Boris Cournède
OECD
OECD Economic Surveys: Israel 2025
Abstract
Israel is well positioned in the global AI market with a strong presence especially in the development and marketing of software. The regulatory approach has been nimble, creating favourable conditions for AI experimentation and growth while protecting privacy and trust in AI systems. Further investment in scientific research and higher education would reinforce the sector. Action to bridge the large gender gap in AI is key to address shortages in the kind of highly skilled labour that is critical to the continued dynamism of the sector. There is also scope for greater participation of currently under-represented population groups, such as ultra-orthodox and Arab Israelis, in the AI economy, through ensuring that young members of these groups have access to high-quality education in AI-relevant subjects. Progress along these dimensions would also facilitate the deployment of AI across the economy, which has been sluggish. For this purpose, continued investment in connectivity infrastructure is a pre-requisite.
The promise of artificial intelligence (AI) is particularly great for Israel’s economy in terms of both production and use (Box 2.1). Israel’s thriving high-tech sector is well placed to expand its already strong AI-creation activities. On the use side, the diffusion of AI holds significant potential to help traditional industries and government, which have long lagged behind high-tech firms in terms of productivity, to increase their productivity. As such, if properly mobilised, AI can further buttress the high-tech sector while reducing duality in the economy.
The OECD AI Principles (OECD, 2019[1]), since their May 2024 revision, define an AI system as such:
An AI system is a machine-based system that, for explicit or implicit objectives, infers, from the input it receives, how to generate outputs such as predictions, content, recommendations, or decisions that can influence physical or virtual environments. Different AI systems vary in their levels of autonomy and adaptiveness after deployment.
AI systems are constructed using machine learning methods or logic and knowledge-based approaches. They have applications in areas such as computer vision, natural language processing, speech recognition, intelligent decision support systems and intelligent robotic systems. The tasks or functions performed by AI systems include, but are not limited to: recognition (identifying and categorising data, e.g., image, video, audio and text), event detection (connecting data points to detect patterns, as well as outliers or anomalies),forecasting (using past and existing behaviours to predict future outcomes),personalisation (developing a profile of an individual and learning and adapting its output to that individual over time), interaction support, finding the optimal solution to a problem, inferring new outcomes that are possible even if they are not present in existing data, and generating new content (OECD, 2024[2]).
Producing as well as using AI both require specific skills, capital and a regulatory framework. Through AI production, the country’s high-tech sector is delivering new output in an area characterised by large global demand. The deployment of AI offers potential for considerable gains in productivity, especially so if labour and capital can reallocate smoothly across occupations and sectors (Filippucci, Gal and Schief, 2024[3]).
This chapter investigates ways for Israel to gain the most from AI. It first looks at how Israel can expand an already strong AI production sector considering its current labour resources, skills base, physical digital infrastructure and regulatory stance. The chapter then investigates possible interventions to accelerate the take-up of AI throughout the rest of the economy.
2.2.1. A buoyant AI sector is becoming a central part of Israeli high-tech activity
Israel is very well positioned on the global AI market. The country ranked 9th in the world for its level of investment, innovation and implementation of AI in a recent benchmarking exercise (Tortoise, 2024[4]). The AI sector is very dynamic, with strong investment and intense startup creation that make Israel a global player in AI innovation on a par with larger economies (Figure 2.2).
Note: The investment data refers to companies, both listed and unlisted.
Sources: OECD AI Observatory; and Stanford AI Index Report 2024.
AI has become a core part of new high-tech activity in Israel, accounting for almost half of all new high-tech startups and funding rounds (Figure 2.3 Panel A). This trend makes AI critical to the continued performance of the high-tech sector, which itself is a key engine of the country’s export and GDP growth. High tech, which employs 12% of the workforce to produce 20% of GDP, accounted for 40% of 2018-2023 GDP growth (Israel Innovation Authority, 2024[5]). The strong outward orientation of the high-tech sector, which generated 53% of exports in 2023, gives Israel a strong basis to participate in the global AI boom (Israel Innovation Authority, 2024[5]).
The AI sector is concentrated in software (Figure 2.3 Panel B). A significant part of software activity relates to cyber-security, an area where the potential of AI is considerable and Israel highly regarded. Reflecting Israel’s well-established strength in high-tech medical and pharmaceutical products, life sciences are another area of significant commercial AI activity. The same holds for agricultural innovations, especially in the use of AI for advanced irrigation systems. Israel also hosts one of the 18 companies worldwide that create foundation models and commercialise associated services (Israel Innovation Authority, 2024[6]).
The ongoing conflict has two-sided effects on AI prospects. On the one hand, many AI professionals have had to join defence forces for reserve duty. Furthermore, agritech and foodtech suffered from the attacks over Israel: a number of experimental fields and installations burned in the south in October 2023 and the north in mid-2024. On the other hand, strong investment in defence-oriented research and development in response to the needs created by the war are likely to generate innovation and spur prospects for the country’s AI sector in the areas of computer vision, automation and cyber defence. The high-tech R&D ecosystem has generally been tightly integrated with defence research (Gandal, Roccas Gandal and Kunievsky, 2021[7]).
Source: IVC data processed by RISE (2024[8]).
2.2.2. Upscaling computing infrastructure will support AI-production growth
Government bodies as well as private observers have identified computing infrastructure as a weakness for the country’s AI sector (Tortoise, 2024[4]; Ministry of Innovation et al., 2024[9]; State Comptroller of Israel, 2024[10]). Latest AI developments, notably generative AI, involve large-scale models that require enormous numbers of computations (OECD, 2024[11]). AI development is anticipated to continue to depend on large-scale computing infrastructure (OECD, 2023[12]). Locally available computing power may be acting as a brake on the development of very large models. As of July 2024, Israel was home to one out of the 125 models trained with more than 1023 floating point operations (FLOP). This model, Jurassic-1-Jumbo, ranked 70th at 3.7.1023 FLOP, two orders of magnitude below the 2.1.1025, 3.8.1025 and 5.0 1025 FLOP underpinning the world leaders GPT4, Llama 3.1-405B and Gemini 1.0 Ultra (Epoch AI, 2024[13]).
More local capacity would be useful despite the international integration of IT systems. For AI model developers, buying computing power from abroad, including from cloud service providers, is an option that offers only an imperfect substitute for local capacity because of the associated need to transfer the corresponding data for processing. This transfer can be costly technically, because of the large sizes involved, as well as legally, because of the interaction between domestic and foreign regulations. Plans by a multinational tech corporation to invest $7.2bn in cloud infrastructure in Israel however open the way for access to large computing capacity with much lower access times than foreign server farms (OECD, 2024[14]).
