Latest / 1 More Paper / A Silent Partner: The Shadow Presence of Generative Artificial Intelligence in Public Administrations — Tangi et al. (2026)

Transcript

A Silent Partner: The Shadow Presence of Generative Artificial Intelligence in Public Administrations — Tangi et al. (2026) · 1 More Paper

• You’re listening to “A Silent Partner: The Shadow Presence of Generative Artificial Intelligence in Public Administrations,” by L. Tangi, A.P. Rodriguez Müller, and M. Combetto. Published in 2026. Luca Tangi1envelope symbol, A. Paula Rodriguez Müller2, and Marco Combetto3 1 European Commission, Joint Research Centre, Ispra, Italy the email address 2 European Commission, Joint Research Centre, Seville, Spain 3 Fincons SpA, Vimercate, Italy Abstract. The rise of generative artificial intelligence (GenAI) is bringing in a new era of technological advancement in the public sector, with sophisticated GenAI tools now readily accessible to all public servants for a wide array of appli-cations. Public servants use GenAI tools to draft reports, refine policy proposals and generate content, often without explicit approval or institutional policies gov-erning their use. This informal and independent adoption of AI, which is referred to as the ‘shadow use’ of GenAI, represents a significant but largely unregulated shift in how AI is integrated into public-sector work. While the benefits are sub-stantial, the challenges are equally significant. The literature on this topic is still in its early stages, emphasising the need to clearly define the research area, outline the main characteristics of the phenomenon and establish terminology relevant to the current debate, particularly concerning the unique features that GenAI intro-duces to the public sector. This study addresses this gap by surveying 576 public managers from seven EU Member States, aiming to shed light on the current state of practice and stimulate discussion about the ongoing transformations. It offers insights into the shifting dynamics and emerging perspectives that GenAI is bringing to the public sector. Generative artificial intelligence (GenAI) is now readily and freely accessible online, enabling anyone with internet access to interact with tools that utilise it. Public servants are part of this wave of enthusiasm and opportunity, increasingly experimenting with GenAI for tasks such as text enhancement, image creation and information retrieval. These tools offer new forms of interaction, often perceived as collaborative or conversational, that are different from those offered by earlier automation technologies. Platforms such as ChatGPT introduce novel user experiences that can often evoke a sense of incredulity and amazement similar to the feeling of encountering a new form of intelligence. While GenAI continues to develop in line with broader trends in AI and – pre-viously – digital government, it also presents new dynamics, particularly in terms of informal, individual-driven use. In some cases, public administrations are piloting or adopting GenAI-based applications, such as Bürokratt in Estonia and UrbanistAI in Helsinki, Finland. However, many public servants are also using GenAI tools inde-pendently, without formal institutional frameworks, oversight or guidance. This kind of autonomous and often undocumented use–which is referred to as the ‘shadow use’ of GenAI – is already widespread and raises important questions for public adminis-tration. In this paper, we focus on this emerging phenomenon. Specifically, we explore how public managers across seven EU Member States engage with GenAI in practice, what motivates or limits its use and what this means for governance, capacity building and future research. The literature on GenAI in public administration is still nascent, with rather frag-mented and scattered research. Much of the current debate is primarily risk oriented, focusing on potential misuse, ethical concerns and the need for mitigation measures, often articulated through common-sense guidelines or high-level governance principles. While these concerns are valid, the discussion tends to prioritise formal adoption and top-down control mechanisms. This overlooks how GenAI is being adopted and implemented into public-administration workflows. As Janssen notes, the increasing ease of AI access and deployment does not automatically translate into higher-quality outcomes or alignment with public values. The more accessible these tools become, the greater the need for governance frameworks that ensure their responsible use while addressing unintended consequences. Rather than assuming that accessibility alone drives impact, there is a need to empirically investigate how GenAI tools are being used in real-world administrative contexts, by whom and for what kinds of tasks. This paper aims to contribute to that effort by offering insights based on an original dataset on the use of GenAI compiled through a survey of 576 public managers across seven Member States. Through an exploratory approach, it examines the current