Latest / 1 More Paper / From understanding resistance to fostering acceptance: how mindsets impact GenAI adoption — Wieland et al. (2026)

Transcript

From understanding resistance to fostering acceptance: how mindsets impact GenAI adoption — Wieland et al. (2026) · 1 More Paper

• You’re listening to “From understanding resistance to fostering acceptance: how mindsets impact GenAI adoption,” by J. Wieland, L. Keating, and A. Mohnen. Published in 2026. ISSN: 0144-929X (Print) 1362-3001 (Online) Journal homepage: the linked source Jennifer Wieland, Lauren Keating & Alwine Mohnen To cite this article: Jennifer Wieland, Lauren Keating & Alwine Mohnen (11 Aug 2026): From understanding resistance to fostering acceptance: how mindsets impact GenAI adoption, Behaviour & Information Technology, DOI: 10.1080/0144929X.2026.2706666 © 2026 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group Published online: 11 Aug 2026. Submit your article to this journal Article views: 551 View related articles View Crossmark data From understanding resistance to fostering acceptance: how mindsets impact GenAI adoption b and Alwine Mohnen a, Lauren Keating a Jennifer Wieland aSchool of Management, Technical University of Munich, Munich, Germany; bDepartment of Law, Management, and Social Sciences, Emlyon Business School, Lyon, France ABSTRACT. Individual resistance to adopting generative artificial intelligence (GenAI) jeopardises its successful implementation in organisations. Across two studies, this paper aims to explain why some individuals embrace GenAI while others oppose it, and investigates whether a growth mindset can facilitate its adoption in organisations. In Study 1, we surveyed 159 German employees to establish a Theory of Planned Behavior (TPB) model explaining the variance in GenAI adoption. The structural equation modelling results revealed that attitudes, subjective norms, and perceived behavioural control predict the intention to adopt GenAI, with attitudes exhibiting the strongest association. While Study 1 enhances understanding of GenAI adoption, Study 2 seeks to promote it. Given the challenges posed by GenAI, we tested whether a growth mindset – the belief that abilities can be developed – can boost GenAI adoption, as it fosters openness to challenges and change. In a randomised experiment with 389 German employees, we observed that a growth mindset intervention positively influences attitudes, subjective norms, and perceived behavioural control. Overall, the findings provide insight into the role that mindsets play in approaching or avoiding digital transformations, as well as offer a path for diminishing people’s reluctance to embrace them. 1. Introduction. Generative Artificial Intelligence (GenAI) is poised to revolutionise the way people work, projected to add $2.6 to $4.4 trillion to the global economy. A recent study found that 90% of employees using GenAI improved their effectiveness, performing 40% better overall than their peers who did not use it. GenAI is therefore likely to become an essential competitive edge, though to leverage its potential, organisations first need to institutionalise GenAI. However, the successful implementation of new technologies relies heavily on the acceptance of users. Adopting GenAI is not merely integrating another IT tool but a transformative change that will fundamentally reshape the workplace, altering not just processes but entire operating models. Such a change can only be successful if supported by employees. Yet, individuals are often resistant to change, with ARTICLE HISTORY Generative AI; Technology adoption; Theory of Planned Behavior (TPB); Mindsets 70% of employees reporting being tired of change initiatives at the workplace. One reason for change resistance is the fear of losing control, which adopting intelligent technologies might imply. This poses a risk to organisations: Employees’ resistance to GenAI might result in a waste of resources and hinder innovation. Understanding when and why individuals adopt versus oppose intelligent technology is thus critically important. Past research reveals the various factors potentially influencing reactions to new technologies, such as age, trust, threat perception, and perceived risk, though such research also remains somewhat inconclusive (Bigman and Gray 2018; Clarke 2019; Kelly, Kaye, and Oviedo-Trespalacios 2023; Stein, Liebold, and Ohler 2019; Waytz, Heafner, and Epley 2014). Moreover, most research on the adoption of AI and intelligent technologies focuses on more ‘traditional’ AI technologies, such as machine learning, rather than GenAI. Yet, this distinction could be crucial because GenAI models produce output that appears ‘human-like’, and individuals react differently when they perceive technology as having higher cognitive and emotional capabilities. Further, most