Counter-positioning the AI Frontier: Lebanon's Playbook

This paper has also been published on substack here. If you prefer reading this paper in pdf, you can also download it for free here. What started as a conversation during my lecture on AI at the American University of Beirut - has evolved into this white paper. Here, Joseph Bakarji , Wassim Maktabi , and I try to answer the question “how should a developing country, like Lebanon, survive the AI revolution?" We hope this sparks a conversation, and action.

Executive Summary

Artificial Intelligence (AI) is transforming global economies at a pace that risks leaving developing countries behind. For Lebanon, a country mired in overlapping financial, fiscal, and governance crises, strategic investment in and adoption of AI offers a potential to accelerate economic recovery, and creates a path for relevance in a world where AI Frontier countries (Appendix 0) dominate. Developing nations that aim to join the ranks of AI leadership, will share in the economic benefits of AI's growth, influence global AI policies and discussions, and protect themselves from AI-related economic, cyber, and military threats. 

This paper argues that nations harboring AI Frontier labs and large compute resources will achieve “escape velocity” (Appendix 1, 2) - leaving developing unable to compete. We advise developing countries to counterposition by investing in AI talent and research that cannot be scaled through monetary resources alone. Using Lebanon as a case study, we describe how a country with limited resources and weak governance, can nurture native AI talent to produce key research of high economic and social viability.

To support our case, we expand on the second-order benefits to governance and economic prosperity from AI Native talent (Appendix 4). Our paper is anchored in endogenous growth literature, which posits that investment in human capital, and research niches, is the leading determinant for sustainable economic growth.. The paper concludes with implications for policymaking and development strategy, contending that in the absence of coordinated state capacity and sufficient political capital, higher education can serve as a catalytic entry point for inclusive, future-proof development.

Background

Worldwide, 72% of organizations have adopted AI in at least one business function (Statista). The speed of performance improvements has been equally significant, with AI coding benchmarks(i.e SWE-Bench Verified) seeing accuracy jumping from 4.4% to 71.7% in the past couple of years. Technology accessibility has also expanded through AI,  with the cost of AI usage reducing by 150x between early 2023 to mid-2024, according to OpenAI’s CEO (creators of ChatGPT) .

The MENA region has responded to this global AI diffusion with large-scale investments and state-led AI strategies. Saudi Arabia and the United Arab Emirates lead this transformation, allocating nearly $1 trillion to AI-related initiatives. More specifically, Saudi Arabia runs a $40B AI fund, in addition to its $20B AI investment in the US, while the UAE launched the $100B MGX firm, and $10B tech fund focused on AI (Semianalysis, May 2025). (See Appendix 3). This coincides with rapidly scaling its AI infrastructure, with Saudi Arabia accounting for 86% of the region’s data center capacity and leading a fivefold increase in compute since 2017. 

In Lebanon, however, the country is still grappling with a financial crisis deemed one of the past century’s most severe (World Bank, 2021). Since 2019, the local currency has lost 95% of its value, poverty has become widespread, and the unemployment rate exceeds 30%. This is exacerbated by infrastructure damage from past conflicts (2006 and 2024 Israel-Lebanon conflicts) and the 2020 port explosion. Recovery is hindered by fractured social cohesion, political polarization, weak institutions, and significant brain drain impacting vital sectors. 

However, certain endowments remain underutilized. These include the country’s multilingual and well-educated labor force, a high-skilled diaspora, and close proximity to GCC economies—a region in high demand for AI talent. In this vein, Lebanon holds potential as a complementary source of AI expertise—if its human capital is mobilized through targeted, systemic interventions.

The promise and peril of AI 

AI brings forward a critical inflection point in the trajectory of global development. Being a general-purpose technology, as significant as the internet, it holds the dual capacity to either mitigate structural divides between developed and developing economies, or exacerbate them—both at an unprecedented magnitude. This duality hinges not only on how broadly and equitably AI is adopted, but also on the speed at which economies and societies can build absorptive capacity. 

