ibantext
  • verified_attimestamptz

Document

  • iduuidPK
  • claim_iduuidFK
  • kindenum
  • storage_keytext
  • extractedjsonb
  • confidencereal

Assessment

  • iduuidPK
  • claim_iduuidFK
  • linesjsonb
  • totalnumeric
  • byenum(ai,adjuster)
  • decided_attimestamptz

FraudSignal

  • iduuidPK
  • claim_iduuidFK
  • ruletext
  • weightsmallint
  • linked_entitytext

Payout

  • iduuidPK
  • claim_iduuidFK
  • beneficiary_iduuidFK
  • amountnumeric
  • bank_reftext
  • statusenum

AuditLog

  • idbigserialPK
  • actor_iduuid
  • entitytext
  • actiontext
  • beforejsonb
  • afterjsonb
  • attimestamptz
  • Policy 1—∞ Claim
  • Claim 1—∞ Document
  • Claim 1—∞ Assessment
  • Claim 1—∞ FraudSignal
  • Claim 1—1 Payout
  • Party 1—∞ Policy
POST /v1/claims Open a new claim from any channel

Request

{
  "policy_number": "POL-MTR-552190",
  "type": "motor",
  "incident": {
    "occurred_at": "2024-11-03T20:42:00+03:00",
    "location": { "lat": 24.7136, "lng": 46.6753 },
    "najm_report": "NJ-1446-0937721"
  },
  "attachments": ["upl_7f3a", "upl_7f3b", "upl_7f3c", "upl_7f3d"],
  "channel": "whatsapp"
}

Response

{
  "id": "clm_01JC8Q4M5TZ",
  "code": "CLM-24-08812",
  "status": "documents_received",
  "sla_due_at": "2024-11-04T20:51:00+03:00",
  "required_documents": [],
  "tracking_url": "/claims/t/8812"
}

Webhook event · claim.status_changed

{
  "event": "claim.status_changed",
  "claim": "CLM-24-08812",
  "from": "under_review",
  "to": "approved_for_payout",
  "amount": 11530.00,
  "currency": "SAR",
  "at": "2024-11-04T09:12:44+03:00"
}
  • Integrations

  • Najm

    Accident reports and fault ratios

  • NPHIES

    Eligibility, pre-auth and medical claims (HL7 FHIR)

  • Yakeen

    Customer identity verification

  • Nafath

    Customer sign-in with the national SSO

  • WhatsApp Business

    FNOL intake and customer updates

  • Bank payouts

    Payouts and IBAN validation

  • Core policy system

    Policies and coverages over API

  • Security & compliance

  • Data and backups hosted inside the Kingdom.
  • National Cybersecurity Authority Essential Cybersecurity Controls (ECC).
  • Personal Data Protection Law (PDPL) alignment, with sensitive data masked in logs.
  • Role-based access and maker-checker on every payout decision.
  • Encryption in transit and at rest with managed keys.
  • Immutable audit trail for every change to a claim.
  • Delivery plan

Clear phases with measurable deliverables.

  1. 01

    Discovery & process mapping

    • Workshops with claims, fraud and finance
    • Current and target process maps
    2 weeks
  2. 02

    UX & architecture

    • Screen prototypes tested with adjusters
    • API and data design
    3 weeks
  3. 03

    MVP build

    • FNOL, workbench and document AI
    • Customer app and WhatsApp bot
    8 weeks
  4. 04

    Integrations

    • Najm, NPHIES, Yakeen and bank
    4 weeks
  5. 05

    Model tuning & UAT

    • Tune damage and fraud models on company data
    3 weeks
  6. 06

    Go-live & hypercare

    • Phased launch by product line
    2 weeks
Total duration
22 weeks
Team
Project manager, business analyst, 3 backend and frontend engineers, a mobile engineer, an ML engineer, QA and DevOps.

Performance

  • < 24h First decision on a simple claim
  • 90%+ Najm fields captured automatically
  • 95%+ Medical claims sent complete
  • 99.9% Availability

Related work

All work

Have a similar project?

Start a project
Contact us