The authorities are preparing to build an AI High-Performance Computing laboratory as part of the NIS 1 billion (USD 270 million) National AI Programme 2024-2027. The objective is to expand local capacity for academic institutions, startups and emerging AI companies. Speedy completion of this project in a difficult budgetary environment is key to ensure that academic research remains at the frontier and that early-stage startups have access to computing capacity to test-run ideas. The extent to which the project fulfils its objectives and is of sufficient scale should also be regularly reviewed to expand it if necessary.
Country breakdown of all known models trained with more than 1023 floating point operations, 2024
Notes: The data were retrieved 25 July 2024. *OTH stands for "other”.
Sources: Epoch AI (2024[13]) data and OECD calculations.
AI computing entails large and rapidly growing power requirements. For instance, the development (in the United States) of OpenAI’s GPT4 required an estimated 60GWh of power (Groes and Ludvigsen, 2023[15]), which is equivalent to the yearly electricity consumption of 6000 Israeli households. Electricity demand from AI is projected to increase tenfold from 2023 to 2026 (IEA, 2024[16]). Power supply is available in Israel at prices that are lower than in EU countries though higher than in the United States. Electricity wholesale costs are around USD50 per MWh in 2024, levels above the United States (USD30) but well below Germany (USD100) or the United Kingdom (USD90). As discussed in Chapter 3, electricity is mostly produced from natural gas, which is abundant since the discovery of large offshore gas fields. Power supply is therefore available to develop and use AI in Israel, but expanding AI while meeting Israel’s greenhouse-gas emission targets will require carbon-free power (see Chapter 3). Locating AI computing and data centres in the north and south of the country will help to reconcile AI-expansion and decarbonisation objectives, as these areas have more abundant renewable energy with surpluses during daytime.
2.2.3. Private funding, strong at early stages, will also need to accompany AI firms’ continued growth
Privately funded investment is central to the growth of the AI sector. The country attracts high volumes of venture capital investment into AI (Figure 2.2 Panel A). Faced with the risk of a dearth of funding for high-tech startups following the 7 October 2023 attack, the Israel Innovation Authority and the Ministry of Finance launched “fast-track” grants at the end of October 2023 (Israel Innovation Authority, 2024[5]). This programme was aimed at early-stage high-tech startups (including but not limited to AI) with a survival horizon below six months. Looking beyond the emergency, a Startup Fund (of 0.5bn NIS in 2024 or $133mn) was launched to fund firm creation in fields involving high-risk R&D. The authorities should monitor the results to evaluate the case for expanding the programme.
The wide-scale computing requirements of many areas of AI development, including generative AI, mean that, for successful AI firms to grow, mature-stage and long-term funding is needed in addition to seed financing. This can be a challenge in Israel, where the high-tech sector is characterised by vibrant start-up activity but a comparatively low base of long-term capital. In April 2024, the Israel Innovation Authority launched an initiative (Yozma 2.0) to attract institutional investors by providing an additional yield to entities that participate in the programme. While this programme is primarily aimed at expanding the investor base for venture capital, it also offers opportunities to create links between long-term investors and high-tech companies which can then facilitate their medium and long-run funding. It will be valuable to assess the results of the programme and evaluate if it is worth adjusting so that it also encourages medium and long-term investment alongside venture capital.
Key to the funding of AI as well as other innovative firms over time is a favourable investment climate. The high country-risk premium creates a hurdle for long-term investment. As developed in Chapter 1, fiscal adjustment alongside monetary prudence can contribute to lowering the risk premium. Supporting framework conditions are essential for AI or high-tech-specific policies to bear fruit.
2.2.4. A broader talent base is key to further AI growth
The strength of Israel’s AI sector largely rests on specialised advanced skills. The AI sector is however facing intensifying labour shortages (Ehrlich and Mekonen, 2024[17]). One indirect indicator is that, over the first seven months of 2024, wages for scientific researchers-and-developers, on whom AI firms rely particularly strongly even if the statistic refers to those active across the whole high-tech sector, were up 11% against 7.3% for other high-tech workers and 4.6% for the rest of the workforce. This section discusses ways to expand the talent pool by expanding AI-relevant higher education, attracting more women and welcoming foreign workers.
Expanding AI-relevant higher education
The authorities as well as researchers have identified a persistent shortage of higher-education graduates in the subject areas that are critical to creating and developing AI (Israel Innovation Authority, 2019[18]; Artificial Intelligence and Data Science Committee, 2020[19]; Ehrlich and Mekonen, 2024[17]; State Comptroller of Israel, 2024[10]). This scarcity, while it also affects the rest of high-tech, is particularly challenging for AI, which involves particularly strong demand for advanced academic skills (Ehrlich and Mekonen, 2024[17]). Compared with the past decades, training a larger number of graduates in AI-relevant fields has become more important, as the previously large potential of highly STEM-skilled immigrants from former Soviet Union countries has been largely exhausted (Gandal, Roccas Gandal and Kunievsky, 2021[7]).
Efforts undertaken over the past ten years to expand scientific higher education have been fruitful at the bachelor level but less so at the master level. Universities have strongly expanded undergraduate capacity in mathematics, statistics, physics and information and communication technology (ICT), the main fields for AI (Figure 2.5). As a result, many young Israelis earn bachelor degrees in these fields by international comparison ( Panel A). Creating AI systems very often requires more advanced degrees: two-thirds of graduates working in AI production hold a master or PhD, compared to 12% for software developers overall in Israel. However, the share of young Israelis earning masters in the most AI-relevant fields has however stagnated over time at a low level by international standards (Figure 2.6 Panel A and Figure 2.6 Panel B). The result is a shortage of workers with advanced degrees in the AI profession (Ehrlich and Mekonen, 2024[17]). This shortage manifests itself through Israel exhibiting the highest share among OECD countries of data scientists and machine-learning experts with annual salaries above $100k per year, at 62%, against 58% in the United States, where GDP per capita is more than 50% higher.
Graduates in mathematics, statistics, physics and ICT
Sources: OECD Education database; and OECD Population database.