land-scape of GenAI use within the public sector. The findings offer practical evidence to inform future governance discussions and guide future research in this area. 2 Generative Artificial Intelligence and Its Use in the Public Sector The adoption of GenAI in public administrations has accelerated rapidly. The broad online availability of tools such as ChatGPT has enabled public servants to experi-ment with GenAI directly, without relying on institutional procurement or integration processes. While comprehensive data remain limited, initial studies suggest growing engagement. For instance, Apolitical reported that, by September 2023, more than 50% of the public servants surveyed had used GenAI tools. As regards the Canadian federal public service, one study found that 11.2% of public servants had applied generative systems for work-related purposes. What distinguishes GenAI from previous waves of digital innovation is not only its accessibility, but also the degree of autonomy it affords to users. Traditional technology implementation in government typically involves managed and controlled processes, including deployment plans and integration into existing systems as part of a digital transformation strategy. However, GenAI can be adopted informally, without top-down approval, training or oversight. This contrasts with many technologies, especially other AI solutions, that remain confined to pilot phases due to complexity, cost or risk aversion. GenAI bypasses many of these barriers, as it is founded on individualised and inde-pendent integration into workloads, which can vary even among individuals with iden-tical responsibilities. Therefore, the potentially large breadth of public-servant users and the simplicity of deployment make the use of GenAI by administrations distinctly different from the use of earlier technologies. On the other hand, this new approach to technology introduces challenges in tracking and managing its use by public servants. Since individuals can easily access these tools online and incorporate them into daily tasks, such as drafting emails, revising text or answering work-related questions, their use is neither embedded in nor traceable through formal organisational processes. This phenomenon can be described as ‘shadow use’. The risk with this type of use is that it can easily go unnoticed and may not comply with professional guidelines at work. Moreover, even individuals with identical roles may adopt GenAI in different ways, resulting in a fragmented usage landscape that is difficult to track and monitor effectively. While the potential benefits of GenAI technology are often emphasised, scholars and practitioners have also begun to explore its possible drawbacks. Although a comprehen-sive list of risks has yet to be established, several recurring concerns are emerging in the literature. Notably, the Council of the European Union published a report in April 2023 outlining the key risks associated with the use of GenAI in public admin-istration. These include challenges linked to accountability, equality and impartiality, efficiency, data security, reliability, citizen involvement and trust, serving the public interest, and quality. Scholars also stress data leakage as a persistent concern, because most of these tools are managed by external providers that often give unclear information about how user data are stored, processed or reused. Public administrations must navigate the tension between the commercial opacity of GenAI providers, the need to foster innovation and the impossibility of fully controlling how these tools are used across organisations. Other frequently cited issues include hallucinations and bias, especially the risk of over-reliance on machine-generated content or uncritical acceptance of discriminatory behaviours. The core concern is not merely the errors themselves, but also the confidence with which they are presented. These errors challenge the mechanistic view of machines, as GenAI’s probabilistic nature does not produce systematic errors. Consequently, GenAI’s errors, akin to human mistakes, bring it closer to human-like behaviour. If not properly understood and managed, this characteristic could lead to inappropriate and potentially harmful technology use. These drawbacks are amplified by the fact that GenAI tools are often not used to meet operational needs. For many public servants involved in policy formulation, communication or advisory roles, interacting with GenAI is less about task automation and more about cognitive support. In this context, GenAI becomes embedded in knowledge work, supporting idea generation, drafting and reflection. This situation represents a new and still evolving challenge for public administra-tions, governments and individual public servants. The widespread yet informal nature of GenAI use creates a context marked by experimentation, learning by doing and varying levels of institutional oversight. However, this does not preclude public administrations from implementing measures to guide and shape the fair use of GenAI. Training sessions and guidelines are already being established to facilitate and regulate the use of GenAI, aiming to enhance efficiency and effectiveness while also addressing potential risks. In this complex and evolving landscape, a clear narrative has yet to take shape. The existing literature remains fragmented, drawing attention to various technical, ethical and organisational aspects, but offering limited insights into how these elements interact in practice. Fundamental questions remain open: ‘What is currently happening?’, ‘What are the implications?’ and ‘How do we tackle them?’. While such uncertainty is expected, given the novelty of the technology, there is a clear need to begin systematically framing the discourse, understanding the phenomenon and identifying the key aspects that require further exploration. In summary, the implications of GenAI for public administrations are extensive and far-reaching, and research in this area remains limited. In particular, the independent – or shadow – use of GenAI in government settings opens up avenues for deeper empirical and conceptual exploration. 3 Methodology 3.1 Sample and Data Collection We conducted a cross-national survey to explore public managers’ use of GenAI tools. The survey was part of a larger survey on AI adoption. The survey specifically targeted individuals holding management roles within local or national governments, with an additional screening question to confirm their employment in the public sector. To ensure a broad perspective, we purposely selected seven Member States, namely Germany, Spain, France, the Netherlands, Austria, Poland and Sweden. This selection includes both larger and smaller Member States, covering diverse geographical regions within the EU and capturing a range of administrative traditions, digital governance maturity levels and AI adoption trends within the EU. Data collection took place in March 2024, with the support of an external company (for details, please refer to ). The survey was conducted online. Responses were screened for completeness to ensure data quality. Given the specific focus on public managers, the survey yielded 576 complete responses, including 53 from Austria, 110 from France, 112 from Germany, 65 from the Netherlands, 64 from Poland, 112 from Spain and 60 from Sweden. The dataset has been published as open data in the Joint Research Centre data catalogue. Our focus on the EU allowed an analysis to be undertaken within the context of a shared regulatory and strategic framework. Since 2021, the EU has prioritised the integration of digital technologies, including AI, into the public sector. The recently adopted AI Actaims to establish a common foundation for AI governance across Member States, emphasising transparency, explainability and fairness while addressing potential risks. By concentrating on the EU, we aimed to explore the factors influencing AI adoption within a structured policy environment, enabling a more nuanced understanding of adoption dynamics. 3.2 Survey Design and Analysis The survey was initially developed in English as a master version. To ensure linguistic accuracy and accessibility, it was professionally translated into the local languages of the selected countries – Dutch, French, German, Polish, Spanish and Swedish – by the Directorate-General for Translation. In addition, researchers with native proficiency reviewed the translations to refine wording and address any ambiguities. The questionnaire covered key aspects of AI adoption in the public sector, including awareness, perceived benefits and risks, barriers to adoption, and actual use of GenAI tools. Moreover, the questionnaire included some demographic questions on the indi-vidual (e.g. age and seniority) and questions on the public organisation (e.g. the area of the public sector and size of the organisation) The survey structure and the distribution of respondents are detailed in the annex. The survey was designed based on established categorisations, primarily those from the European Commission’s Public Sector Tech Watch, which monitors AI adoption, including GenAI, across EU public administrations. More information about the method-ology for the survey design and the categories can be found in the dedicated report. As this study aims to initiate discussion rather than provide a comprehensive quanti-tative statistical analysis, the results section focuses on descriptive statistics and prelim-inary trends. This approach enables an initial understanding to be gained of the patterns in GenAI adoption among public managers. 