Economic Research Forum (ERF)

The political economy of turning AI adoption into firm productivity in MENA

12
Generative AI can make individual tasks faster without making firms more productive. This column argues that across the Middle East and North Africa, turning AI adoption into better performance depends on organisational capabilities and on agreements within firms over authority, workloads and rewards. Public support can help by financing those capabilities through transparent selection, milestone-based assistance and independent evaluation.

In a nutshell

Generative AI can raise productivity in individual tasks, but firm-level gains depend on complementary investments in processes, data, skills and management – investments that remain limited in much of MENA.

Adopting AI redistributes authority, workloads and rewards inside firms; unless managers, workers and owners agree how adjustment costs and gains are shared, resistance can block the organisational change that productivity requires.

Public support should finance capabilities rather than licences: transparent selection, disbursement tied to verified milestones, competition reforms and independent evaluation of productivity, wages and working hours.

How should MENA governments support AI when purchasing technology is easier than changing how firms operate? Gatti, Islam and Torres (2025) identify weak private-sector dynamism, limited training and uneven competition as regional development challenges. AI policy must address these conditions if technological upgrading is to improve productivity.

In my earlier column, I examined MENA’s investment-to-productivity conversion gap (Yeganegi, 2026). Here, the question is who can change workflows, who bears adjustment costs and who captures the benefits. An investment can promise gains for the firm while threatening the authority, income or security of particular groups.

What the evidence establishes

Brynjolfsson, Li and Raymond (2025) studied generative AI assistance among 5,172 customer-support agents, finding a 15% average increase in issues resolved per hour. Less experienced workers benefited most. The study demonstrates potential within a service workflow, but it neither estimates gains for MENA nor isolates the contribution of organisational redesign.

Firm productivity means quality-adjusted output relative to the inputs used. An AI tool may shorten response times while adding verification, supervision and integration costs. The relevant question is whether the whole operation becomes more productive. Answering it requires attention to the capabilities supporting implementation.

Regional evidence points to limited complementary investment. Figure 3, Panel C of Gatti, Islam and Torres (2025), also published on the World Bank website, shows  formal training in 14.5% of firms in the MENA sample, compared with 30.1% in lower-middle-income and 34.3% in upper-middle-income economies. These figures use the latest available surveys for seven MENA countries, not a common survey year. They measure training provision, not AI adoption.

More direct evidence comes from Udomsaph’s (2025) exploratory study of Enterprise Surveys from 2022–2025 (Online Appendix Table A3, p. 6). In Bahrain, AI was identified in 6.9% of innovating firms, rising to 14.9% among innovators meeting stricter digital-readiness criteria. In Jordan, the corresponding shares were 1.5% and 4.7%. These criteria combine skilled employment and an online presence with fixed broadband and electronic payments. The percentages refer to innovators, not all surveyed firms, and should not be interpreted as economy-wide adoption rates.

The study identifies AI from firms’ descriptions of their innovations, so it may miss unreported uses. The country comparisons are descriptive and do not measure AI-induced productivity gains. The concentration within digitally prepared businesses supports diagnosing missing capabilities before subsidising purchases.

Organisational change redistributes costs and authority

Brynjolfsson, Rock and Syverson (2021) explain why general-purpose technologies require complementary investment in processes, organisational knowledge and human capital. These investments also redistribute authority and adjustment costs.

Managers may support an AI purchase yet resist changes that expose discretionary approvals or transfer decisions to frontline staff. For employees, sharing practical knowledge becomes less attractive if automation threatens jobs or intensifies monitoring without improving pay. Owners seeking rapid savings may leave workers to absorb the learning burden. Individually understandable responses can collectively obstruct improvement.

Tunisia provides a historical example. Rijkers, Freund and Nucifora (2014) found that firms connected to the Ben Ali family enjoyed especially large profit and market-share advantages in sectors subject to entry authorisation and restrictions on foreign investment. Their findings illustrate how incumbents can benefit from restricted competition. For AI support, the policy implication is that preferential access should reward capability-building, not connections.

The Tunisian evidence shows how incumbent interests can shape market rules, but the allocation of authority, adjustment costs and rewards within AI-adopting firms still requires firm-level diagnosis. Such diagnosis should identify bottlenecks and the interests sustaining them: who controls approvals, who supplies data and who bears responsibility for errors.