Training more post-graduate students in AI-relevant fields is feasible. Different options exist to expand the number of master-level graduates in AI-relevant subjects starting with an expansion of the current system. The previous Survey called for increasing university places in mathematics, statistics and computer science faster than the general trend increase (OECD, 2023[20]). The public education authorities could allocate a larger budget to the sector to increase capacity in higher-education institutions, especially research universities. A prerequisite for doing so is expanding the pool of scholars who can teach AI-relevant master and PhD courses (State Comptroller of Israel, 2024[10]). Action initiated in this direction under the 2019 National Program for the Advancement of Data Science in Higher Education is continued under the second step, adopted in September 2024, of the National Artificial Intelligence Program. The funding of this initiative should be regularly reviewed
An additional challenge for public-sector universities is to retain AI scholars given strong business demand (Ben-Israel, Matania and Friedman, 2020[21]). One way of reconciling these forces would be to allow AI faculty members to work in industry alongside part-time academic duties, as done for instance in France with associate faculty members or the United States with professors of practice (State Comptroller of Israel, 2024[10]; OECD, 2024[22]). Attracting and keeping scholars in AI-relevant fields also requires sufficient flexibility in university salary schemes to provide competitive levels of pay.
Another avenue is to allow market forces to operate more freely at the post-graduate level. The high wages paid to workers producing AI provide a return on the acquisition of post-graduate diplomas in AI-relevant fields. As a result, there can be economically viable scope for private universities or graduate schools to offer training in these fields. For the labour market to work efficiently, employers need to be in a position to trust the degrees that such new private higher-education providers would award. The Council for Higher Education would have a role to play by standing ready to certify programmes teaching these master and PhDs: the criteria should be academic quality only, since the private origin of the funding takes away budgetary costs from the points to consider.
Notes: Countries are ranked according to the combined share among their age cohort of 2021 graduates in (i) mathematics, statistics, physics and (ii) information and communication technologies. STEM stands for science, technology, engineering and mathematics.
Sources: OECD Education database; and OECD Population database.
A specific challenge in Israel for earning a post-graduate degree is the long compulsory military service (32 months for men, 24 for women). This raises the opportunity cost of master and PhD degrees for those who would prefer to start working by a given age. Blended-learning programmes combining in-person intensive sessions with online coursework while adhering to high academic standards offer a way to reconcile the pursuit of advanced AI-relevant degrees with the start of work and/or family lives. Existing programmes in AI-relevant fields could be expanded at the Open University while facilitating their creation by other universities.
AI qualifications can also be acquired during military duties. Military authorities are enabling a limited number of conscripts to earn degrees, mostly in technological fields, under the Atuda Programme. As part of the National Artificial Intelligence Programme (Box 2.1), the Israeli Defence Forces (IDF) will develop a specialised AI training programme. It will be worth connecting this programme with academic curricula so that its participants can follow it up with further AI studies in higher education institutions after their military duty if they wish to. Nurturing links between the IDF AI programme and academia as well as the private sector will maximise economic spillovers.
Bridging the large gender gap
There is scope to expand the pool of highly skilled labour for AI by attracting and retaining more women. Women are underrepresented in the AI workforce across OECD economies (OECD, 2024[23]). Currently, in Israel, women perform 23% of professional-level AI-production jobs whilst only 13% of AI startup founders are women (Ehrlich and Mekonen, 2024[17]). These shares are close to the ones observed across the broader high-tech sector, where women make up 29% of technological positions and 12% of startup founders. One study found that high-tech firms treat male and female applicants equally (Neyer and Soroker, 2022[24]) and instead attributed the difference to a “funnel” effect with women being less likely than men to move into high-tech-relevant streams at each stage of their education and professional curriculum. Women earning master’s degrees in mathematics, statistics, physics and ICT account for a declining share of their age cohort, declining to below 30% by 2021 from 36% in 2013 (Figure 2.5 Panel B).
A first avenue for action is to create more links between female scientific graduates and AI firms. One step in this direction would be to provide better information within higher-education programmes about AI labour-market prospects. Military duties are a direct way for public authorities to provide more women with training and experience in AI-relevant fields. Defence forces could make greater efforts to recruit more women in software-development and cybersecurity (two areas where they occupied only 23% of positions in 2019) and ensure high female presence in new AI programmes.
Second, the authorities can promote greater take-up of AI-relevant subjects by female students to develop information about career and wage prospects by topic in middle and high-schools before pupils choose their major (Encinas-Martín and Cherian, 2023[25]). It is also helpful for education authorities to ensure that middle and high-school curriculums fully overcome gender biases and stereotypes regarding math and science (Pawelec and Lesher, 2022[26]). A partnership between the Ministry of Labour and the non-profit organisation She Codes in mid-2023 launched a new initiative to encourage more female pupils who completed mathematics credits in high school to opt for academic studies relevant for high tech. The authorities should closely monitor this initiative to expand it if it proves successful. Furthermore, military authorities could increase the share of women (19% in 2019) following science and engineering courses under the Atuda programme.
Raising the employment of under-represented groups in AI production
Large segments of the population currently have comparatively few members working in high-tech, offering considerable potential to provide additional human resources in the future for high tech in general, also benefiting AI. Within the working-age population, while on average 11.6% of Israelis work in high-tech, this share is only 4.3% and 3.5% for Haredi (ultra-Orthodox) women and men, respectively, and even lower at 1.7% and 1.1% for Arab men and women, respectively (Israel Innovation Authority, 2024[5]).
Improving the quality and quantity of primary and secondary education for the under-represented groups is essential. Doing so could equip large numbers of young Israelis with the skills necessary to undertake higher education in scientific fields and later have opportunities to work in high-tech including AI. The economic benefits for the economy would materialise in terms of both greater participation and higher productivity. As documented in the previous Survey (OECD, 2023[20]), mean proficiency levels among 15-year-olds lag behind those of most other OECD countries, with low outcomes among particular population groups being important factors behind this average result. Four in every five Haredi boys (making up about 8% of their age cohort) receive secondary education from yeshivas, which have reduced or no requirements to teach mathematics or English (Finkelstein, 2023[27]). The Arab education system, which receives lower funding than state Hebrew streams especially at the secondary level, produces substantially lower learning outcomes. There is considerable scope to reform primary and secondary education including by increasing funding for the Arab system and requiring the teaching of core subjects in full across all education streams (Table 2.1). Additional barriers include limited public transportation from Arab towns to high-tech employment hubs, and a shortage of childcare facilities for children under three in Arab municipalities.