4 Results This section presents the descriptive statistics derived from the survey. Detailed statistics for each question are available in the annex, and the corresponding question numbers are given in brackets in the text. The results outlined in this section cross-analyse responses to illustrate key patterns in the data. 4.1 Adoption Trends First and foremost, this exploratory approach aims to understand the extent of GenAI use and the characteristics of its users. A key finding from the survey is that 30% of public managers actively use GenAI in their daily work (Q3.2). While this figure is considerable, it is equally notable that a further 44% of respondents (255 out of 576) reported plans to adopt these tools in the future. This indicates that adoption is already under way and likely to expand further as awareness and familiarity with these technologies increase. An analysis by age group (Q1.1) reveals a clear trend in anticipated GenAI adoption (Fig. 1). As expected, younger civil servants are at the forefront; however, a substantial share of respondents across other age brackets also express interest, suggesting that GenAI adoption in the public sector is not limited to the youngest cohort. This emphasises that interest in GenAI tools spans multiple generations. When the data are cross-referenced with the public manager’s awareness of GenAI tools (Q3.1), it becomes evident that those who do not use the tools are often less informed (Fig. 2). Fig. 2. Average self-assessed awareness of GenAI tools (on a scale from 0 to 100), by usage group (n = 576) Finally, adoption rates vary considerably across sectors (Q2.1), potentially reflecting differences in operational needs, regulatory constraints and levels of digital readiness (Fig. 3). The data show that certain governmental functions, such as defence and eco-nomic affairs, already report comparatively higher adoption rates than other sectors. In contrast, sectors such as education and housing show lower current usage, although a substantial number of respondents from these areas indicated a strong willingness to adopt GenAI in the future. However, given the low number of responses in some sectors, further research is needed to confirm these findings. It is worth noting that no clear differences in adoption rates were observed among government levels or administration sizes. 4.2 How Public Managers Use Generative Artificial Intelligence An intriguing aspect of the data concerns how public managers access GenAI platforms (Q3.4). Unsurprisingly, most report using freely available web-based versions. However, a notable proportion use internal tools provided by their administration. Further research is needed to better understand these responses. It would be valuable to explore what types of tools administrations are providing, whether they have begun integrating GenAI through application programming interfaces, whether they are paying for employee subscriptions and whether there is confusion among public managers regarding the distinction between internal and publicly accessible tools. When examining the tasks for which GenAI is used (Q3.5), the data reveal that GenAI is used to support a wide range of activities (Fig. 4). A considerable share of public managers reported using GenAI for most of the tasks listed several times per week. However, a notable number also reported ‘never’ using GenAI for each task. This suggests that usage varies widely and may depend on several factors, such as job responsibilities, individual skills and personal preferences. For example, depending on a manager’s writing proficiency, GenAI may be used regularly for text enhancement or not at all. Regarding the perceived benefits among current users (Q3.6), the data clearly show that the narrative around GenAI extends beyond mere efficiency gains (Fig. 5). While efficiency remains one of the most frequently cited advantages, many public managers also recognise GenAI’s potential to enhance creativity and support learning. As GenAI tools are still emerging and not yet fully integrated into most public-sector workflows, some of the reported benefits may be based on first impressions rather than on sustained use. Nonetheless, this finding highlights a growing interest in how GenAI might augment, not just automate, human capabilities. 4.3 Potential Barriers to Adoption The primary reason cited for not adopting or for lacking interest in adopting GenAI tools (Q3.3) is insufficient knowledge about how to use these tools, or even an unawareness of their existence, rather than data-sharing concerns or a lack of trust. In addition, environ-mental issues are considered of little importance. This finding suggests that the barriers to GenAI adoption may stem not from resistance to the technology itself but rather from the absence of formal capacity-building initiatives within public administrations. Some differences emerge when examining the potential barriers by awareness level (Q3.1) (divided into four classes for clarity) (Fig. 6). Respondents with a high level of awareness primarily refrain from using GenAI due to concerns about not sharing data with providers. Although this data-sharing concern spans all awareness levels, respondents with lower levels of awareness are more affected by a lack of knowledge and confidence. Interestingly, a lack of confidence in using GenAI tools is much less common among those with a high level of awareness, suggesting that there is a potential link between trust in these systems and an understanding of how they function. 