On one hand, the reduced cost of cognitive tasks and widespread availability of AI tools create opportunities for democratizing access to expertise, and, by extension, unlock productivity gains. And on the other hand, early-movers and concentrated infrastructure investments risks further widening the gap between lagging nations and developed economies. Whether AI becomes an engine of inclusive growth or a catalyst for deeper global stratification will depend on domestic policy priorities, rapid capacity-building, and strategic investment in human capital. 

AI’s promise for inclusion

AI is a democratizing force—for now. It is pushing the cost of expertise to near-zero. In 2024, GitHub reported that AI copilots were writing nearly half of all code on its platform. While in 2025, AI-generated code has become a dominant force in software development. Notably, Y Combinator (popular US-based Startup incubator) reported that 25% of startups in its Winter 2025 batch had codebases that were 95% AI-generated. Regionally, 65% of businesses in Saudi Arabia and the UAE recommend the use of AI at work (Hays Salary Guide, 2025).

With tools like ChatGPT, Cursor, Claude.ai, Replit, and Perplexity, individuals anywhere can access capabilities once reserved for elite institutions or wealthy corporations. Tasks generating business plans, building websites, analyzing legal or financial documents, or streamlining workflows, previously requiring years of training or a dedicated team, can increasingly be performed by a fraction of the resources with the help of  low-cost AI tools. For the first time, intelligence itself has become an accessible utility, and that shift is unlocking human potential in ways we are only beginning to understand.

Escape velocity: AI’s risk of widening the development gap

Despite the growing mainstreaming of AI, its adoption is largely concentrated in large enterprises, with McKinsey’s 2025 Global Survey showing significantly faster adoption in companies earning over $500 million annually (McKinsey, 2025). This trend is also mirrored by the country level, where wealthier nations are racing ahead in integrating AI across public services and infrastructure. Over $1 trillion has been earmarked across the richest five countries—led by the U.S., China, France, the U.K., and Saudi Arabia (Semianalysis, 2025). Meanwhile, all frontier labs (e.g., OpenAI, DeepMind, Anthropic, DeepSeek, Mistral) are situated in the world’s most advanced economies. 

This concentration raises critical concerns about a widening technological divide. The growing gap between societies building the AI frontier and those still idle will yield an “escape velocity” phenomenon, where lagging economies will be left behind, unable to benefit from the same economic growth, national security, and deterrence.

Furthermore, advanced models (i.e., AGI, ASI, etc.) will likely be limited to frontier lab host nations. Their potential as a cyberweapon makes them especially dangerous in the hands of a few, especially when developing countries lack any deterrence. Countries that lead in AI are likely to gain irreversible advantages, not only economically but militarily; deploying large-scale cyber operations and asserting technological dominance. (For further reading on a scenario of “escape velocity”, please see https://ai-2027.com/).

Opportunities in the Lebanese Context

AI as a Transformative Lever for National Recovery

In the absence of structural macroeconomic and state-building reforms, AI cannot be viewed as the singular panacea to Lebanon’s complex situation. Yet, out of all existing mega trends beyond geopolitics, investing in AI remains the most profound lever with the highest return. We argue that AI can enable Lebanon to recover at a lower cost basis by emerging as a digitized AI-native country. We highlight how this transformative opportunity is possible for Lebanon's recovery through radically impacting the key inputs of governance and development. (We expand on each area in Appendix 4).

1. Accelerating Governmental Digital Transformation: AI can significantly reduce the time associated with developing custom software and digitizing paper-based processes. It can bridge the gap between unstructured data and digital processes, facilitating a move to a fully digitized environment.

2. Cost Reduction in Public Services: AI can automate and streamline government operations, leading to substantial cost reductions and improved service delivery. Savings can be reinvested to further enhance digital infrastructure and talent.

3. Training and Up-Skilling Public Servants: AI coaching programs can improve the skills and efficiency of public servants. AI assistants can provide on-the-job training and generate training materials, enhancing the technical capacity of the government. (See Appendix 4 for a deeper dive into these three areas).