Bringing in more foreign experts
An additional way of overcoming highly-skilled-labour shortages is to hire non-Israeli workers. One option has been to employ West Bank workers and link with the nascent AI sector around Birzeit University in Ramallah; the Israeli authorities in 2021 issued 500 permits for Palestinians to work in the high-tech sector, but take-up was very limited (below ten persons). In mid-2023, the Israeli authorities opened a tech hub in East Jerusalem. However, the suspension of work permits for West Bank workers following 7 October 2023 closed this avenue for potential labour.
Since 2018, the Israeli authorities have put in place special work visas for high-tech experts. These visas are granted for one year, can be extended to five years, allow spouses to work and benefit from an expedited procedure (6-10 business days). However, the reach of these visas may be limited by the yearly renewals and involvement of the employer, which can create considerable uncertainty for would-be immigrants into a sector that is characterised by strong labour-market churn. This is a small programme in quantitative terms: for the entire Israeli economy, foreign workers holding expert visas represented 0.12% of employment as of mid-2024 (while numbers specific to high-tech are unavailable).
The high-tech expert visa programme could be reformed to become more attractive. A way forward is to tie eligibility criteria to offer permits for a longer period than one year while weakening the link between the visa and a specific employer. Immigration systems putting the employer at the centre have typically been unable to attract very highly qualified migrants in sufficient numbers (Chaloff and Lemaître, 2009[28]).The link to the employer could be weakened by considerably lengthening the permit duration and allowing an extended period of job search if the employment contract is terminated before requiring to depart. However, the absence of a citizenship perspective would limit the attractiveness of the scheme compared with other destination countries.
2.2.5. The regulatory environment is supportive
The authorities are operating a regulatory environment implementing the OECD AI Principles by incorporating them in sector-level regulatory efforts rather than through over-arching cross-sector regulation. In line with the OECD framework for the classification of AI Systems, Israel’s AI Policy follows a risk-based approach. Similarly to a number of other jurisdictions, Israel requires sector-specific regulators to assess AI-related risks and take appropriate measures (OECD, 2023[29]; OECD, 2022[30]; Ministry of Innovation et al., 2023[31]). This decentralised framework allows regulators to implement the principles in ways that can be adapted to sector specificities. A government decision of February 2023 established a centre for AI regulation assisting and coordinating the work of sectoral regulators to ensure that regulatory efforts are consistent (OECD, 2023[29]; Office of Legal Counsel et al., 2023[32]).This approach also favours soft tools such as ethical principles, standards and recommendations for voluntary adoption as well as areas for controlled regulatory experimentations (so-called “sandboxes”).
The sector-by-sector approach has so far served AI development well. It has avoided potential unnecessary costs that might arise from a centralised, cross-sectoral regulation in the hypothesis that it was applied too rigidly (Martens, 2024[33]). Given the rapid advances in AI, the authorities should at regular intervals review the regulatory framework to check that it continues to encourage innovation and competition for trustworthy AI in line with OECD AI Principle 2.3b (OECD, 2019[1]).
Access to data is central to AI development. This area, long identified as a challenge (Israel Innovation Authority, 2019[34]), received priority status in September 2024, as part of the adoption of the second phase of the National AI Programme, which is welcome. The country’s adherence to the OECD Recommendation on Artificial Intelligence contributes to the objective of fostering adoption of trustworthy AI. The legal framework has been strengthened through the adoption of AI Policy Principles in December 2023, the revision in August 2024 of the privacy law and the signature in September 2024 of the Council of Europe framework convention on artificial intelligence and human rights, democracy and the rule of law.
The privacy protection framework is evolving. The review currently underway about the use of AI in financial services (see below) may result in recommendations on measures for privacy and data protection that could provide a framework to address the privacy implications of AI systems in other sectors. In particular, the August 2024 legislative revision put in place privacy-protection obligations backed by the threat of potentially large fines under the watch of the Privacy Protection Authority. The authorities should monitor the effects of this regulation to guarantee a proper balance between on the one hand protecting privacy and on the other hand facilitating data access and sharing while providing legal certainty to businesses. Establishing tight co-operation between the national centre for AI regulation and the Privacy Protection Authority would be a way of fostering coherence between the two regulatory activities.
The National AI Programme gathers the institutions and ministries responsible for AI policy: the Israel Innovation Authority, the Ministry of Finance, the Ministry of Innovation, Science & Technology, the Council for Higher Education and the Directorate of Defence Research and Development. Launched in February 2023 through Government Resolution No. 173, the program “aims to secure Israel’s position as a global leader in AI innovation and technology.” It rests on three pillars: a government strategy, investments in AI infrastructure and the promotion of AI adoption across the economy.
After the adoption of its second phase in September 2024, the Programme aims to:
Integrate AI in public service delivery
Foster research by
establishing a National AI Research Institute and
launching a “moonshot” challenge to fund AI projects aiming at scientific breakthroughs and their application in the industry and defence sectors.
Build human capacity by
Expanding academic research and advanced degree programmes
Developing a specialised AI training program in the Israel Defence Forces (IDF)
Facilitating access to data, especially government datasets
Encouraging transformative AI ventures, including by facilitating experimentation in heavily regulated fields by temporarily removing barriers while monitoring results (“sandboxes”).
AI represents a major change in the digital revolution, transforming many tasks and reshaping large planks of economies and societies. Firms and workers report across a wide range of surveys conducted in many countries that the use of AI has considerably enhanced their productivity (Filippucci et al., 2024[35]). The promise of AI goes beyond productivity with users reporting that the technology has improved their job enjoyment as well as their mental and physical health (Lane, Williams and Broecke, 2023[36]).
The deployment of AI across sectors is particularly challenging in a dual economy such as Israel’s. While skills, productivity and pay are elevated in the high-tech sector, two thirds of workers are employed in sectors where productivity is below the OECD average (Koelle, 2023[37]). Digitisation, including the spread of AI, has been lagging in conventional industrial sectors such as manufacturing and utilities (Figure 2.7).
Shares of AI users in ICT as well as manufacturing and utilities, 2020 or latest available year
Note: The survey was conducted in 2020 in most of the countries (2018 in France and 2019 in Germany, Japan and Korea).
Source: Calvino, F. and L. Fontanelli (2023[38]), “A portrait of AI adopters across countries: Firm characteristics, assets’ complementarities and productivity,” OECD Science, Technology and Industry Working Papers, No. 2023/02.