4.4 Overall Opinion on Generative Artificial Intelligence The final data point concerns the survey respondents’ overall opinion on GenAI (Q3.7). The results can be interpreted in two ways (Fig. 7). On the one hand, a considerable proportion of the respondents did not express a clear opinion, with over 25% selecting a neutral response (‘neither disagree nor agree’) for each statement. On the other hand, the remaining respondents appear polarised between positive and negative views. While an overall optimistic stance slightly prevails, a noteworthy share of respondents expressed a negative perspective. The data reveal some differences when responses are compared across awareness levels (Fig. 8). Specifically, in terms of perceived usefulness, the results show that the more aware public managers are of GenAI tools, the more likely they are to find them helpful. This stresses the relevance of knowledge and skills in enabling effective use. In contrast, among those with very low awareness, there is a tendency towards more negative perceptions, including stronger agreement with statements such as ‘it is not trustworthy’ or ‘it is dangerous’. When comparing responses between users and non-users (Q3.2) some differences emerge (Fig. 9). First, a clear difference emerges in the intention to use GenAI exten-sively (Q3.2 – second response): non-users tend to disagree with this statement, whereas public managers who are using GenAI or plan to use it in the future are more likely to agree. Conversely, agreement with negative statements, such as concerns about danger or discrimination or a lack of trust, is slightly lower among users than non-users. 5 Discussion This section reflects on how GenAI tools may be reshaping technology adoption and use in the public sector, potentially challenging established assumptions in both research and practice. Given the preliminary and exploratory nature of this study, the discussion does not aim to provide an exhaustive analysis but instead draws attention to a set of intriguing insights that may inform future research. In particular, it seeks to identify emerging themes in how public managers perceive and engage with GenAI and to consider what these patterns might suggest for broader debates around digital transformation in the public sector. 5.1 From Working with to Thinking with The literature on AI has already pointed to a shift from machines simply supporting human tasks to more collaborative interactions between humans and machines (see, for example, ). This is often captured by the concept of ‘augmentation’, in which AI enhances human capabilities. Traditionally, augmentation has involved AI assisting humans in completing specific tasks, for instance by providing recommendations or explanations to support decision-making. In such cases, humans might also fine-tune the system to observe changes in output, using AI to improve efficiency or quality. With the advent of GenAI, this understanding of augmentation warrants rethinking. While AI still supports task completion, GenAI increasingly augments cognitive pro-cesses. Findings from our sample suggest that public managers experience GenAI not only as a tool for efficiency, but also as a catalyst for creativity and learning. While efficiency remains a core perceived benefit (consistent with dominant narratives around digitalisation in government), 61% of GenAI users reported using it weekly for learning purposes and 52% reported using it weekly for creative tasks. These patterns suggest a shift away from purely instrumental uses towards more intellectual and exploratory inter-actions with AI (i.e. AI enhances speed and quality). In light of this, GenAI becomes not only a tool for working with, but also a partner for thinking with. As Cantens sug-gests, there is a need to move towards ‘in-intellectualisation’, which involves integrating machines into the cognitive processes of public servants. The implications of this shift for public administration are still unfolding. On the one hand, innovation in the public sector is often driven by the individual creativity of public servants, who must navigate rigid structures and regulatory constraints. GenAI has the potential to support and amplify this creativity by offering new perspectives, accelerating ideation or reducing cognitive load. On the other hand, because GenAI is trained on historical data, it may reinforce dominant patterns and limit the emergence of genuinely novel ideas. This underscores the continued importance of maintaining space for unconventional, human-driven thinking to prevent the stagnation of the innovation cycle. 