4. Zero-trust auditing & monitoring: The use of AI agents to audit and monitor public services can contribute towards lowering corruption and increasing transparency. Leveraging multi-modal agents that can read, hear, and see will allow the scaling of audits and monitoring to flag issues as a form of prevention.

Role of the Lebanese Government

The Lebanese government must take on a dual role: acting both as a customer of AI innovation and a supplier of valuable public datasets. The government must open access to structured, anonymized public data—such as census, health, land, and economic records—to catalyze applied research. As demonstrated by cases in Estonia and Singapore, public-private-academic (PPA) collaboration built on shared data infrastructure can dramatically accelerate innovation while improving state capacity. In Estonia, platforms such as Texta (Open-Source Text Analytics Toolkit), Hans (Speech Recognition for Parliamentary Records), and X-Road (Secure Data Exchange Platform) were all developed through government collaboration (e-estonia, 2019).

By actively commissioning digital transformation projects in health, taxation, education, and public utilities, the government creates a ready market for AI applications built by Lebanese graduates and startups. These early contracts offer both training opportunities and commercial validation.

Challenges in Government Execution and Structural Issues 

Governments have played central roles in setting and operationalizing their national AI strategy (Appendix 4) - this is more difficult to achieve in developing countries such as Lebanon. The Lebanese political and economic leadership are limited in their ability to effectively implement policies (i.e., in-fighting, nepotism, political deadlocks, corruption, lack of funds). We are concerned that over indexing on the government’s role in AI strategy, despite the immense benefit of AI in governance, lowers the likelihood of achieving meaningful progress in the AI field. The massive investments made by MENA frontier countries accurately represents the urgency - a developing country’s government will falter due to lack of funds, and ineffective governance.

Although we are encouraged by the current Lebanese government and non-governmental efforts (i.e., National Council for Entrepreneurship and Innovation’s AI Strategy). Multiple global agencies and local actors have already published AI strategies for Lebanon - but none have led to the urgent and large investment needed. Lebanon’s AI ministry’s five year roadmap (Appendix 6) to turn Lebanon into an AI-powered digital nation (LBC, 2025) will spend its first year on passing new legislation and digital rights. From an investment perspective, the minister of AI stated that $50M is needed to modernise core government services and restore public trust (National, 2025) - this is 10-100x less than counterparts in the MENA region are spending on AI alone. Even when reviewing reports published by other organizations we find a similar theme (See Appendix 7). Reliance solely on the government to provide the necessary urgency and leadership for Lebanon to achieve AI relevance is unlikely to yield timely or sufficient progress. A more agile and resource-efficient approach, driven by educational institutions and private sector innovation, is essential to counter-position Lebanon in the rapidly evolving AI landscape.

Lebanese Educational Institutions in Counter-Positioning the AI Frontier.

Counter-positioning is a strategy that involves adopting a novel approach that incumbents cannot easily replicate. This is due to the conflict between the new approach or model and their existing economic or strategic choices - thus creating a sustainable advantage. This strategy allows developing countries, such as Lebanon, to avoid direct competition on the same terms (i.e., via compute capacity), and instead carve out unique, defensible niches. 

We propose Lebanon to achieve this counter-positioning through focusing on key research areas that are particular to the country’s context (low data, compute, technical talent) and that are at odds with the research agenda of the AI Frontier. This will enable the country, as a result, to produce a defensible IP base, nurture local AI talent, and capture the economic benefits of an AI revolution. Naturally, the country’s Educational Institutions play an outsized role in achieving this strategy. Hence the next few sections will focus deliberately on this role.

Lebanon's Research Niches

Lebanon is unlikely to compete in raw compute, but it can contribute uniquely to global AI advancement by targeting underexplored and high-leverage research niches. Economic impact from AI innovation originates from research labs. OpenAI, the creators of ChatGPT, were started purely as a research laboratory rather than a tech corporation. To capture a share of the economic lift from AI, Lebanon must invest in AI research - and this should be instigated by private universities. In this section, we highlight multiple areas of interest that have real-world applications across domains. This type of research will not always lead to profitable applications. Rather, a single breakthrough, from a large body of research, can lead to significant economic impact.