Policy settings matter, as the impact of AI is far from given. The aggregate-output effects of AI are highly uncertain: anticipations range from a cumulative output level increase over ten years of 1% (Acemoglu, 2024[39]) to a rise of 18% over the same period followed by a permanent boost to the growth rate of one percentage point (Baily, Brynjolfsson and Korinek, 2023[40]). A recent OECD study points to permanent effects on output growth in the order of 0.25 to 0.6 percentage points (Filippucci, Gal and Schief, 2024[3]). Besides reflecting uncertainty, the higher numbers across these studies hinge on assumptions that the technology spreads widely and fast while also boosting the innovation rate. In other words, realising the potential of AI across the economy depends on having framework conditions and policies that facilitate its adoption across the economy (Filippucci et al., 2024[35]).
AI is likely to matter for a large number of workers though with very different implications depending on whether AI tools complement or replace the tasks that workers execute. Estimates using the sector and occupation structure of the Israeli economy suggest that as many as 30% of workers are highly exposed to AI with AI complementarity while 23% are highly exposed with high substitutability (Figure 2.8 Panel A).
The role of public policies varies between situations of complementarity and substitutability.
Where AI complements workers’ activity, the effect is higher productivity, which opens the potential for higher wages and profits: incentives are in this case aligned between workers and employers. As identified in previous Economic Surveys, Israel is lagging behind in the digitisation of its non-high-tech sector with considerable scope for improving framework conditions for the take-up of digital technologies, among which AI, through quality education, lifelong learning and competitive product markets. Management and business skills have been identified as most in demand among occupations that are highly exposed to AI, a finding that provides an orientation for public efforts to support lifelong learning programmes (Green, 2024[41]; Borgonovi et al., 2023[42]). Training can also help build support for AI use, since AI users who have received training are more likely to expect AI to lead to higher wages (Lane, Williams and Broecke, 2023[36]).
Where AI can substitute for workers, the onus for policy is on easing reallocation, to reap the economy-wide gains from AI, while creating new opportunities for displaced workers. Maintaining, and where possible strengthening, the policies that have led to a vibrant labour market with low unemployment is essential to facilitate reallocation. It is also important to ensure that workers, especially those that are potentially or actually displaced, can benefit from training opportunities and other active labour market programmes.
Impacts from AI deployment is set to differ across sectors. AI is likely to be most important for education, finance, ICT and real estate (Figure 2.8 Panel B). This heterogeneity has implications for public policy regarding specific sectors. For education, since the sector is largely publicly run, this calls for the authorities to embrace the roll-out of AI. For finance, the potential of AI will be more likely to be realised if authorities allow competitive pressures to operate in particular by facilitating the entry of fintech companies into retail markets. The regulatory framework for the use of AI in finance is currently under review to encourage innovation while preserving privacy, financial stability, accountability and absence of discrimination as well as mitigating cybersecurity, third-party and misinformation risks (Interagency Task Force, 2024[43]). The review focuses on critical areas in the financial sector that are likely to be most affected by AI such as investment advice, portfolio management, consumer credit and insurance underwriting. This review exemplifies the current regulatory approach in Israel: sector-by-sector regulation with the objective of fulfilling the potential of AI while protecting privacy and other public-policy objectives (including financial stability in this case). Besides, the large substitutive potential of AI for labour in finance and insurance also means that accompanying policies, including through retraining opportunities, are essential.
The level of education also matters for the worker-level effects of AI. The occupational structure of the Israeli economy implies that high-skilled workers are overall more exposed to AI than middle or low-skilled ones with an impact that for the vast majority will be complementary (Debowy et al., 2024[44]). This contrasts with previous waves of technological innovation, such as automation, the impacts of which concentrated more on middle-skilled workers in manufacturing and typically more substitutive. This means that, for most high-skilled workers, AI is likely to bring greater productivity and potentially higher wages. As mentioned above, however, there may be exceptions in sectors such as finance and insurance, where substitution may dominate.
The policy settings that matter for AI adoption while enhancing social inclusion very largely overlap with the ones that determine successful digitalisation. This observation re-emphasises the relevance of the recommendations made in the previous Survey to broaden the use of digital technologies across the economy (Table 2.2).
A central requirement for widely beneficial AI use is to bridge digital, mathematics and literacy divides. Internet use varies more between low and high-education groups than in most other OECD countries (OECD, 2023[20]). This reflects to some extent choices by some groups to refrain from internet access at home or smartphones for religious reasons but also geographic disparities in the availability of high-quality digital infrastructure. Israel has a comparatively low stock of public ICT capital given its GDP (Axelrad, Sumkin and Haver, 2022[45]). Fibre-optic connections however are being deployed apace across the country, narrowing spatial gaps in access to very high-speed internet. Remarkably, the share of fibre connections in total broadband has risen from 5.5% in 2019 to 48% in December 2023, which places Israel above the OECD average (43%) though still well below top performers (with above 85% rates in Iceland, Korea and Spain).1
Estimated level and nature of AI exposure across the economy, 2018-2023
Note: See the source for the methodology.
Source: Debowy et al. (2024[44]), Artificial Intelligence and the Israeli Labour Market.
Average digital skills among adults are low by international comparison, including for young people. Only 35% of the surveyed 25-to-35-year olds proving able to solve problems in technology-rich environments against an OECD average of 45%; the average achievement gaps are much larger among the ultra-Orthodox and Arab groups (OECD, 2023[20]). As discussed in the previous Economic Survey, digital upskilling requires improvements to early, primary and secondary education, strengthening links across the different education streams, enhancing teacher quality and strengthening work-based vocational education as well as facilitating lifelong learning in this area. Efforts that are currently underway to incorporate initiation to AI in secondary education are welcome, as they can encourage pupils to choose relevant fields later or take up AI more easily in their professional life.
Israel distinguishes itself by having workers that are among the likeliest in the OECD area to report possessing AI skills, even after controlling for the large share of the high-tech sector in Israel (Figure 2.9, Panel A). This self-evaluation contrasts with the relatively low share of adults identified as possessing digital skills in the 2015 PIAAC survey. While this apparent contradiction might partly stem from limitations intrinsic to international surveys, it reflects a positive attitude of Israeli workers towards taking up AI across the economy.
Notes: Panel A chart shows the prevalence of workers with AI skills – as self-reported by LinkedIn members from 2015-2022 – by country and against a benchmark set at the OECD average. A country’s AI skills penetration of 1.5 means that workers in that country are 1.5 times more likely to report AI skills than workers in the benchmark. Average from 2015 to 2022 for a selection of countries with 100 000 LinkedIn members or more.
