Deloitte says most Middle East AI projects are still stuck in pilot mode

Deloitte Middle East returned to LEAP 2026 in Riyadh for what it says is its fourth consecutive year at the event, presenting research that puts a specific, uncomfortable number on a problem most of the technology vendors at the show would rather not dwell on: only a third of organisations in the region say they are using AI to actually change how they operate. The research is Deloitte's 2026 State of AI in the Enterprise report, released in June and covering the Middle East. Its headline findings describe two different speeds of progress happening at once. Access to AI tools has expanded fast — the firm says sanctioned AI tool access across the region's workforce grew from under 40 percent of employees to nearly 60 percent over the past year, a roughly 50 percent expansion — and more than half of organisations surveyed expect at least 40 percent of their current AI experiments to reach production within three to six months. Those are adoption numbers, and by that measure the region looks like it is moving quickly. The transformation numbers tell a slower story. Deloitte says just 34 percent of organisations report using AI to deeply change products, processes or business models, while 37 percent describe their AI use as surface-level productivity gains with little structural change to how work actually gets done. The gap between those two sets of figures is the report's real subject: plenty of pilots, plenty of tool access, and a much smaller number of organisations that have moved past experimentation into anything that changes the underlying business. Deloitte attributes the stall to integration difficulty, governance complexity and infrastructure limitations rather than to any shortage of enthusiasm or budget. Deloitte's LEAP programme is built around that finding, organised into three stated pillars — scaling AI with value, with resilience and with confidence — aimed at helping clients move projects out of what the firm calls the proof-of-concept cycle: a pattern of launching pilot after pilot without any of them graduating into production because the governance and data-access questions were never resolved the first time around. Those questions are the same ones a cluster of smaller LEAP 2026 exhibitors built their own pitches around this year. ARIS argued that enterprise AI agents fail without process context — an understanding of how work actually flows through an organisation — while Master Works pitched consolidated strategy, PMO and compliance software as the missing governance layer beneath AI ambitions. Deloitte's report gives that pattern of vendor pitches some independent backing: the gap between adoption and transformation is real enough that a major advisory firm built a whole LEAP programme around closing it. The percentages themselves come from Deloitte's own survey methodology, which has not been independently audited by TechScoop, and the report is described as covering the Middle East broadly rather than breaking Saudi Arabia out as a separate figure. No named client engagements or contract values from Deloitte's LEAP 2026 presence have been published. TechScoop is reporting from LEAP 2026 in Riyadh.