Low-resource data, training, and inference: Students, researchers, and tech entrepreneurs in Lebanon are unlikely to access the scale of GPU resources, or petabytes of curated data available to AI Frontier nations. Instead, we propose turning exactly these constraints - limited compute, and data - into a technical focus area. Available data sources in Lebanon are heterogenous, complex, scarce, and unique to a local context (i.e., Ottoman era hand-written records for local urban planning). Researchers can pioneer methods for generating data from low-data sources, training and deploying models in environments with limited computational or data resources. This area can unlock the ability to serve on-device models, open new commercial niches, and enable other “Global South” economies to achieve AI breakthroughs.

We draw inspiration from compact and elite research teams (i.e., DeepSeek), who proved that they can optimize reinforcement learning systems with more efficient data and infrastructure resourcing, focusing on low-cost training regimes for agentic AI systems.

Arabic contextual red-team and evaluation hub: A key factor in developing the AI frontier is the ability to test, and uncover misalignment within a system or a model. This capability is also of interest for businesses and labs regionally looking to be supported with evals that ensure their ability to achieve business impact. We propose research into red-teaming and eval for Arabic and niche dialect models - beyond the mainstream. This can be an area of deep collaboration with labs in the MENA region.

Context specific domains such as Arabic cultural, and artistic heritage: Instead of investing in building Arabic-speaking chatbots, we propose research into context-driven domain rather than language and dialect alone. Local context in arts, music, literature, history, and archaeology, all open up new modalities (i.e., archaeological site scans), data-sources (i.e., oud sheet music) for applied AI. These rich domains are targets for research that lean into the Lebanese niche.

Applied multi-modal agentic systems  for digital transformation:  The 2025 wave of GPT-4o, Gemini 2.5 Pro/Flash and open-source Layout-aware models (e.g., LayoutXLM, LLaVA-Next) shows that a single vision-and-text backbone can now classify, extract and reason over messy PDFs, photos of forms and voice notes on commodity GPUs, beating classic OCR in both accuracy and speed. Lebanon could position itself as the MENA test-bed for “multimodal e-government”. Regional startups (i.e., Zink AI) have proven that fine-tuning these models can offer value to local governments. Lebanon can leverage Arabic and French civil-registry scans to fine-tune and  open-source adapters so that municipalities and startups can plug them into low-code RPA stacks. Success here would generate reference datasets and tooling that every data-poor administration across the Global South needs.

AI for fragile state governance: Developing systems for public service delivery, fraud detection, and decision support in politically and institutionally fragile settings. Building on recent OECD “States of Fragility 2025” diagnostics, Lebanese researchers could prototype open-data audit agents that review government and local authority spending, detect fraud and corruption, while maintaining open forum for the public to participate.

Positioning Lebanese Universities as an AI Anchor

Lebanese universities must urgently cultivate AI-native talent to maintain their historical leadership. Prioritizing AI-first programs or new AI-focused faculties is crucial. Lebanon's resource-constrained environment can drive innovation by focusing research on low-resource model optimization, multi-modal techniques, and deep-reasoning systems. By emulating successful small teams in similar settings, Lebanese academia can target high-impact areas like adaptive learning, civic AI, and localized solutions for healthcare, education, and financial inclusion.

To succeed, universities should:

1. Tap into the Lebanese AI diaspora to design a world-class curriculum based on key domains such as machine learning engineering, reinforcement learning, applied NLP, and AI ethics in low-resource contexts.

2. Embrace the need to continuously improve and update learning material as the new normal.

3. Launch targeted recruitment campaigns aimed at the Lebanese scientific diaspora and international visiting scholars, with incentive packages co-funded by industry and philanthropic partners.

4. Develop a student pipeline with a focus on regional diversity and inclusion, offering needs-based scholarships, AI bootcamps, and bridge programs to support students from underrepresented regions or academic backgrounds.

5. Secure funding from philanthropic, institutional, and regional partners through joint research initiatives, GCC industry fellowships, and multilateral development agencies.