5.2 From Ensuring Use to Ensuring Fair Use Discussions on AI adoption in the public sector often focus on how to move beyond pilot programmes and achieve sustained use by public servants. Scholars have noted that many AI solutions remain confined to the laboratory due to constraints such as limited funding, technical complexity and the need for ongoing training and maintenance. However, the scenario considerably shifts when it comes to GenAI. These tools are widely available and can be used without specialised implementation processes, making them readily available for public servants. Our data support this shift: 30% of respondents already report using GenAI tools, and adoption is likely to grow, as 41% of non-users cite a lack of knowledge as the main barrier. These findings suggest that the main challenge is no longer ensuring adoption but rather understanding the diversity of use. In contrast with more traditional IT systems, GenAI does not have fixed usage paths or standardised workflows. Its open-ended and general-purpose nature enables a high degree of individual interpretation and improvi-sation. Two public servants may approach the same task with the same tool and apply it in entirely different ways or not at all. Moreover, if one public servant leaves and is replaced, the way they perform their duties may be different due to a different use of GenAI tools. There is no one-size-fits-all instruction manual. Moreover, the use of GenAI raises new concerns. For instance, 22% of respondents reported avoiding these tools due to data-sharing concerns, and, while less prominent in our sample, environmental impact was also noted. In this context, the question is no longer only about whether public servants use GenAI, but how it is being used and how to ensure that this use is equitable and responsible. Therefore, there is a need to shift the focus from ensuring use to ensuring fair use. ‘Fair’ in this context refers to both procedural equity, acknowledging that individuals may engage with GenAI differently or not at all, and the need to anticipate and mitigate well-known risks associated with its use. This might include addressing data-protection concerns, ensuring equal access to tools and guidance and recognising the variability in how technologies are enacted in practice. 5.3 From Technical to Cultural Transformation Digital government transformation is primarily a technical process: public servants learn to use the software, comprehend its advantages and limitations and apply it in their work. Public servants are expected to transition and adapt their workflows to new systems and understand how their roles will evolve with the introduction of such technology. GenAI, however, appears to introduce a cultural and affective dimension into digital transformation. Similar to past disruptions such as the rise of social media, although different in its use and workplace implications, GenAI elicits strong individual reactions. Our data reflect this dynamic. While approximately one quarter of respondents selected a neutral stance (‘neither disagree nor agree’) regarding their overall opinion of GenAI, the rest were polarised between enthusiasm and concern. Notably, 43% of respondents agreed (responding ‘somewhat agree’, ‘agree’ or ‘strongly agree’) with the statement that ‘GenAI must not be used’, a figure that is higher among those with lower awareness levels. These findings suggest that the initial impact is not only technical or procedural but also cultural, marking a more significant cultural transition than in the past and drawing attention to the need to engage with individual perceptions and promote critical understanding. This means equipping public servants with the skills to grasp how GenAI functions, where its limitations lie and how to make informed decisions about the use of this tool. Such efforts may help to alleviate public anxieties while reinforcing the idea that GenAI remains a tool to be adopted or rejected, depending on the context and specific needs and responsibilities. 5.4 From Mapped Software to Shadow Partner The implementation of new software, whether AI based or not, has traditionally been transparent or often formally codified within public administrations. With codification and mapping, monitoring and accountability have been relatively straightforward. When a tool is intended for a specific task and used in a predefined way, its deployment can be mapped, audited and assessed for compliance. This logic, however, is evolving. First, the question of what GenAI should be used for is now difficult to define. In our survey, we provided a non-exhaustive list of 10 potential tasks, and a considerable share of respondents reported using GenAI for all of them. Second, defining – and therefore monitoring – correct usage is challenging. GenAI is emerging as a silent or shadow partner, often operating under the radar, and thus is largely invisible to traditional forms of monitoring and evaluation. Its use is shaped by individual preferences, skills, levels of digital literacy and, as discussed earlier, personal attitudes. For instance, a public servant who is confident in writing may never use GenAI for drafting or editing, while others may rely on it heavily for the same tasks. This results in widespread variation in use that is neither standardised nor easily trackable. Therefore, this raises new governance questions. To what extent should GenAI use be monitored, documented or even regulated? The decision to facilitate or prohibit its use varies considerably across different administrations and countries, as shown in prior studies. Moreover, in many cases, use goes undeclared, making transparency difficult to enforce. Whether and how GenAI use should be declared, at least in principle, remains an open issue, and practical results are even more varied. 