Efficiency gains arrive quickly. Organizational transformation arrives much more slowly. In the Middle East, according to Deloitte, 66% of surveyed organizations report productivity and efficiency gains thanks to AI, but 84% have not redesigned the jobs or workflows that support them. This gap between adoption and reorganization reveals diverging trajectories across organizations.
The essentials
- The Deloitte survey observes distinct degrees of declared transformation with AI: adoption without job or workflow redesign, or adoption accompanied by overhaul of products, processes, or business models.
- In the Middle East, 66% of surveyed organizations report efficiency and productivity gains thanks to AI, but only 34% have overhauled their products, processes, or business models (Deloitte Middle East, 2026).
- All three declared approaches, including superficial use, are associated with efficiency and productivity gains; the survey does not establish causality or a timeline for overhaul, while business reinvention mainly characterizes the group deeply transforming its operations.
- The IMF estimates that AI offers an opportunity for productivity gains in Gulf economies, provided that risks of displacement, skills gaps, and labor reallocation are managed through active labor market policies, training, and targeted social protection.
- The distribution of gains and the terms of decision-making before adoption remain to be explicitly defined.
84% of organizations adopting without transforming
The Deloitte figures do not describe a lag. They describe a deployment architecture. Middle Eastern organizations have massively shifted from pilots to large-scale deployment: this is precisely what the 2026 edition of the report measures. But in 84% of cases, this deployment occurred without rethinking jobs or workflows. AI was grafted onto the existing organization.
This choice, deliberate or by default, produces a specific type of gain: efficiency in executing existing tasks. An analyst manages more files, a lawyer drafts faster, a customer service representative handles more tickets. Speed increases. The organization’s structure remains intact.
The 34% of organizations that have overhauled their products, processes, or business model pursue a transformation of a different nature than superficial adoption. The source distinguishes transformation approaches and publishes declared gains, without comparing their respective trajectories.
South Korea experienced an earlier version of this divide with industrial automation: the country had a robot density more than seven times the global average in 2020, but available studies do not establish a direct link between this automation and wage stagnation. The Middle East presents a similar dynamic: technology without reorganization tends to concentrate gains rather than distribute them widely.
Pre-distribution corrects what wage bargaining cannot reach
Economist Mariana Mazzucato has drawn a distinction that directly illuminates this gap. In her recent work on the common good economy, she argues that fiscal redistribution—raising taxes on companies capturing gains, financing training for displaced workers—arrives too late. It is necessary, but it corrects after the fact a distribution already fixed upstream.
“Pre-distribution” means redesigning the rules of value creation before the sharing. Control over data produced by employees using AI, the choice of which tasks to automate first, and the allocation of time freed by the tool must be decided before deployment, because these trade-offs influence how gains are split between companies and workers. Posed after deployment, these questions no longer find useful answers.
They must be posed before, inscribed in the very conditions of adoption.
The Deloitte data document limited job redesign around AI: in 84% of organizations, positions had not been redesigned around AI’s capabilities. Jobs were not rethought. Workflows were not redrawn. Efficiency gains concentrate in organizations and functions capable of capturing them quickly.
A competing reading exists, notably from economists close to Schumpeterian analysis of growth. Philippe Aghion and work on creative destruction argue that reorganization naturally follows adoption: companies that capture efficiency gains reinvest, create new jobs, and technological diffusion eventually lifts the entire labor market. The timeline differs, not the outcome.
This reading is not false. It describes well what happens when markets function smoothly, when professional mobility is high, and when companies compete for talent. But in Gulf economies, these conditions are only partially present. Labor markets are segmented between nationals and migrant workers, sectoral mobility remains limited, and large public enterprises, which carry weight in AI adoption, do not have the same incentives as firms exposed to competition. The Schumpeterian argument requires structural conditions that the Gulf has not yet built.
The tension between the two readings does not resolve in the abstract. It plays out in the institutions and incentives that regional governments will, or will not, put in place.
The 34% that transformed: a lever, not an exception
Organizations that have overhauled their processes and business models are not outliers. They are an indicator. They show that transformation is possible in the short term, in the same regulatory context, with the same market constraints.
Deloitte observes that some enterprises evolve systems and work organization together, and recommends anticipating operational constraints from the design stage of deployment. The 34% corresponds to organizations reporting the use of AI to deeply transform products, processes, or business models.
This distinction has a direct implication for public policy. Gulf states—Saudi Arabia with Vision 2030, the Emirates with their national AI strategy—are investing heavily to accelerate adoption. These investments are real and results in terms of deployment are measurable. But if incentives do not distinguish between simple adoption and adoption accompanied by reorganization, public funds mainly favor the first.