Curriculum, Faculty, and Talent Strategy

Universities must evolve their curriculum in partnership with leading AI practitioners from the global Lebanese diaspora and regional industry leaders. Curriculums should include modular learning tracks in core AI subfields—machine learning, systems architecture, and AI ethics—as well as applied domains such as digital governance, health tech, and financial automation. We warn that seeking to hold a snapshot of the current state of the art into a curriculum with content that does not change will be a failure. For example, hundreds of Prompt Engineering courses in 2023 have become irrelevant in 2025 (FastCo, 2025). Even programs that are centered around specific techniques or model architectures can become irrelevant overnight (Stackoverflow, 2024). 

A global search for academic talent in AI should prioritize candidates who apply artificial intelligence to domain-specific knowledge; such as AI for bioengineering, physics, medicine, the arts or the humanities. Many academic disciplines are on the verge of transformation, and leading experts who can adapt will not only revamp existing curricula but also anticipate how their fields will evolve and shift their research accordingly. At the same time, the search should include exceptional candidates advancing the frontiers of machine learning theory and its foundational frameworks, knowing that innovative solutions come at the intersection of disciplines.

While we recognize that not all leading candidates will be willing to relocate to Lebanon, many are open to returning or contributing under the right conditions. The university should cultivate an environment that inspires creative thinkers to be at the forefront of global innovation. This requires providing sufficient resources for international collaborations and building a critical mass of talent that attracts more talent. The message should be clear: what is possible abroad can also be done in Lebanon.

In the short term, we recommend establishing a robust visiting lecturer program that draws from both industry and academia, targeting experts working at the intersection of AI and other fields. These visiting scholars can contribute through seminars, workshops, and modular teaching, helping build a vibrant community and stimulating collaborations on locally relevant projects. 

In the medium term, existing faculty should be empowered—with time, training, and resources—to rethink how AI tools can be integrated into their teaching. As many current jobs may disappear or be radically altered in the coming years, our educational programs must adapt, ensuring students are grounded in the fundamentals while also gaining the capacity to use AI to augment and accelerate their expertise.

Ultimately, the long-term goal is to create an agile, scalable ecosystem; a hub that empowers other faculties with AI capabilities, fosters interdisciplinary innovation, and drives a new model for research and education in the age of intelligent systems. AUB has already taken important steps in this direction with the launch of the School of Computing and Data Science (AUB, 2025). Now is the time to build on that momentum.

Securing Compute

While sovereign data centers are a rising trend, Lebanon is not positioned to support them today due to infrastructure vulnerabilities and energy instability. Instead, we recommend leveraging regional compute infrastructure and cloud-based platforms through partnerships. Feasible options include joining existing Gulf-based initiatives such as Saudi Arabia’s AI cloud consortiums, partnering with the UAE’s G42 and MBZUAI for academic compute credits, and collaborating with regional hyperscalers such as STC Cloud or e&, which have both infrastructure and policy alignment with Lebanon’s needs. To mitigate risks, Lebanon must ensure that data residency and compliance protocols are built into these partnerships to address data sovereignty concerns and maintain control over sensitive information.

This approach ensures that Lebanese AI researchers and startups gain timely access to high-performance computing resources, while simultaneously building capacity for future localized infrastructure as the country stabilizes.

University-Startup Collaboration

Since AI lowers the barriers to build and launch technological products, universities are ideally positioned to become hotbeds for new ventures. With robust research programs and integrated AI education, academic institutions will increasingly be sources of applied AI innovation in Lebanon, surpassing traditional industry avenues. This university-driven innovation provides multiple benefits. It facilitates transformative solutions for Lebanese enterprises, delivers essential technological services to the Lebanese government, and creates immediate employment opportunities for university graduates as ventures grow. Crucially, strong ties between universities and startups create essential feedback loops, allowing academic researchers to refine their programs based on practical insights and real-world experiences from applied AI projects. Strengthening this collaboration is essential to fully harness Lebanon's AI potential.