6 Conclusions GenAI is already being used by public servants in their daily work, marking a notable shift in the digital transformation of the public sector. This technological shift introduces characteristics that challenge existing assumptions and call for a renewed perspective and research framing. As with any emerging technology, it is essential to understand both its underlying logic and its practical implications, recognising where it aligns with past innovations and where it introduces new dynamics. This paper offers an initial empirical contribution by analysing survey data from 576 public managers across seven Member States. The distinctiveness of this dataset helps shed light on how GenAI is currently being used, perceived and understood by public servants. Through its exploratory approach, the study draws attention to early trends in GenAI adoption, the diversity of usage patterns and the evolving challenges. While the findings are preliminary, this study lays the groundwork for a rapidly developing research agenda (Table 1). 7 Annex (continued) 1. Aryfiyanto, H., Alamsyah, A.: Public service with generative AI: exploring features and. applications. In: Muharram, A. T., Rosalina, M., Kurniawati, D. and Yulianti, S. D. (eds), 2024 7th International Conference of Computer and Informatics Engineering (IC2IE), IEEE, Bali, Indonesia, pp. 1–7 2024. the linked source 2. Bright, J., Enock, F., Esnaashari, S., Francis, J., Hashem, Y., et al.: Generative AI is already. widespread in the public sector: evidence from a survey of UK public sector professionals. Digital Government: Research and Practice 6, 1–13 (2025). the linked source 0140 3. Cantens, T.: How will the state think with ChatGPT? The challenges of generative artificial. 4. European Commission: Directorate-General for Digital Services, Brizuela, A., Montino, C.. Galasso, G., Polli, G. et al., Adoption of AI, blockchain and other emerging technologies within the European public sector, Publications Office of the European Union, Luxembourg (2024). the linked source 5. European Commission: Joint Research Centre, Tangi, L., Combetto, M., Hupont Torres, I.. Farrell, E. et al.: The potential of generative AI for the public sector: Current use, key questions and policy considerations, European Commission, Ispra (2024) JRC139825 6. Janssen, M.: Responsible governance of generative AI: conceptualizing GenAI as complex. 7. Tangi, L., Janssen, M., Benedetti, M., Noci, G.: Digital government transformation: a struc-. tural equation modelling analysis of driving and impeding factors. Int. J. Inf. Manage. 60, 102356 (2021). the linked source 8. van Noordt, C., Tangi, L.: The dynamics of AI capability and its influence on public value. 9. Beltran, M. A., Ruiz Mondragon, M. I., Han, S. H.: Comparative analysis of generative AI risks in the public sector. In: Liao, H.-C., Duenas Cid, D., Macadar, M. A. and Bernardini, F. (eds), Proceedings of the 25th Annual International Conference on Digital Government Research, ACM, Taipei, Taiwan, pp. 610–617 (2024). the linked source 7125 10. Council of the European Union, ChatGPT in the Public Sector – Overhyped or over-. looked?, Brussels (2023). the linked source 11. European Commission: Joint Research Centre, Grimmelikhuijsen, S., Tangi, L.: What factors. influence perceived AI adoption by public managers? A survey among public managers in seven EU countries, Publications Office of the European Union, Luxembourg (2024). the linked source 12. European Commission, Coordinated plan on artificial intelligence 2021 review. Euro- pean Commission website (2021). the linked source nated-plan-artificial-intelligence-2021-review 13. European Commission: Joint Research Centre, Tangi, L., Combetto, M. and Martin Bosch, J.. Methodology for the Public Sector Tech Watch use case collection: Taxonomy, data collection, and use case analysis procedures, Publications Office of the European Union, Luxembourg (2024). the linked source 14. Maragno, G., Tangi, L., Gastaldi, L., Benedetti, M.: AI as an organizational agent to nurture:. 15. Mikalef, P., Lemmer, K., Schaefer, C., Ylinen, M., Fjørtoft, S.O., et al.: Enabling AI capabil-. ities in government agencies: a study of determinants for European municipalities. Gov. Inf. Q. 39, 101596 (2021). the linked source Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (the linked source), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made. The images or other third party material in this chapter are included in the chapter’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the chapter’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.