Before a supported trial begins, managers and employees should agree who can change the workflow, how training time will be paid, and how gains will affect workloads and rewards. Programmes could make such an agreement a condition of support, so that public money backs reorganisation rather than only the purchase. Such agreements can reduce uncertainty and make cooperation more credible. Training alone cannot resolve a dispute over who benefits.

Market incentives and public support

Gatti, Islam and Torres (2025) emphasise competition and equal treatment of private and state-affiliated firms. Applied to AI, protected revenues may allow firms to postpone disruptive changes. Competitive firms, meanwhile, may lack finance for integration and training. Policy must distinguish weak incentives from missing resources.

Subsidising licences alone can reward visible procurement while leaving organisational constraints untouched. Opaque grants may benefit well-connected incumbents over entrants. Suppliers have incentives to sell systems, while public agencies may prefer reporting purchases to waiting for measurable improvements.

Published eligibility criteria, transparent selection and time-limited assistance can reduce these distortions. Disbursement should follow verified milestones, including process diagnosis, usable data and staff preparation. Independent evaluation should assess outcomes beyond purchases. Competition reforms remain necessary where privileged market access weakens pressure to improve.

Priorities should follow firms’ capabilities. Shared advisory services could lower diagnostic costs for smaller businesses. In digitally advanced Gulf supplier networks, large buyers could help suppliers integrate systems, with open participation and safeguards against dependence on a single technology vendor. Where electricity, connectivity or digital records remain unreliable, those foundations may come first.

Evaluating public support

Enterprise-development agencies should pilot packages that address identified financing constraints or generate learning benefits beyond participating firms. Cost-sharing should be calibrated so that viable smaller businesses are not excluded.

Buying the software is only the beginning. Each trial needs a baseline and a specific operational problem. Evaluation should track quality-adjusted output per worker-hour, delivery times and total implementation costs, including staff time and rework. Tracking employment, wages and working hours would show whether gains reflect efficiency, intensified workloads or changes in benefit distribution.

Where feasible, transparent randomised phased entry among equally eligible applicants can provide a comparison group. Results should be reported by sector and firm size, while protecting commercially sensitive information. Scaling should depend on additional benefits relative to firm and programme costs.

MENA’s AI challenge is partly a negotiation over organisational change. Public policy can help by financing capabilities, reducing uncertainty over adjustment and limiting preferential access to support. Success should be judged by durable improvements in production and worker capabilities.

Further reading

Brynjolfsson, E., Li, D. and Raymond, L. (2025). Generative AI at work. The Quarterly Journal of Economics, 140(2), 889–942. https://doi.org/10.1093/qje/qjae044

Brynjolfsson, E., Rock, D. and Syverson, C. (2021). The productivity J-curve: How intangibles complement general purpose technologies. American Economic Journal: Macroeconomics, 13(1), 333–372. https://doi.org/10.1257/mac.20180386

Gatti, R. V., Islam, A. and Torres, J. (2025). Shifting gears: How the private sector can be an engine of growth in MENA. The Forum, Economic Research Forum, 27 May. https://theforum.erf.org.eg/2025/05/27/shifting-gears-how-the-private-sector-can-be-an-engine-of-growth-in-mena/ World Bank version, 3 June, Figure 3, Panel C: https://blogs.worldbank.org/en/arabvoices/shifting-gears–how-the-private-sector-can-be-an-engine-of-growt

Rijkers, B., Freund, C. and Nucifora, A. (2014). All in the family: State capture in Tunisia. Policy Research Working Paper 6810. World Bank. https://documents.worldbank.org/curated/en/440461468173649062/pdf/WPS6810.pdf

Udomsaph, C. (2025). AI adoption in MENAAP: Evidence from Enterprise Surveys. Research paper, Georgetown University, 20 September. Online Appendix Table A3, p. 6 (PDF p. 37). https://thedocs.worldbank.org/en/doc/e07a6825455ab28afc1867ec00482546-0280032025/original/87-udomsaph20250920-MENAAP.pdf

Yeganegi, K. (2026). From mega-investment to mega-productivity: Closing MENA’s conversion gap. The Forum, Economic Research Forum, 8 September. https://theforum.erf.org.eg/2026/09/08/from-mega-investment-to-mega-productivity-closing-menas-conversion-gap/

Tags: , , , , , , ,

  • Kamran Yeganegi
    Kamran Yeganegi
    Assistant Professor, Department of Industrial Engineering, Islamic Azad University

Most read

How trade policy can reduce MENA’s cereal import vulnerability

Heavy dependence on imported cereals, combined with climate change, water scarcity and geopolitical uncertainty, continues to threaten food resilience across MENA. This column explains how an inclusive trade policy can play a key role in making the region’s food security less vulnerable to shocks.