Conclusion

In conclusion, Lebanon stands at a critical juncture. Despite facing significant socio-economic challenges, it possesses unique endowments, particularly its human capital, which can be leveraged to not just survive but lead in the AI revolution. By strategically counter-positioning itself, focusing on niche research areas tailored to its constraints, and repositioning its universities as regional AI anchors, Lebanon can unlock substantial opportunities. Embracing AI as a transformative lever of recovery will enable Lebanon to emerge as a digitized, AI-native nation, capable of addressing its immediate challenges while also securing a competitive edge in the global landscape. The journey demands urgency, collaboration, and a commitment to continuous learning, but the potential rewards for Lebanon and the wider Global South are immense.

Appendix

Appendix 0. AI Frontier Countries Definition. AI Frontier Labs are the leading research groups that develop and release the highest performing models based on publicly available benchmarks. At the time of writing this whitepaper, some of these labs are OpenAI (USA), DeepMind (UK), Anthropic (USA), DeepSeek (China), Meta (USA), and Grok (USA) (LLM Leaderboard, Vellum). We refer to AI Frontier Countries as those that host these labs.

Appendix 1. Definition of Artificial Intelligence (AI). Artificial Intelligence (AI) refers to computer systems and algorithms that perform tasks typically requiring human intelligence. These include learning from data, recognizing patterns, making decisions, understanding language, and generating content. Modern AI ranges from narrow systems (like chatbots or image classifiers) to large models capable of coding, reasoning, and multi-modal understanding. AI enables machines to adapt, improve performance over time, and automate complex cognitive tasks.

Appendix 2. Escape Velocity. Escape velocity in this context refers to a scenario where certain countries, particularly those with advanced AI labs and ample computational resources, achieve such rapid and accelerating progress in AI development and adoption that they create an insurmountable gap with other nations. Essentially, these leading countries move so far ahead, at an ever-increasing pace, that those lagging behind are unable to catch up, effectively being "locked out" of the next wave of technological and economic development. This creates a perpetual disadvantage and deepens global stratification.

Appendix 3. GCC AI Investments. Major investments, like Qatar’s $550M expansion with Nvidia and the UAE’s AI supercomputing initiatives, are enhancing regional AI capabilities. Additionally, open data efforts in the UAE are supporting AI research and innovation across sectors.  AI adoption is accelerating across key sectors—banking, healthcare, and IT—where it's projected to contribute significantly to GDP (est. 12-15%) and regional tech spending.

Appendix 4. Examples of MENA Government’s role in AI strategy. The Saudi National Strategy for Data & AI (SDAIA) aims to make Saudi Arabia a global leader in AI and data-driven innovation by 2030., focusing on building a skilled ecosystem, attracting investments, and advancing key sectors like healthcare, education, and energy.

While in the UAE, the Ministry of State for Artificial Intelligence has set the UAE's National Strategy for Artificial Intelligence 2031  (UAE AI 2031 Strategy) - a plan that aims to position the country as a global leader in AI by expanding its assets, developing a robust AI ecosystem, and attracting international talent, with a focus on sectors like healthcare, energy, logistics, and tourism.

Appendix 5. Deep dive into the benefits of AI in the Lebanese Government.

1. Accelerating Governmental Digital Transformation.

We believe that AI will deliver the digital transformation needed by lowering the cost of custom software, and translating paper into bits. AI can bridge the gap between unstructured data and digital processes, facilitating a smoother transition to a fully digitized environment that meets contemporary needs. This digital and computational transformation is long overdue. Building custom software, digitizing paper processes into simple processes, and even making the use of these tools widely available and easy to use is radically cheaper and faster. See Appendix 3 for existing examples of digital transformation in Lebanon. These concrete projects show that Lebanon already has the datasets, the tools and the talent to leapfrog from paper files to API-ready information pipelines that can power both public-service apps and data-hungry startups.

2. Radical Cost Reduction in Public Services.

Deloitte estimates conversational AI can free 1.2 billion government work hours annually. AI can further automate, streamline, or eliminate unnecessary government operations. By using AI to optimize services, the government can reduce operational costs significantly. This shift not only alleviates budgetary pressures but also improves service delivery to citizens. The savings feed the “Lebanese AI Flywheel,” reinvesting in faster connectivity, richer data, and talent upskilling—creating a self-reinforcing cycle that accelerates national recovery and positions Lebanon as a regional AI leader. See Appendix 4 for examples of how developing countries have leveraged AI to create savings in public services.

