Defense & aerospace
Spare-parts logistics, simulator support, sovereign deployments.
3Lines provides aircraft spare parts, simulator support, and sovereign technology solutions to defense, aerospace, and government operators across the Kingdom. We are the AI Division — building the agents, dashboards, and infrastructure that run the company's day-to-day operations and ship sovereign AI to other Saudi enterprises.

We design, build, and deploy custom AI systems for ERP, HR, finance, and field operations — wired into the tools your team already uses.
Reviewed 142 invoices in Odoo overnight.
3 contract renewals require approval this week.
Drafted 8 supplier replies in Outlook.
Trusted across the Kingdom
Six core practices, one operating model. Each engagement is scoped, priced, and shipped against measurable outcomes — not slideware.
Unified dashboards that surface ERP, email, calendar, HR, and reports through a single Claude-powered briefing.
RAG over policies, contracts, RFPs, and SOPs. Cited, auditable answers in Arabic and English.
Spare-part orders above SAR 250k require dual-signature approval before…
Vendor response window closes 30 days after publication; clarifications via…
Liquidated damages capped at 12% of total contract value; force majeure…
Triage, draft, and route Outlook traffic. Approval gates and full audit logs on every send.
Native integrations with Odoo and Jisr — turn natural language into invoices, leave requests, and reports.
Voice-first agents for procurement, maintenance, and dispatch. Works offline, syncs when online.
On-prem or sovereign cloud. We handle evaluation, fine-tuning, guardrails, and red-teaming.
model: claude-opus-4-7region: ksa-riyadh-1deployment: sovereign-cloudguardrails: stricteval: continuousaudit: per-token
Our flagship internal platform — Claude wired into Odoo, Outlook, Calendar, Jisr HR, and reports. Operators run a full workday from one window. Built and shipped to 3Lines staff in under 8 weeks.
Search, approve, and reconcile purchase orders, sales orders, and invoices through chat. Generic search across any Odoo model — contacts, products, employees, leads.
Triage inbox, draft replies grounded in your contracts, and create or move calendar events without leaving the chat. Microsoft Graph wired end-to-end.
Approve leave requests, pull headcount reports, and answer policy questions against your handbook. Bilingual responses, full audit per action.
Sales summary, invoice aging, purchase overview, accounts receivable, product performance — generated on demand from live ERP data, with charts and PDF export.
Create, prioritize, and complete tasks via chat. Tags, due dates, and status. The AI proactively surfaces what needs attention each morning in the daily briefing.
Full Arabic + English UI with automatic RTL flip. Self-hosted at acc.ai3lines.com on KSA infrastructure. No data leaves the Kingdom.
Seven AI systems we've built and deployed for 3Lines across the Kingdom. Each one ships with audit trails, bilingual UX, real integrations into the systems your team already uses, and measurable time savings.
AI-powered RF circuit design assistant for the engineering team. Describe what you need in plain English, the agent designs the circuit, simulates it with scikit-rf, plots S-parameters and Smith charts, runs Monte Carlo yield analysis, and exports to Touchstone or SPICE.
Field reports, contract review, RFQ analysis, and an Arabic voice agent are next on the roadmap. Want to see something specific built for your workflow?
Talk to the teamEvery engagement follows the same disciplined cadence. No mystery, no scope drift.
Workshops with your team. We map the workflows, data, systems, and decisions worth automating.
Architecture, prompt patterns, evaluation criteria, and a working prototype against your real data.
Production rollout with observability, guardrails, audit trails, and bilingual UX.
Continuous evaluation, model updates, and quarterly reviews tied to business KPIs.
Combined output of every agent we've shipped, measured against the workflows they replaced.
Built for environments where reliability, traceability, and bilingual support are not optional.
Spare-parts logistics, simulator support, sovereign deployments.
Reconciliation, KYC, regulatory reporting, investor relations.
Clinical documentation, patient comms, claims and authorisation.
Citizen services, document processing, multilingual support desks.
Procurement copilots, supplier comms, fleet and dispatch.
Field reports, asset inspections, compliance documentation.
Six specialised systems we are scoping or piloting for defense and security operators in the Kingdom. Every system is positioned as defensive, monitoring, or human-in-the-loop decision-support — never autonomous targeting.
Real-time fusion of radar, ELINT, and SIGINT streams. Deep models classify and de-noise signals; anomaly scoring routes findings to operators.
Tracks fused with ADS-B and intel feeds into one situational picture. Unusual approach patterns flagged for the operator to review.
Wearable + sensor data (HR, SpO₂, fatigue, heat). Edge models surface medical risk before it becomes critical. Fully on-prem.
Identifies and scores potential security risks for personnel, locations, and operations. Output is advisory — every action gated by a trained human.
Spare-parts forecasting, supply-chain risk, perimeter surveillance summaries, and a natural-language layer over secure command dashboards.
Cited Q&A over manuals, ROE, and orders in Arabic + English. Air-gapped deployment, full audit trail per question and per source citation.
Building sovereign AI in the Kingdom means hitting walls that don't exist for hyperscalers. These are the ones we're investing in — partners and collaborators welcome.
Senior AI engineers and bilingual prompt designers are scarce in the Kingdom. We're investing in apprenticeships and pairing local engineers with frontier-lab veterans to grow the bench fast.
Modern Standard Arabic is well-served, but Saudi dialect, government documents, and defense terminology are not. We're curating high-quality bilingual corpora with permission-cleared sources.
Defense and healthcare workloads require fully offline, audit-traceable LLM stacks. We're hardening Llama and Qwen-based pipelines that run without phoning home and meet KSA classification requirements.
Real-time bilingual voice agents need < 800 ms turn-taking. Off-the-shelf TTS for Arabic struggles with dialect, and ASR drops 10-15 % on Saudi accents. We're tuning open models on local speech.
Most public benchmarks measure English. Building automated eval pipelines that catch hallucinations, dialect slips, and policy violations in Arabic responses is an open research area for us.
Cross-encoder rerankers and small task-specific models trained on defense MRO, financial reconciliation, and clinical Arabic data — open-sourced where the data allows.
Tell us about the workflow you want to transform. We'll come back within 48 hours with a scoped recommendation.