The IMF, in its 2026 assessment of AI prospects for the Middle East and Central Asia, emphasizes that the region has an opportunity for productivity gains, provided that risks of displacement, skills gaps, and labor reallocation are managed through active labor market policies, training, and targeted social protection.
The productivity stagnation affecting more mature economies, like Australia, where insufficient business renewal blocks the spread of technological gains, shows what this type of signal’s absence costs over the long term. Technology spreads, but aggregate productivity does not follow, because the organizations capturing gains and those stagnating do not renew at the same pace.
The decisions the Gulf can still make
AI is already transforming work in Gulf economies: 66% of organizations report efficiency gains. The question concerns the beneficiaries of this transformation and its modalities.
Two trajectories are plausible over a ten-year horizon, and current data allow us to distinguish them without ranking them as forecasts.
In the first, the 84% of organizations that have not rethought their jobs continue on this trajectory. Efficiency gains concentrate in already-performing enterprises. Workers, particularly low-skilled migrants who represent a significant share of the private sector workforce in the Gulf, remain confined to the same tasks, with no visible pathway toward roles redefined by AI. Adoption is rapid, professional mobility remains low. The gap between organizations that know how to transform and those that know how to adopt widens until it becomes a market divide.
In the second, regional governments condition part of AI-related aid and tax incentives on the production of a plan for transforming jobs and workflows. The 34% that have already transformed their processes become a documented model, whose practices can be examined, compared, and diffused. National observatories measure real professional mobility in sectors with high adoption, making visible the gap between declared gains and jobs actually created or upgraded. The signal becomes legible for companies and workers alike.
The gap between the two trajectories depends on both public incentives and measures, as well as on organizational capacities, skills, and complementary assets necessary for productive AI adoption. Conditioning part of public aid and incentives on criteria for documented transformation and monitoring of professional mobility could help align public incentives with these objectives. Aid for adoption that does not distinguish between reorganization and technical deployment risks financing approaches without structural transformation.
The signals to follow are precise. The 34% share corresponds to organizations beginning to use AI to deeply transform products, services, core processes, or business models; it constitutes an indicator to track. The evolution of upward professional mobility rates in sectors with high adoption will tell, within three to five years, whether efficiency gains have translated into individual trajectories or only into enterprise margins.
Collective needs that adoption alone does not cover
Both scenarios share a common deficit: workers need access to training that matches tasks redefined by AI, and this training must be available before reorganization is complete, not after. Waiting until jobs are redefined to train employees amounts to excluding them from decisions about their own work.
The second collective need concerns gain-sharing. Organizations reporting efficiency gains rarely publish the rules by which these gains distribute among shareholders, executives, and employees. Without explicit rules, distribution tends to follow existing power balances. This is exactly what pre-distribution seeks to prevent.
These two needs—anticipatory training and visible distribution rules—are not filled by AI adoption itself. They require collective decisions, public or negotiated, that precede or accompany deployment.
The Emirates and Saudi Arabia have the institutional capacity to set these conditions. They have the financial resources to fund training, and the regulatory levers to condition incentives. The window described by the IMF is real. It lasts as long as organizations lock in their adoption models. In economies that have already gone through a wave of automation, like public procurement in France, where innovation remains blocked due to lack of structural incentives, the moment for institutional decision closes quickly.
Measurement as a condition for mobility
A methodological remark deserves to be stated clearly. Available data measure declared efficiency gains and the share of organizations having transformed their processes, but not yet the actual professional mobility that results.
This is precisely why national observatories of mobility in sectors with high adoption constitute a concrete path, not a rhetorical suggestion. The absence of measurement of professional mobility limits evaluation of the real effects of adoption on individual careers. The 66% of declared gains will be presented as a success in both cases. Observing the difference between adoption creating mobility and adoption concentrating gains requires systematic measurement of professional mobility in sectors with high adoption.
Gulf governments have invested in ambitious AI strategies. The next step is deciding what they want to measure, and therefore what they want to encourage.
Sources
- Deloitte Middle East, State of AI in the Enterprise 2026, press release
- Mariana Mazzucato, The Common Good Economy: a new compass, Substack, 2024
- IMF, Harnessing Artificial Intelligence’s Potential in the Middle East and Central Asia, 2026 (no link, report available on the IMF website)
- Philippe Aghion, work on Schumpeterian growth and creative destruction (corpus reference, no single URL)