3. Educating and Up-Skilling Public Servants.

Well-designed AI coaching programs have demonstrated significant positive outcomes in international pilot studies. For example, a 2024 GovTech-Singapore study on government software teams using GitHub Copilot showed a 21–28% increase in coding speed and a 95% rise in developer satisfaction, with junior staff experiencing the most rapid skill development (Ng et all, 2024)t. Singapore's "Pair" chatbot quickly gained widespread adoption, attracting 11,000 officers across over 100 agencies within two months and maintaining approximately 4,500 weekly active users (GovTech). This highlights that when AI tools are reliable and secure, they will be utilized for self-directed, on-the-job learning.

Lebanon's public services have suffered from years of nepotism and corruption. AI now offers an opportunity for public servants to enhance their skills, learn new abilities, and increase their efficiency beyond basic task automation. AI assistants can function as "on-the-job" trainers, providing advice and even generating training materials at scale. A key area for improvement is the technical capacity of the Lebanese government, which AI "co-developers" can help strengthen. See Appendix 6 for risks and mitigations of digital-skill shortage.

4. Improved Governance and Public Outcomes.

Global pilots already show how AI can tighten oversight and speed up audits. An OECD survey of 59 audit- and integrity-focused agencies found that the single most valued use case (ranked first by 37 % of respondents) was automating evidence-gathering and document review, freeing auditors for higher-order judgment (OECD, 2024). Brazil’s Federal Court of Accounts operationalised this insight with “ChatTCU,” an LLM interface that lets 1,400 staff instantly summarise case files and surface anomalies during live audits (OpenData, 2024). Likewise in the UK, the Cabinet Office’s new “Humphrey/Consult” engine processed 2,000 consultation responses as accurately as human analysts and is projected to save 75,000 civil-service work-days and about £20 million a year once rolled out across 500 consultations (Gov.uk, 2025).

To scale such gains, governments must first build the right data and talent backbone. The OECD’s Governing with AI report stresses that trustworthy public-sector AI “relies on high-quality, shareable government data” and calls robust, whole-of-government data-governance frameworks an essential enabler (OECD, 2024). Open-data programmes, such as the US’s  Data.gov, has shown robust growth - an increase from 47 datasets in 2009 to more than 310,000 in 2025, and attracts +1M monthly page-views from researchers and entrepreneurs (Data.gov, 2025). See Appendix 5 for additional data on public service savings in global economies.

In addition to digitization, AI tools can enhance governance by improving audits, and decision-making based on data-driven insights. By further digitizing public datasets and making them accessible, the government can support research and innovation that addresses core issues like healthcare, infrastructure, and public welfare, thereby improving the quality of life for citizens.

Appendix 6. Lebanon AI Ministry LEAP Plan. In the first step, “Launch,” plans to establish the ministry as a leading entity with initial digital AI programs. The “Enact” step in the first year involves deploying regulatory frameworks, partnerships and investments that can enable wider adoption and utilization of AI. For the next three years, Dr. Shehadi’s vision is to “Advance:” Position Lebanon as a regional leader in AI and technology policy with scalable investment growth and sector-wide AI adoption. The last step, “Promote,” hopes to realize a fully operational, tech-enabled economy whereby Lebanon is a recognized regional hub for AI talent and innovation, in five years’ time. (LAU, 2025)

Appendix 7. Other Lebanese AI initiatives and publications. Lebanon National Center for Electronic Innovation in Lebanon (NCEIL) launched a program called 9816 AI that also focuses on policy making and monitoring. While the UN ESCWA AI Strategy for Lebanon has not been updated since 2020 - 2 years prior to the advent of OpenAI’s GPT 3.5 and ChatGPT.