Digitalisation, global value chains and regional integration in MENA & SSA

Participation in global value chains is vital for countries pursuing structural transformation and inclusive economic development. This column summarises new evidence on how much production processes have been globalised in Africa and the Middle East relative to other regions; whether this process has taken place with partners within or outside the region; and whether it has taken place more in manufacturing or services.

The risks of a two-speed AI economy in MENA

Much of the debate about artificial intelligence in the Middle East and North Africa focuses on how many jobs it will replace. As this column explains, the more important challenge is that the new technology is likely to create a two-speed labour market, where countries with the capacity to invest in digital infrastructure and skills move rapidly ahead while others struggle to keep pace. Unless governments combine credible national reforms with targeted regional cooperation, AI could widen, not reduce, the economic divides across MENA.

Financing Lebanon’s Reconstruction: A Compact for Finance, Governance, and Accountability

As Lebanon embarks on another round of reconstruction, the key question is not how much financing can be mobilised, but whether each commitment is linked to a verified need, involves a transparent procurement process, is linked to clearly defined executing and operating institutions, and is based on a liability that is in line with the state’s ability to repay. This column proposes an operational mechanism for establishing that connection: a Reconstruction Finance and Accountability Compact.

Sanctions and conflict: why corruption changes the equation

Why do sanctions produce such different political outcomes across the countries that have been targeted? This column reports research examining the role of pre-existing corruption in shaping the relationship between sanctions and subsequent conflict or political instability. The evidence indicates that trade sanctions are associated with higher conflict-related deaths where pre-existing corruption is high.

The weak bread chain: cereal supply risks in six MENA economies

The Middle East and North Africa is one of the most food-import-dependent regions of the world, particularly for staple cereals such as barley, maize and wheat, and particularly in six countries – Algeria, Egypt, Jordan, Lebanon, Morocco and Tunisia. As this column reports, diversifying import sources has become an essential strategy for strengthening food security and improving resilience to geopolitical and climate-related disruptions.

From mega-investment to mega-productivity: closing MENA’s conversion gap

Across the Middle East and North Africa, governments are investing heavily in infrastructure, technology, diversification and new industries. Yet capital accumulation alone does not guarantee productivity growth. As this column argues, the region’s deeper challenge is its ‘investment-to-productivity conversion gap’: ensuring that large-scale investment translates consistently into stronger firms, technological diffusion, skills and better jobs. A different metric of success is required: not how much capital is deployed, but how much productive capability that capital leaves behind.

Why energy efficiency still struggles to attract finance in Egypt

Egypt has reduced energy subsidies and gradually adjusted domestic energy prices toward cost-recovery levels, improving the economic case for energy efficiency investments – but financing has not kept pace. Drawing on structured interviews with banks, developers and public officials, this column traces why energy efficiency projects in the country still struggle to attract capital, and sets out how a ‘financing ladder’ used elsewhere – from revolving funds to risk guarantees to green bonds – could close the gap.

Iran is still selling oil. But can it replace tomorrow’s barrels?

Iran’s oil exports have not disappeared, but continued sales may hide a deeper vulnerability. Arash Irani argues that selling barrels already in the logistics system is not the same as generating new export flows, and proposes five indicators to help policymakers judge whether trade under disruption is sustainable or merely running down its stocks.

The political economy of turning AI adoption into firm productivity in MENA

Generative AI can make individual tasks faster without making firms more productive. This column argues that across the Middle East and North Africa, turning AI adoption into better performance depends on organisational capabilities and on agreements within firms over authority, workloads and rewards. Public support can help by financing those capabilities through transparent selection, milestone-based assistance and independent evaluation.