Skip to content

White paper November 2026

The Live Gap: AI in Saudi operations, 2026

Most organisations in the Gulf use AI. Few have scaled it, and fewer can show the value. This paper sets out the reality, the rules that are coming, the compass SDAIA has already published, and what to do in the next 90 days.

Definition: The Live Gap is the distance between an AI pilot that works in a demo and a system running in daily operations, with someone accountable, live data and measured value.

Octopus research team11 min read

Download the PDF

Read the full paper on the web. The PDF needs a work email.

1. Adoption is wide. Value is thin.

AI is no longer new in the Gulf. In McKinsey's November 2025 survey, 84% of GCC organisations surveyed said they had adopted AI in at least one function, up from 62% in the 2023 survey[1]. Official Saudi statistics point the same way. GASTAT reports that 27.6% of Saudi establishments used AI technologies in 2024[2], and 33.1% in 2025[3].

Use is uneven across sectors. In 2025, information and communications led at 61.1%, followed by financial and insurance at 52.9% and education at 51%[3]. In 2024, manufacturing stood at 26% and electricity and gas at 25.7%[2]. The sectors with the heaviest operations are not yet the heaviest users.

31%

of GCC organisations surveyed say AI is scaling or fully scaled. 84% use it in at least one function.

Source: McKinsey, The state of AI in GCC countries, Nov 2025

Scale is another matter. Only 31% of the same organisations say AI is scaling or fully scaled. The rest are not using it, experimenting or piloting. In McKinsey's words, "more than two-thirds haven't moved beyond pilots"[1].

Value is thinner still. Of every 100 organisations, only 11 see a clear impact on profit (5% or more of earnings). McKinsey calls them value realizers: they scale AI and attribute that share of their earnings to it. Yet 89% plan to increase their AI budgets in the coming year[1]. Spend is rising faster than value.

The prize is large, but it is an estimate. In 2018, PwC estimated that AI could contribute US$135.2 billion to the Saudi economy in 2030, equal to 12.4% of GDP (forecast)[4]. Estimates like this are a reason to act. They are not a measure of progress.

2. Why pilots stall

The gap between pilot and production is not only a Gulf problem. In a Gartner survey of 644 respondents in the US, Germany and the UK, only 48% of AI projects reached production, and the move from prototype took 8 months on average[5]. BCG found that 74% of companies have yet to show tangible value from AI[6].

The cause is rarely the algorithm. BCG estimates that about 70% of AI challenges are people and process issues[6]. McKinsey's GCC data points the same way. Workflow redesign "has the strongest link to earnings improvement" from AI, and in the GCC it is "still at an early stage"[1]. All but one of the value realizers surveyed had adoption and scaling strategies backed by change management, against 41% of other organisations[1].

How the work is organised matters too. McKinsey cites research in financial services where 70% of organisations with centralised AI operating models had moved pilots into production, against about 30% of those with decentralised models[1]. Central ownership of strategy and value, with AI working inside each operation, is a pattern that recurs.

In the Kingdom, the stall shows up in six places.

  • Pilots that never reach production. The pilot ran on a clean export, and nobody planned the integration with live systems.
  • Tools bought, but the work unchanged. Approvals, handoffs and forms stay the same, so people do the old work plus one new step.
  • Governance moving from voluntary to binding. Compliance arrives after the design is fixed, and the redesign takes longer than the pilot.
  • Data that must stay in the Kingdom. Hosting is decided late, and the model that was tested cannot run where the data has to live.
  • A talent gap. Few people have taken AI into live operations, and hiring them takes time.
  • Consulting spend that has to show a return. Large programmes are harder to approve, so every stage has to earn the next.

A pilot is not progress. A system in operations is.

Octopus

3. The rules are moving from voluntary to binding

For the last few years, AI governance in the Kingdom has rested mostly on principles. That is changing. The direction is clear, even where the text is not yet final.

