Sovereign Enterprise AI Intelligence Platform

NourNexus

One governed layer over unstructured enterprise content — discovered, classified, enriched and made retrievable for search, RAG and agents.

In preparationFinal release before the end of 2026.

The problem

Most enterprise knowledge never reaches the systems meant to use it.

The estate is not short of information. It is short of information in a shape that search, analytics and AI can act on — and every new AI project rebuilds the same ingestion problem from the beginning.

Where it breaks down

  • Trapped in documents. Obligations, decisions and history live inside scans and PDFs no system can read.
  • Structure discarded. Flattening a page to plain text throws away tables, headings and reading order — and the meaning with them.
  • Arabic mishandled. Generic pipelines mis-order Arabic script and break on mixed Arabic/English layouts.
  • Fragmented repositories. ECM, file shares, mail and databases each hold a fragment of the same answer.
  • Cloud restrictions. Regulated content cannot simply be posted to a public API endpoint.
  • Rebuilt every time. Each AI initiative re-implements ingestion, OCR and chunking from scratch.
The consequence downstream

Enterprise AI quality begins before the model.

A retrieval system can only be as good as the structure it was given. Poorly extracted text produces poorly grounded answers regardless of which model sits at the end of the pipeline — which is why NourNexus is built to fix the problem upstream rather than compensate for it downstream.

Platform overview

One governed intelligence layer, from the repository to the answer.

NourNexus is being built as a single layer that connects to enterprise content where it already lives, understands it, structures it, governs it, and makes it retrievable for search, RAG, agents and analytics.

01
Connect & ingest
Reach into existing content management systems, file shares, object storage, databases and APIs rather than requiring migration first.
02
Content intelligence
Layout-aware parsing, OCR and Arabic OCR, classification, extraction and metadata — structure recovered before text is taken.
03
Multimodal understanding
Documents, images, speech and video handled through one intelligence layer instead of four disconnected tools.
04
Knowledge, search & RAG
Structure-aware chunks, metadata, embeddings and entity relationships built for retrieval — with citations back to the source region.
05
Agents & analytics
Agentic and analytical workloads operate on governed knowledge rather than on raw files.
06
Governance throughout
Identity, classification, policy, lineage and audit wrap every stage — not a compliance section bolted on at the end.

NourNexus is in preparation, with final release stated for before the end of 2026. It is described here as a platform in active rollout: capability status is being confirmed with NourSoft product leadership, and roadmap framing is used deliberately throughout this page.

Architecture

Security, governance and sovereignty wrap the whole stack.

Select any layer to see what it does. The outer wrapper is not a stage in the flow — it applies to every stage inside it.

Security • Governance • Sovereignty
Content & document intelligence

Recover the structure the page always had.

Extraction quality is decided before a single character is read. NourNexus is designed to identify what a page is made of — then take the text out of it in the right order.

Layout-aware parsing
Headings, columns, reading order, headers and footers identified as structure rather than as a stream of characters.
OCR and Arabic OCR
Printed and scanned content in Arabic, English and mixed-language pages, processed through the same path.
Table extraction
Rows, columns and spanning cells recovered as a table, not collapsed into a run of text.
Field & key-value extraction
Named fields pulled into a defined schema, each one traceable back to the region it came from.
Entity extraction
Organizations, people, dates, amounts, references and durations identified and normalized.
Classification
Document type, department and sensitivity assigned so downstream policy has something to act on.
Signatures, stamps & seals
Detected as regions and surfaced for human confirmation rather than silently asserted as verified.
Multimodal inputs
Images, speech and video processed into the same structured, governed output shape as documents.
Human-in-the-loop
Extract Validate Approve Learn

Confidence is expressed as High confidence or Needs review — never as an invented accuracy percentage. Anything marked for review is routed to a person before it is trusted, which is what a regulated buyer needs to understand about the uncertain case, not only the confident one. Workflow shown as an illustrative future capability — confirmation pending

Data enrichment

Add the business meaning raw text does not carry.

Extraction produces values. Enrichment makes those values comparable across an entire estate — which is what turns a pile of parsed documents into something an analyst or an agent can reason over.

Normalization
Dates to a single calendar and format, amounts to a currency and scale, references to a consistent pattern — including Hijri and Arabic-Indic numerals.
Semantic labelling
Content tagged against enterprise vocabulary so that departments searching in different words still find the same document.
Summarization
Document- and section-level summaries generated as retrievable content in their own right.
Relationship linking
A contract, its annexes, its invoices and its correspondence linked as one connected record rather than four unrelated files.
Metadata enrichment
Department, document type, dates, parties and sensitivity attached to derived knowledge, not just to the source file.
Quality signals
Low-contrast scans, ambiguous fields and uncertain conversions flagged for review instead of quietly passed downstream.
Enterprise knowledge & RAG

Give AI better knowledge.