Appendix 8. Examples of Digital Impact in Lebanon. Volunteers and academics are already proving how quick wins are possible. AUB Libraries and its School of Computing are digitising more than 150,000 manuscripts, photos and maps and layering AI-based Arabic OCR so the content is searchable and ready for new machine-learning models [AUB Libraries]. Beirut Madinati has posted a 40-MB, fully timestamped archive of its protest-era tweets so researchers can mine engagement patterns and geotagged images—an instant proof of concept for structuring “messy” civic data [AUB ScholarWorks].  In 2020, Nonprofits (i.e. Open Map Lebanon) led a data-driven recovery effort for Lebanon with data ingestion pipelines and interactive visualisations that guided on-the-ground efforts more efficiently (OpenMapLebanon).

Appendix 9. Use of AI in Developing countries. AI-led digitisation is already slashing public-sector costs and upgrading infrastructure worldwide. India’s biometric-AI Aadhaar platform eliminated millions of ghost beneficiaries and saved about US $12 billion. Estonia’s open-source X-Road backbone automates 99% of inter-agency data exchange, worth 844 work-years each year.; Brazil’s tax authority boosts recovery rates with machine-learning fraud detection; and Ericsson shows AI network-optimisation software raises telecom capacity 15–20 % while halving fault-resolution time. Adapting these proven tools to Lebanon would let a lightweight e-ID layer purge patronage from cash transfers, an X-Road-style fabric collapse queues and bribes, Arabic/French chatbots deliver 24/7 services without new hires, AI risk scoring recapture much of the country’s ≈US $2 billion tax gap, and smart-network software lift user speeds immediately without heavy capital outlays. 

Appendix 10. Better data translates into social value. The World Bank’s Digital-in-Health review estimates that interoperable health data can cut public-health costs by up to 15 %—savings already being released in large tele-health roll-outs (World Bank, 2024). Yet both the OECD and G7 reports flag that shortages of specialised AI and data-engineering skills as the top implementation barrier, ahead of regulation or culture. In short, pairing open, well-governed data with a cadre of in-house AI engineers is the prerequisite for AI tools that strengthen audits, deliver data-driven decisions and—ultimately—raise citizens’ quality of life. 

Appendix 11. Risks and mitigation of Digital Skill Shortage. Global evidence indicates that productivity gains from AI can be lost if digital skills gaps are not addressed. The 2024 G7–OECD Toolkit identified "digital-skills shortage" as a major obstacle to AI implementation, recommending that governments create AI-engineering and safety roles within each ministry (OECD, 2024). Similarly, a World Bank governance note concluded that administrations lacking internal technical expertise risk becoming overly dependent on opaque AI systems, potentially leading to service decline (World Bank, 2024). The suggested solutions include establishing central "AI hubs" and actively recruiting or upskilling specialized engineers and data stewards before large-scale AI deployment.

About the authors.

Sherif Maktabi is an award winning AI Product Leader and advisor. Currently, Sherif is a Sr. Principal Product Manager for Core AI at UiPath. Previously, Sherif founded an AI tech team at Amazon - delivering millions of dollars in incremental revenue from scaling AI-driven products, and launching the company’s largest AI conversational agent. Previously, he was Head of Product at Michelin Connected Fleet, where he successfully pivoted the group toward electric vehicle fleet solutions. 

Joseph Bakarji is an Assistant Professor within the AI, Data Science and Computing (AI-DSC) Hub, and the Mechanical Engineering department at the American University of Beirut. He spent 3 years as a postdoctoral fellow at the AI institute in Dynamic Systems, University of Washington where he developed machine learning algorithms for scientific modeling. Joseph received his PhD at Stanford University in 2020, where he developed multiscale stochastic models for granular materials. He also heads the music intelligence lab at AUB that develops innovative technology at the intersection of artificial intelligence, music instrument design and music composition and performance.

Wassim Maktabi is an economist and public policy researcher covering the MENA region. His work spans the political economy of reform, labor economics, and public finance and has been published by leading media outlets, international development institutions, and academic journals. Wassim previously helped build Lebanon’s first public tool that uses legislation data to assess what policymakers prioritize and currently serves as a Research Manager at a management consulting firm.