Use cases

The work that could not go to an offshore model.

Every sector has documents too sensitive to paste into a public chatbot. Those are usually the documents where AI saves the most time.

حالات استخدام لكل قطاع، والبيانات تبقى في المملكة

Government

Restricted to Top Secret

Correspondence, policy and citizen services in Arabic, with classified data kept at home.

  • Draft and summarise official correspondence
  • Search policy, regulation and circulars with citations
  • Citizen-service assistants grounded in published guidance

Banking & insurance

Restricted to Secret

Read the documents, keep the customer data in the Kingdom.

  • KYC and onboarding document extraction
  • Credit memo and policy drafting
  • Complaint triage and response drafting

Healthcare

Secret

Clinical notes and medical records never leave the Kingdom.

  • Summarise clinical notes in Arabic and English
  • Code and classify encounters
  • Patient letters and discharge summaries

Energy & industry

Restricted to Secret

Decades of engineering documents, searchable in one place.

  • Question answering over manuals and procedures
  • Incident and maintenance report analysis
  • Contract and tender review

Education & research

Public to Restricted

AI for students and researchers on accredited, in-Kingdom infrastructure.

  • Course and research assistants
  • Arabic content generation and review
  • Research data analysis without export

Data labels show the classification range these use cases typically involve. The deployment follows the classification.

Arabic-first

Built for Arabic, not translated into it.

Most models are trained mostly on English. Arabic quality, dialects and right-to-left documents are where sovereign AI has to be better, not just closer.

Evaluated on your Arabic

Model choice is decided on your own documents and questions, in Modern Standard Arabic and the Gulf dialects your users write in.

Right-to-left done properly

Bidirectional text, Arabic numerals and mixed Arabic-English documents handled in the interface and in extraction.

Arabic-aware search

Retrieval that copes with diacritics, spelling variants and morphology, so a query finds the word in all its forms.

Bilingual by default

Ask in Arabic, answer from English sources with citations, or the reverse.

Patterns

Four patterns cover most of what organisations build

Assistant

A secure chat workspace for staff, with approved models and no data leaving the Kingdom.

Grounded search

Questions answered from your documents, with the passage cited and source permissions respected.

Document pipeline

Read, classify and extract fields from documents at volume, into your systems.

Agent with approval

A multi-step workflow that drafts, checks and prepares, with a person approving before anything is sent.

Have a use case in mind?

Bring one workload and a sample of the documents. We will scope a six-week pilot around it.

Become a design partnerSee the platform