Personal data

The Personal Data Protection Law (PDPL) took effect on 14 September 2023, with a one-year grace period to 14 September 2024, and SDAIA is the regulator[7]. Cross-border transfers need adequacy or appropriate safeguards[7]. Where a breach must be reported, SDAIA has to be notified within 72 hours of the controller becoming aware of it[8]. For AI, that makes the location of hosting and inference a design decision, not an IT detail.

Cybersecurity

NCA's Essential Cybersecurity Controls (ECC-2:2024) set 108 main controls across four main domains. They apply to government agencies, their affiliated companies and entities, and private entities that own, operate or host critical national infrastructure[9]. The Cloud Cybersecurity Controls (CCC-2:2024) add requirements for cloud tenants in scope[10]. In the 2024 versions, data localisation controls moved out of the NCA texts to the National Data Management Office (NDMO) at SDAIA, which entities must consult before acting on localisation[9].

AI ethics

SDAIA's AI Ethics Principles set seven principles, from fairness to accountability, and four risk levels, from little or no risk to unacceptable. High-risk systems need conformity assessments before and after deployment[11].

What is coming (draft)

In November 2026, SDAIA's Responsible AI Policy was still a draft, after a public consultation that closed in May 2026. As summarised by AS&H Clifford Chance, the draft would make AI governance binding: four risk tiers with critical systems prohibited, registration on an AI Governance Platform, ethics badges, safety reports reviewed before high-risk deployment, and annual third-party audits for high-risk systems[12]. A companion draft, the AI-First Policy, would make AI the preferred option for public services where it is shown to be suitable, and recommends that state-owned companies apply its principles[12]. Both texts may change. Their direction is what matters for planning.

Rules are coming. Build for them now, not after.

Octopus

4. The foundations are in place

The Kingdom is not starting from zero. Government cloud adoption reached 41% by Q3 2025, up from 14% in 2021, with more than 156 government entities migrated[13]. Across all establishments, 51.3% used cloud services in 2025[3]. For many organisations, the infrastructure AI needs is already there.

14% → 41%

Government cloud adoption, 2021 to Q3 2025

Source: DGA, Nov 2025

Strategy is strong too. In the Tortoise Global AI Index 2024, Saudi Arabia ranked 1st for government strategy and 14th overall. Its weakest areas were talent and the operating environment[14]. The National Strategy for Data and AI targets more than 20,000 data and AI specialists and experts by 2030 (target)[15]. McKinsey notes that new roles such as AI product owners and forward-deployed engineers have "relatively few people" qualified to fill them[1].

So the constraint has moved. Budgets, cloud and strategy are largely in place. What is scarce is people who have taken AI into live operations, and an operating model that lets them do it again.

5. The compass SDAIA has already published

Organisations do not need a new framework to steer by. SDAIA's AI Adoption Framework, first launched in September 2024 and updated in 2025, covers all government and private entities. It has three pillars and seven components: Directions (strategy, governance), Enablers (data, infrastructure, human capacity) and Outcomes (applications, impact)[16].

Three things in it deserve more attention than they get.

  1. It names a KPI for tracking the transition of projects from pilot to production[16]. The official signal of maturity is how many projects reach production, not how many pilots run.
  2. It asks for ROI measurement, a quarterly review of each AI initiative, and KPIs per initiative, such as hours saved, satisfaction and model accuracy[16].
  3. It recommends a dedicated AI office, a defined AI unit, an AI strategy linked to Vision 2030 and the national data and AI strategy, and alignment with the AI Ethics Principles[16].

SDAIA's National AI Index measures government entities on the same structure: 3 pillars, 7 axes and 23 sub-fields[17]. For semi-government entities in particular, this is the yardstick they will be read against.

We map our six practices to these seven components, so clients can see where each piece of work lands. The framework is SDAIA's. Our mapping is a reading of it, not a replacement.

6. What it means for each kind of organisation

State-owned companies

A board that sets an AI mandate should also ask what that mandate has returned. Aramco, which publishes its numbers, reports 442 AI use cases identified, more than 200 deployed, and US$2.6 billion of AI-driven technology realised value in 2025[18]. That is the kind of ledger a board can ask for. Consulting spend is under scrutiny too. In May 2026 the Ministry of Finance said the government remains focused on ensuring consultancy investments deliver measurable returns[19].