The path from a scanned page to a grounded answer, with the citation still attached at the end of it.

01
Original document
Scanned, mixed-language, multi-column source in its native form.
02
Layout-aware parsing
Reading order, tables, headers and figures preserved as structure.
03
Semantic chunking
Split on meaning and document structure, so clauses stay intact.
04
Metadata
Department, type, dates, parties and sensitivity attached to each chunk.
05
Knowledge / index
Semantic and keyword retrieval over governed, access-aware content.
06
Retrieval
Hybrid retrieval that respects the permissions the source already carried.
07
Grounded answer + citation
The answer arrives with the page and region it came from.
Structure-aware chunkingChunks follow document structure rather than a fixed character count.
Metadata preservationWhat the document was about travels with every chunk taken from it.
Access-aware retrievalRetrieval honours the permissions attached to the source content.
Knowledge graph linksEntity relationships where the content supports them.
Arabic Intelligence

Built for Arabic Enterprise Knowledge.

Arabic is a first-class intelligence layer here, not a localization pass at the end. Below is the same NourNexus Intelligence Studio used across the site — loaded with an Arabic archival document by default. Hover any extracted field to highlight the exact region it came from.

ترتيب القراءة من اليمين إلى اليساريُحافَظ على اتجاه النص وترتيبه في الصفحات العربية والمختلطة.
تخطيطات مختلطة عربي/إنجليزيمعالجة الصفحات ثنائية اللغة دون فقدان البنية أو الجداول.
الكيانات والبيانات الوصفية العربيةالجهات والتواريخ والمبالغ والأرقام المرجعية بصيغة منظمة وقابلة للاستعلام.
التقويم الهجري والأرقام العربيةتطبيع التواريخ الهجرية والأرقام العربية-الهندية مع الإبقاء على الأصل.
الاسترجاع الدلالي بالعربيةبحث قائم على المعنى يدعم مصطلحات القطاع الحكومي والمؤسسي.
الوثائق التاريخية والأرشيفيةقدرة ضمن خارطة الطريق — تخضع لتأكيد فريق المنتج قبل النشر.

Historical and handwritten Arabic support is presented as a roadmap capability pending product-team confirmation. Confidence is reported as ثقة عالية / يتطلب مراجعة rather than as a numeric figure.

Security & governance

Governed from ingestion to intelligence.

Every control below applies to derived knowledge — chunks, entities, summaries and indexes — not only to the source files they came from.

Security • Governance • Sovereignty
ConnectIngestContent IntelligenceKnowledge / RAGAgents & Analytics
Identity & accessEnterprise IAM
Role-based accessLeast privilege
Data classificationSensitivity labelling
EncryptionAt rest & in transit
Audit trailTraceable actions
Data lineageSource to answer
Policy enforcementApplied at retrieval
Content governanceLifecycle rules
Data residencyDeployment-controlled
Private deploymentNo forced public cloud
Model governanceControlled model use
Human validationConfirmation pending
Sovereign deployment

Your Data. Your Infrastructure. Your Intelligence.

NourNexus is designed to keep sensitive processing inside your own infrastructure wherever your deployment model requires it. Four models, no hidden fifth tier.

01
On-Premise
Runs on customer-controlled infrastructure. Content, derived knowledge and audit stay in your own data centre.
02
Isolated / Air-Gapped
Operation without external network connectivity, for the most controlled government and enterprise environments.
Planned capability, confirmation pending
03
Private / National Cloud
A controlled hosted environment aligned to your data-residency requirements.
Terminology pending validation
04
Hybrid
Workload placement according to business, security and governance requirements.
Regulatory posture. NourNexus is designed to support alignment with Saudi data governance and cybersecurity requirements. No claim of NCA, NDMO or SDAIA certification is made on this page.

Full deployment detail, model comparison and regulatory posture

Developer experience

Build on NourNexus.

An intelligence layer your team calls from an existing pipeline — the same interface surface regardless of which deployment model you run.

API-first accessSubmit a document, receive structured content, entities and metadata in a predictable shape.
Language-aware by defaultArabic, English and mixed-language documents handled through the same interface.
Deploy where policy requiresOne integration surface across on-premise, private cloud and hybrid environments.
Review states in the responseFields needing human validation are surfaced in the payload rather than hidden behind a score.

Official installation and authentication details will be published once confirmed by the NourSoft engineering team. No package name or endpoint on this page should be treated as a published specification.

Enterprise use cases

Built for the document realities of Saudi organizations.

Six industries, each with a document estate that predates the systems now expected to reason over it. No customer names, logos or testimonials appear on this page.

Next step

Your enterprise knowledge is already there. Make it AI-ready.

Bring a real document and a real governance constraint. We will show you what NourNexus reads from it, and where that processing would run.