In Panel B, Adaptive Problem Solving (APS) refers to the ability of adults to adapt to new circumstances and learn throughout life. This new domain in the 2023 Survey of Adult Skills replaces the assessment of problem solving in technology-rich environments in the previous cycle of the Survey.
Sources: OECD AI Policy Observatory (OECD.AI); and OECD (2024), Do Adults Have the Skills They Need to Thrive in a Changing World?: Survey of Adult Skills 2023, OECD Skills Studies, OECD Publishing, Paris, https://doi.org/10.1787/b263dc5d-en.
This widespread readiness for AI reported by workers seems unmatched by take-up among firms (Figure 2.10). In 2020, when they were last surveyed on this subject, comparatively few Israeli firms indicated having deployed big data analysis, a required underpinning for many strands of AI (OECD, 2024[11]). A first step for the authorities to prepare to promote broader deployment would be to repeat the survey on AI adoption. Given how fast adoption has spread in other countries since 2019, newer indicators are required to ascertain the extent to which the issue remains.
Businesses of more than ten employees using artificial intelligence (AI)
Source: OECD ICT Access and Usage by Businesses database, https://oe.cd/dx/ict-access-usage.
Public authorities can foster AI use among firms through a number of avenues, starting with legal security and stability. As for AI creation (see above), a policy framework that provides legal certainty over data access and AI deployment will facilitate its use across the economy, especially among SMEs that lack resources to evaluate legal risks.
The public sector can facilitate the take-up of AI tools by easing access to its data. Israel’s national AI program includes a goal to facilitate the use of public sector datasets in line with OECD recommendations (OECD, 2008[46]; OECD, 2016[47]; OECD, 2021[48]). Government Decision 1933, adopted in 2016, includes an “Open by Default” policy, which materialised through the data.gov.il platform under the responsibility of the Israel National Digital Agency (then named the National ICT authority). Furthermore, the Ministry of Health has been managing a platform providing access to health data for researchers, with privacy protection and security safeguards, including employing privacy-enhancing technologies.
Programmes are currently underway to encourage the use of government data. The Ministry of Innovation, Science and Technology in collaboration with the National Digital Agency has in recent years launched calls for proposals for data-based applied AI research in collaboration between academia and public bodies. The government allocates funds to the researchers and coordinates the access to public sector data sets. Nevertheless, the use of big data, which is key for many AI deployment cases, remains limited among AI companies: there is scope to further enhance data availability, with appropriate safeguards such as privacy-enhancing technologies, as a way of fostering broader AI use (State Comptroller of Israel, 2024[10]).
Furthermore, it should be noted that in the context of LLMs and adoption of AI tools by the local market, the Hebrew language can be a challenge. AI developers in other countries have little incentive to develop models for a small market. A benefit of the government promoting development based on public sector information is to encourage the development of AI tools based on Hebrew language datasets. This can facilitate the emergence of Hebrew-fluent tools that local firms and SMEs can more readily adopt.
Reinforcing education at all stages will ease AI deployment, with a particular role for lifelong learning to facilitate AI take-up by the existing workforce. As pointed out in the previous Economic Survey, however, workplace training is generally low in Israel, and a fragmented accreditation system complicates the recognition of qualifications across the economy. Expediting plans to establish a National Qualifications Framework would help to define clear learning outcomes, providing greater transparency about the skills acquired (OECD, 2023[20]; OECD, 2024[49]). Especially if integrated with high-tech skill initiatives by the Innovation Authority and the Ministries of Labour and the Economy, such a framework would make it easier for firms to hire staff with skills useful for rolling out AI while sharpening incentives for workers to acquire them.
Key skills for AI deployment go well beyond the scientific and ICT competencies that are central to the creation of new AI products. A survey conducted by the OECD across ten countries identified that management and business skills are the most demanded ones in occupations with high AI exposure (Green, 2024[41]). Social and digital skills are also found to be in strong demand in high-AI-exposure occupations. These results underline that, alongside a knowledge of digital technologies, managerial competence, including the social skills required to promote successful business transformations, are essential to the take-up of AI.
A possible option to accelerate the take-up of AI, especially by small and medium-sized enterprises (SMEs), is to implement support programmes targeted at them. When considering AI adoption, many SMEs face barriers regarding technical, managerial and legal skills (OECD, 2021[50]). Over one in four SMEs point to bottlenecks including training (OECD, 2024[51]). Access to finance does not appear to be a major barrier owing to the widespread availability of AI technologies in the form of software-as-a-service with data hosted through cloud computing, which allows costs to be scalable by comparison with in-house solutions (OECD, 2021[50]). Against this background, government support appears to be particularly valuable if geared at facilitating access to data and training in digital and managerial skills (OECD, 2024[52]; Kergroach, 2021[53]). It is preferable to target businesses by criteria other than size, such as the age of the firm, to avoid creating disincentives to scale up (OECD, 2020[54]).
Government institutions can also take advantage of AI to improve the quality and efficiency of their services. The national AI programme (Box 2.2) promotes the integration of AI technologies across the public sector to improve public services by making them more easily accessible and personalised, enhance decision-making processes through greater use of data, and increase public-sector efficiency. Besides reducing costs, the automation of repetitive cognitive tasks can allow public-sector employees to focus on activities that require human judgement or creativity. Many initiatives are already being implemented with most at early stages of development but one already providing highly promising results in the fight against VAT fraud (Box 2.3).
Scaling up AI initiatives to modernise government however raises the challenges of developing digital skills in the public sector and attracting skilled human capital with high-tech skills. The National Digital Agency is operating a digital training school that enables public-sector employees to train in big data management and AI deployment. Specific AI training programmes are also available for local-government officials. As regards hires, the OECD (2021[55]) review of public sector pay and previous Economic Survey highlighted that the rigidity of the government wage structure makes it difficult to attract highly skilled professionals in areas of need: in particular, pay rises in one job classification typically trigger pay rises in other job classifications. Public authorities should use the flexibility clause introduced by the May 2023 wage agreement to offer higher wages when seeking professionals with the skills required for AI deployment.
The “Israel Invoices” programme
An operational AI project is providing strong results in the tax area. On May 5, 2024, the Israel Tax Authority started the operational implementation of the "Israel Invoices" programme, aimed at reducing the issue of fictitious invoices. Fictitious invoices allow criminals to receive VAT refunds for sales of goods or services that never occurred. These fictitious invoices also reduce the amount of income tax businesses need to pay at the end of the year, as the associated apparent spending on intermediate products reduces their profits. It is estimated that the system could yield NIS 2.5bn (0.1% of GDP) in VAT revenue that is currently lost to fictitious-invoice fraud (State-Comptroller, 2024[56]) .
