AI for Government & Public Sector —
Arabic-First Consulting, Training, and Delivery

Citizen-service call centers answer in Modern Standard Arabic scripts while callers speak Gulf dialect, so the global chatbot vendor piloted last year cannot tell a Najdi complaint from a Hijazi inquiry. Three AI pilots demoed well at the innovation showcase last year; none reached production, and the next budget review will ask why. National AI strategies hand down KPIs for adoption and trained staff, but the free vendor courses are generic English slideware — teams finish the certificate and still cannot automate a single workflow. Institutional knowledge lives in thousands of Arabic policy PDFs and circulars, so new employees take months to answer correctly and even veterans give inconsistent answers across channels. Every proposal that lands on the transformation office's desk is either a Big-4 deck priced at ten times the budget or a startup demo with no data-residency answer — nobody shows evals, monitoring, or an Arabic accuracy number, and with no internal framework to score feasibility or vendor claims, everything stalls in committee. Meanwhile reports, summaries, and inter-departmental correspondence consume senior staff hours, drafted bilingually in Arabic and English with quality that varies wildly from one desk to the next.

MIT research found that 95% of enterprise generative-AI pilots produce no measurable P&L return — a failure mode that lands even harder in government, where pilot theater at an innovation showcase is a known pattern and the next budget cycle asks hard questions about what actually shipped. We contract to a measurable outcome — resolution rate, answer accuracy, hours saved — and instrument it from day one rather than demo it once. A citizen-inquiry assistant delivered via a prime would be measured on first-contact resolution rate, Arabic intent-classification accuracy on a held-out dialect eval set, and escalation rate. An applied workshop program for a government academy or SOE would be measured on pre/post skills-assessment delta and the number of participant-built automations still in production 30 days later. We do not publish invented numbers or client names before an engagement completes and the client signs off — every case study follows our standard production process, only populated from real instrumented work.

Saudi Arabia: SDAIA regulates and drives adoption under the National Strategy for Data & AI, which targets 20,000 AI specialists — a mandate that creates training budgets before local Arabic-training supply exists. Procurement runs largely through Etimad with heavy local-content scoring, and classified workloads carry in-Kingdom data-residency and NCA cybersecurity-controls obligations, alongside PDPL for personal data — so we work as a specialist subcontractor to licensed primes for tendered scopes, and directly for training, advisory, and SOEs. UAE: national upskilling programs are funded and mandated at scale (Microsoft alone committed to certifying 100,000 UAE government employees in AI), making training our strongest direct lane there, while federal tenders again run through local or free-zone entity requirements that favor a partner route. Qatar, Kuwait, Bahrain, and Oman all run national digitization programs with AI components; we track them for content and future partnerships but are not yet actively prospecting there. Egypt is our delivery base and proving ground under Law 151/2020 — valuable for credentials and talent, but we do not pursue Egyptian government contracts directly. Across every market, procurement realities are stated plainly rather than glossed over: vendor-registration lead times, 60-90+ day payment terms routed through the prime, and committee-driven decision cycles that lengthen the sales process — we plan engagements around that reality rather than promise otherwise.