Every board in the GCC has an AI slide. Very few have an AI P&L.
That gap is the most important thing to understand about enterprise AI in the region today. It is also, in our view, the reason AI workflow automation is the most underpriced infrastructure category in the Gulf right now, rather than AI itself.
The numbers are blunt. According to McKinsey's 2025 State of AI in GCC Survey (QuantumBlack, November 2025), 84% of regional organisations have adopted AI in at least one business function, up from 62% in 2023. But only 11% qualify as “value realizers,” meaning they have scaled AI and can attribute 5% or more of their earnings to it. Two-thirds of organisations are still stuck in pilots.
Adoption is not the binding constraint. Execution is. And execution in Arabic-speaking, regulated GCC enterprises carries specific architectural requirements that many current deployments were not designed to meet.
The investable opportunity is not “AI in the GCC.” It is the infrastructure layer that closes the gap between 84% adoption and 11% value capture.
Why Global Models Fall Short in GCC Enterprise Workflows
Three practical issues sit between most GCC enterprises and the EBIT impact their CIOs committed to.
Language. Global foundation models were trained overwhelmingly on English. They handle Arabic as a secondary layer, often struggle with Modern Standard Arabic morphology, and can perform poorly on Khaleeji and Levantine dialects, or on the Arabic–English code-switching that defines real customer interactions in Dubai, Riyadh, or Doha. For a region serving more than 400 million Arabic speakers, this is not a rounding error in user experience. It is the core product requirement.
Data. The UAE Data Protection Law, Saudi Arabia’s NDMO and SDAIA frameworks, SAMA for banks, TDRA for telecoms, and ADGM for regulated financial institutions increasingly require strict data residency, in-region hosting, or architecture-level compliance controls for sensitive workloads. For high-risk deployments, sovereign architecture is often the entry ticket. That means on-premise, in-region cloud, or air-gapped. Global LLMs accessed by default through US hyperscaler APIs frequently cannot meet this bar without significant and sometimes uneconomic reconfiguration.
Execution. The commercial prize in enterprise AI is no longer conversational. It is agentic. The shift investors should be tracking is from LLMs that answer questions to AI agents that execute workflows from start to finish. In our experience, translation layers bolted onto English-native agents tend to break under the demands of regulated production. Domain-tuned, Arabic-capable agents plugged into enterprise systems with auditable guardrails are what the market is actually asking for.
The Investment Case: Arabic Is Infrastructure
The category we are focused on is not “Arabic AI” as a product. It is the workflow infrastructure that is Arabic-capable, deployable inside sovereign environments, and agentic in execution. This is the layer that lets a Saudi bank, a UAE ministry, or a Qatari telecom turn an AI pilot into measurable operating impact.
The category has, in our view, an unusually attractive investment profile. Demand is increasingly mandated rather than discretionary, as sovereign AI frameworks move from preference toward regulatory expectation. The moat is real and compounds over time. Arabic-language data, regulated-sector deployment experience, and workflow integration cannot be replicated quickly with compute budget alone. Enterprise stickiness accelerates once an AI agent is embedded in a core workflow such as KYC or claims. The consolidation pathway is visible: sovereign AI champions such as G42 and HUMAIN, regional telecoms building their own vertical stacks, and global hyperscalers needing credible in-region partners together create a workable exit environment.
What a Winning Platform Looks Like
Four characteristics, taken together, define the businesses most likely to lead this category. Importantly, very few will own a foundation model end to end, and those that do not can still win.
Arabic-native capability. Not necessarily ownership of a foundation model, but ownership of enough of the stack (curated data, post-training, fine-tuning, evaluation benchmarks) to deliver production-grade Arabic performance across dialect and domain.
Sovereign deployment architecture. On-premise, air-gapped, and in-region cloud options that allow regulated banks, ministries, and telecoms to deploy without triggering data residency or audit risk.
Agentic execution, not chat UX. KYC is the cleanest test case. A Gulf retail bank is onboarding tens of thousands of customers each month. Documents arrive in Arabic and English, often mixed in the same file. Sanctions and PEP screening depends on Arabic name transliteration, which is a well-known source of false positives and false negatives. A real agentic platform does the whole workflow (ID verification, name matching, document parsing, risk scoring, exception routing) inside the bank’s sovereign environment, with a complete audit trail. A chatbot that answers customer FAQs in Arabic is a different product. It does not solve the same problem and does not build the same moat.
Commercial discipline. Profitability or a credible path to it, real enterprise traction in regulated sectors, and the operational muscle to drive change inside the client. The winners will have to do more than ship software.
Sovereign foundation model efforts such as Jais (Inception and MBZUAI), Falcon (TII), and Saudi Arabia’s emerging model ecosystem have established that the regional technical layer exists. The investable frontier, for us, is the execution layer that turns those models into enterprise workflow infrastructure.
What Could Go Wrong
Three risks matter more than the rest, and it is worth being honest about each.
The Arabic foundation model landscape is competitive, and global models are closing the Arabic gap faster than many expected. The moat therefore cannot rest on the model alone. It has to be the combination of language capability, enterprise distribution, vertical fine-tuning, deployment architecture, and an operating track record.
Compute concentration is a second risk. A large majority of global AI compute runs through Nvidia, and GCC sovereign AI plays are ultimately downstream of that supply chain, however ambitious the regional narrative becomes.
The adoption-to-value gap is itself double-edged. It is the investment opportunity, but it also reflects that many GCC enterprises are not yet operationally ready to capture AI’s full P&L impact. Change management inside the client will separate the businesses that scale from the businesses that stall.
The M Capital Thesis
At M Capital, AI & Enterprise Automation is one of our three Core Sectors, alongside Cybersecurity & Digital Trust and Payments & Financial Infrastructure. What these sectors share is a demand profile driven by regulation rather than discretionary spend, together with a regional consolidation pathway that sits ahead of global multiples catching up.
MENA technology spending is projected to reach USD 169 billion in 2026 (Gartner), and regional data-centre capacity is forecast to triple by 2030. The capital is in place. The compute is being built. The regulatory architecture is being written. What is still being built, and what is investable today at growth-equity valuations, is the Arabic-capable execution layer that actually sits between sovereign infrastructure and enterprise earnings.
AI workflow automation in the GCC today, in our view, resembles cybersecurity 18 to 24 months ago. Demand is mandated. The moat is real when built properly. The consolidation path is visible. And the window before global multiples arrive is open but not wide.
That is the category we are backing.
Sami Besbes is Chief Investment Officer at M Capital Limited, an ADGM-regulated (FSRA Category 3C) growth equity investment manager focused on digital infrastructure across the GCC, Africa, and South Asia. M Capital invests across Payments & Financial Infrastructure, Cybersecurity & Digital Trust, and AI & Enterprise Automation.
This article reflects the author’s views and does not constitute investment advice or a solicitation to invest. Market data sourced from McKinsey QuantumBlack, Stanford HAI, Gartner, IMF, and publicly available disclosures.
#AI #EnterpriseAutomation #GCC #DigitalInfrastructure #SovereignAI #ArabicAI #AgenticAI #GrowthEquity #UAE #SaudiArabia #MCapital #AbuDhabi #Dubai