Under the programme, tax authorities’ issues assignment numbers for tax invoices through an online system. These allocated numbers are required as a condition for deducting input tax in transactions exceeding the ceiling established by law (NIS 25,000, around USD 7,000, for 2024). With the new system, every real-time transaction goes through the credit clearing system to the tax authority. An artificial intelligence system applies ten (unpublished) criteria (“suspicious indicators”) to determine the genuineness of each transaction, on the basis of which it then assigns a grade. If a transaction raises a certain number of red flags, it receives no automatic number.
Every day, between 50,000 and 100,000 invoices exceeding the minimum amount are submitted for approval. As of mid-September 2024, the total value of fictitious invoices detected in real-time was NIS 9 billion, which has avoided NIS 1.5 billion in VAT loss due to fraud. In September, a first arrest was made following fraud detected by the system, amounting to potentially 135 million shekels (USD 37 million). Following the success of the program, the ministry of finance approved an increase in manpower at the Tax Authority by about 100 employees, including some to fill technological positions.
New government AI projects launched following a national competition
In 2023, the ministry of innovation launched a competitive mechanism for national projects to incorporate AI in the public sector. The ministry of innovation offers budgetary and professional support alongside the National Digital Agency, which, attached to the ministry of the economy, promotes and pilots AI use by government departments. Priority was given to cloud-based projects relying on the government’s central cloud platform (Nimbus, run by the National Digital Agency). Among the nine selected projects, initiated in March 2024, a remarkable one is an initiative by customs authorities to automatically classify goods based on the import invoice. This project is expected speed up goods arrival as well as reduce storage fees, taxes and administrative costs: as developed in Chapter 4, advances in trade facilitation can reduce the cost of living. Another promising project, by the ministry of transport, intends to use AI to improve the identification of priority areas for traffic-enhancing investment. The ministry of justice plans to develop an AI model to help detect non-profit organisations that might launder money or finance terrorism.
Source: Government Resolution 173 of February 24, 2023, (Ministry of Innovation, Science & Technology, 2024[57]), (Israel Tax Authority, 2024[58])
References
[39] Acemoglu, D. (2024), “The Simple Macroeconomics of AI”, Economic Policy forthcoming.
[19] Artificial Intelligence and Data Science Committee (2020), National Strategy for Artificial Intelligence and Data Science: Findings and Recommendations, https://tinyurl.com/bdz2rxex.
[45] Axelrad, H., S. Sumkin and S. Haver (2022), Promoting and Developing Digital Transformation in Israel Toward 2030.
[7] Ben-Bassat, A., R. Gronau and A. Zussman (eds.) (2021), The High-Tech Sector, Cambridge University Press.
[21] Ben-Israel, I., E. Matania and L. Friedman (2020), The National Initiative for Secured Intelligent Systems to Empower the National Security and Techno-Scientific Resilience: A National Strategy for Israel - Special Report to the Prime Minister.
[42] Borgonovi, F. et al. (2023), “Emerging trends in AI skill demand across 14 OECD countries”, OECD Artificial Intelligence Papers, No. 2, OECD Publishing, Paris, https://doi.org/10.1787/7c691b9a-en.
[40] Brookings (ed.) (2023), Machines of mind: The case for an AI-powered productivity boom, https://www.brookings.edu/articles/machines-of-mind-the-case-for-an-ai-powered-productivity-boo.
[38] Calvino, F. and L. Fontanelli (2023), “A portrait of AI adopters across countries: Firm characteristics, assets’ complementarities and productivity”, OECD Science, Technology and Industry Working Papers, No. 2023/02, OECD Publishing, Paris, https://doi.org/10.1787/0fb79bb9-en.
[28] Chaloff, J. and G. Lemaître (2009), “Managing Highly-Skilled Labour Migration: A Comparative Analysis of Migration Policies and Challenges in OECD Countries”, OECD Social, Employment and Migration Working Papers, No. 79, OECD Publishing, Paris, https://doi.org/10.1787/225505346577.
[44] Debowy, M. et al. (2024), Artificial Intelligence and the Israeli Labour Market, Taub Center for Social Policy Studies in Israel, Policy Paper No. 06.2024.
[17] Ehrlich, E. and T. Mekonen (2024), Israel’s Artificial Intelligence Landscape, https://www.storydoc.com/5d662c0f663f17d29047f62992d9038e/d65f927c-1a16-46de-8621-a71ff45800ec/663fde437845e765ff62f66b.
[25] Encinas-Martín, M. and M. Cherian (2023), Gender, Education and Skills: The Persistence of Gender Gaps in Education and Skills, OECD Skills Studies, OECD Publishing, Paris, https://doi.org/10.1787/34680dd5-en.
[13] Epoch AI (2024), Tracking Large-Scale AI Models, https://epochai.org/blog/tracking-large-scale-ai-models (accessed on 25 July 2024).
[35] Filippucci, F. et al. (2024), “The impact of Artificial Intelligence on productivity, distribution and growth: Key mechanisms, initial evidence and policy challenges”, OECD Artificial Intelligence Papers, No. 15, OECD Publishing, Paris, https://doi.org/10.1787/8d900037-en.
[3] Filippucci, F., P. Gal and M. Schief (2024), Miracle or Myth? Assessing the macroeconomic productivity gains from Artificial Intelligence, https://doi.org/10.1787/b524a072-en.
[27] Finkelstein, A. (2023), A Very Foreign Language.
[41] Green, A. (2024), “Artificial intelligence and the changing demand for skills in the labour market”, OECD Artificial Intelligence Papers, No. 14, OECD Publishing, Paris, https://doi.org/10.1787/88684e36-en.
[15] Groes, K. and A. Ludvigsen (2023), The carbon footprint of GPT-4, https://towardsdatascience.com/the-carbon-footprint-of-gpt-4-d6c676eb21ae.
[16] IEA (2024), Electricity 2024: Analysis and forecast to 2026.
[43] Interagency Task Force (2024), Interagency Task Force Publishes Interim Report on AI Use in the Financial Sector for Public Comments, https://www.new.isa.gov.il/en/nav-index/supervised-publications/new051124.
[6] Israel Innovation Authority (2024), Study on Gen-AI Companies in Israel.