  • Turn your AI mandate into a ranked, funded roadmap.
  • Measure value the board can audit.
  • Scale what works across business units.

Private companies

Adoption is highest in information and communications, finance and insurance, and education[3]. In most companies the tools are in use and the work around them has changed less. That matters, because workflow redesign is the move most strongly linked to earnings[1].

  • Redesign two or three core workflows end to end.
  • Choose models and hosting without lock-in.
  • Name someone in-house who is accountable for AI, and train them.

Government and semi-government entities

The ECC and CCC apply directly to government entities and their affiliated companies[9,10]. Under the draft Responsible AI Policy, public entities would work through their AI office to inventory AI systems, set up governance, classify systems by risk, add human oversight and explainability, monitor after deployment, and prepare for registration (draft)[12].

  • Set up the AI office and governance before the Responsible AI Policy becomes final.
  • Keep data in the Kingdom, in line with PDPL and NDMO rules.
  • Track pilot-to-production, the KPI SDAIA's framework names.

7. The next 90 days

Whatever kind of organisation you lead, one quarter is enough to move one system much closer to live operations. This is the plan we would follow.

None of this needs a new framework or a large programme. It needs a clear choice of where to start, decisions made in the right order, and a team that stays until the system runs. The gap between pilot and live operations is not a technology gap. It is a gap in design, ownership and discipline, and it can be closed one system at a time.

Value is what finance signs off, not what a demo shows.

Octopus

The 90-day plan draws on two of our practices: AI strategy, investment and value realisation, and AI governance, risk and regulatory readiness.

About this paper

This paper draws on public sources available on 16 November 2026. Figures are quoted as published and are not rounded. Survey figures are self-reported, and the McKinsey GCC sample is 139 senior executives and board directors. Forecasts, targets and drafts are labelled. Draft policy texts may change before they are final. Some figures come from press or law-firm reports of the original source, as listed below. This paper is a practical guide to start from; your legal and security teams sign off.

Sources (19)

Figures are quoted as published. Forecasts and drafts are labelled.

  1. 1McKinsey and GCC Board Directors Institute, The state of AI in GCC countries: In pursuit of scale and value, Nov 2025
  2. 2GASTAT, Establishments' ICT Access and Usage Statistics 2024
  3. 3GASTAT, Establishments' ICT Access and Usage Statistics 2025
  4. 4PwC Middle East, The potential impact of AI in the Middle East, 2018 (forecast)
  5. 5Gartner, survey press release, 7 May 2024
  6. 6BCG, AI adoption in 2024: 74% of companies struggle to achieve and scale value, press release, 24 Oct 2024
  7. 7Clyde & Co, Saudi Arabia's Personal Data Protection Law, Sep 2024
  8. 8DLA Piper, Data Protection Laws of the World: Saudi Arabia
  9. 9NCA, Essential Cybersecurity Controls (ECC-2:2024)
  10. 10NCA, Cloud Cybersecurity Controls (CCC-2:2024)
  11. 11SDAIA, AI Ethics Principles, Sep 2023
  12. 12AS&H Clifford Chance, briefing on SDAIA's draft Responsible AI Policy and AI-First Policy (drafts), 4 Nov 2026
  13. 13Digital Government Authority, reported by Sharikat Mubasher, 6 Nov 2025
  14. 14Tortoise Media, Global AI Index 2024, reported by Arab News, Sep 2024
  15. 15SDAIA, National Strategy for Data and AI: objectives and targets
  16. 16SDAIA, Artificial Intelligence Adoption Framework, 2025
  17. 17SDAIA National AI Index, reported by Arab News, Jul 2025
  18. 18Aramco, Building the AI future
  19. 19Consultancy-me, on consulting spend in Saudi Arabia, May 2026

Download the PDF

Read the full paper on the web. The PDF needs a work email.

Keep reading

All insights

From pilot to live

New articles and practical notes, by email.

Want to talk about your own operations?