[5] Israel Innovation Authority (2024), The State of High-Tech: 2024 Annual Report, Israel Innovation Authority.
[34] Israel Innovation Authority (2019), Bolstering Artificial Intelligence, https://innovationisrael.org.il/en/report/bolstering-artificial-intelligence/.
[18] Israel Innovation Authority (2019), Innovation Report, https://innovationisrael.org.il/en/report/bolstering-artificial-intelligence/.
[58] Israel Tax Authority (2024), Israel Invoices.
[53] Kergroach, S. (2021), “SMEs Going Digital: Policy challenges and recommendations”, OECD Going Digital Toolkit Notes, No. 15, OECD Publishing, Paris, https://doi.org/10.1787/c91088a4-en.
[37] Koelle, M. (2023), “Addressing labour market challenges for sustainable and inclusive growth”, in OECD Economic Surveys: Israel 2023, OECD Publishing, Paris, https://doi.org/10.1787/727fa4b8-en.
[36] Lane, M., M. Williams and S. Broecke (2023), “The impact of AI on the workplace: Main findings from the OECD AI surveys of employers and workers”, OECD Social, Employment and Migration Working Papers, No. 288, OECD Publishing, Paris, https://doi.org/10.1787/ea0a0fe1-en.
[33] Martens, B. (2024), The European Union AI Act: premature or precocious regulation?, https://www.bruegel.org/analysis/european-union-ai-act-premature-or-precocious-regulation.
[31] Ministry of Innovation et al. (2023), Israel’s Policy on Artifical Intelligence: Regulation and Ethics, https://www.gov.il/BlobFolder/policy/ai_2023/en/Israels%20AI%20Policy%202023.pdf.
[57] Ministry of Innovation, Science & Technology (2024), The winners of the artificial intelligence implementation project in the government offices are revealed, https://www.gov.il/he/pages/most_ai_government_agencies_open_call_winners.
[9] Ministry of Innovation, S. et al. (2024), AI National Program, https://aiisrael.org.il/.
[24] Neyer, D. and I. Soroker (2022), Ladies, you are invited to Hi-Tech!, https://hethcenter.colman.ac.il/2022/11/21.
[23] OECD (2024), Algorithm and Eve: How AI will impact women at work, OECD Publishing, Paris, https://doi.org/10.1787/a1603510-en.
[2] OECD (2024), “Explanatory memorandum on the updated OECD definition of an AI system”, OECD Artificial Intelligence Papers, No. 8, OECD Publishing, Paris, https://doi.org/10.1787/623da898-en.
[14] OECD (2024), “Financing broadband networks of the future”, OECD Digital Economy Papers, No. 365, OECD Publishing, Paris, https://doi.org/10.1787/eafc728b-en.
[11] OECD (2024), OECD Digital Economy Outlook 2024 (Volume 1): Embracing the Technology Frontier, OECD Publishing, Paris, https://doi.org/10.1787/a1689dc5-en.
[52] OECD (2024), OECD Digital Economy Outlook 2024 (Volume 2): Strengthening Connectivity, Innovation and Trust, OECD Publishing, Paris, https://doi.org/10.1787/3adf705b-en.
[49] OECD (2024), Skills First for Inclusive and Efficient Labour Markets.
[51] OECD (2024), “SME Digitalisation to manage shocks and transitions: 2024 OECD D4SME survey”, OECD SME and Entrepreneurship Papers, No. 62, OECD Publishing, Paris, https://doi.org/10.1787/eb4ec9ac-en.
[22] OECD (2024), The state of academic careers in OECD countries – an evidence review.
[12] OECD (2023), “A blueprint for building national compute capacity for artificial intelligence”, OECD Digital Economy Papers, No. 350, OECD Publishing, Paris, https://doi.org/10.1787/876367e3-en.
[20] OECD (2023), OECD Economic Surveys: Israel 2023, OECD Publishing, Paris, https://doi.org/10.1787/901365a6-en.
[29] OECD (2023), “The state of implementation of the OECD AI Principles four years on”, OECD Artificial Intelligence Papers, No. 3, OECD Publishing, Paris, https://doi.org/10.1787/835641c9-en.
[30] OECD (2022), “OECD Framework for the Classification of AI systems”, OECD Digital Economy Papers, No. 323, OECD Publishing, Paris, https://doi.org/10.1787/cb6d9eca-en.
[48] OECD (2021), Recommendation of the Council on Enhancing Access to and Sharing of Data, https://legalinstruments.oecd.org/en/instruments/OECD-LEGAL-0463.
[50] OECD (2021), The Digital Transformation of SMEs, OECD Studies on SMEs and Entrepreneurship, OECD Publishing, Paris, https://doi.org/10.1787/bdb9256a-en.
[55] OECD (2021), The Public Sector Pay System in Israel, OECD Publishing, Paris, https://doi.org/10.1787/3b6ad37f-en.
[54] OECD (2020), “Going Digital integrated policy framework”, OECD Digital Economy Papers, No. 292, OECD Publishing, Paris, https://doi.org/10.1787/dc930adc-en.
[1] OECD (2019), Recommendation of the Council on Artificial Intelligence, https://legalinstruments.oecd.org/en/instruments/oecd-legal-0449.
[47] OECD (2016), Recommendation of the Council on Health Data Governance, https://legalinstruments.oecd.org/en/instruments/OECD-LEGAL-0433.
[46] OECD (2008), Recommendation of the Council for Enhanced Access and More Effective Use of Public Sector Information.
[32] Office of Legal Counsel et al. (2023), Principles, policy, regulation and ethics in the field of artificial intelligence, https://www.gov.il/he/pages/ai_23.
[26] Pawelec, H. and M. Lesher (2022), Why don’t more women code?, https://oecdstatistics.blog/2023/03/08/why-dont-more-women-code/.
[8] RISE (2024), Communication to the OECD, RISE Israel.
[10] State Comptroller of Israel (2024), National Preparation for Artificial Intelligence, https://www.mevaker.gov.il/sites/DigitalLibrary/Pages/Reports/7944-7.aspx.
[56] State-Comptroller (2024), The Tax Authority’s handling of the phenomenon of fictitious invoices.
[4] Tortoise (2024), The Global AI Index, Tortoise, https://www.tortoisemedia.com/intelligence/global-ai/.
[59] Unsal, F. et al. (2024 forthcoming), “The Impact of AI on Aggregate Productivity Growth”, OECD Economics Department Working Papers.
← 1. Source: OECD Broadband Statistics database.