existing research focuses on AI acceptance from a consumer rather than employee perspective, with only a handful of exceptions (e.g. Greiner, Jovy-Klein, and Peisl (2021) and Xu and Wang (2021)). Taken together, employees’ adoption of generative AI in organisational contexts remains underexplored, with existing studies having largely focused on identifying correlates of technology adoption rather than examining how adoption-related attitudes and behaviours can be influenced through targeted interventions. The reality of increasing integration of GenAI in organisations, together with the fact that employees are often resistant to change, suggests that organisations may waste considerable resources (e.g. time and money) in the hope that employees will adopt this new technology. As a step toward addressing this problem, we first explore what drives or hinders employees’ usage of GenAI in an organisational context. To do so, we draw on the Theory of Planned Behavior (TPB) in Study 1 to explain GenAI adoption. The TPB is a widely applied framework to explain behaviour in general and technology adoption more specifically, wherein attitudes towards a specific behaviour, the perception of social norms, and perceived behavioural control are proposed to predict whether individuals intend to and then actually perform the behaviour. Although understanding the factors that can predict GenAI adoption is needed and timely, it is also important to understand how attitudes and behaviour regarding GenAI can be changed, so as to minimise potential organisational waste and employee demoralisation. While identifying psychological predictors of GenAI adoption requires a theory-driven, explanatory approach, examining whether such predictors can be actively influenced necessitates an experimental design. As GenAI poses a grand challenge for many employees, requiring completely new skills and openness to change, a growth mindset – the belief that personal attributes and abilities can be developed – could be a promising approach, given that individuals with a growth mindset are more open to challenges and change. Indeed, Dang and Liu (2022) observed that a growth mindset enhances acceptance of robots. In line with such burgeoning research, in Study 2, we conduct an experiment to examine whether an induced growth mindset can foster the adoption of GenAI. The present research pursues two primary research objectives: to explore a psychological mechanism that could explain adoption of GenAI and to evaluate whether a growth mindset can serve as a lever to promote GenAI adoption. In pursuing these two objectives, we apply and empirically test a TPB-based model to explain variation in the adoption of GenAI in an organisational context, thereby extending the application of the TPB to emerging technologies. This research extends the TPB by incorporating mindset as an upstream psychological mechanism shaping the formation of TPB belief structures. Specifically, we conceptualise mindset as an antecedent influencing attitudes, subjective norms, and perceived behavioural control, which in turn drive intention and behaviour in the emerging context of GenAI adoption. We also apply mindset theory to an unexplored context and population for growth mindset interventions, namely adoption of GenAI by employees in organisations. Finally, we offer practical guidance for organisations to understand and support employee adoption of GenAI. We begin by reviewing the literature on the adoption of GenAI. In Study 1, we then outline relevant research on the Theory of Planned Behavior, from which we derive our initial set of hypotheses and test our theoretical model. In Study 2, we demonstrate how mindset theory relates to technology adoption, as well as describe our experiment to explore whether mindsets can foster the adoption of GenAI. Finally, we discuss the research and practical implications of our work. 2. Determinants of GenAI adoption. Artificial intelligence (AI) – that is, capabilities allowing computational devices to perform tasks typically requiring human intelligence – has already had considerable influence on individuals, organisations, and societies. Yet to fully leverage the potential of such technologies, user adoption is key. Past studies demonstrate broad consensus about certain factors fostering AI acceptance, such as being part of a younger generation, having higher levels of trust, as well as perceiving lower level risk and threats. However, research remains inconclusive about the human response to intelligent technologies, particularly because attitudes are often ambivalent. While intelligent technologies might be perceived as convenient and helpful, these technologies can also be interpreted as a challenge or threat. Context thus plays a crucial role: While individuals may hesitate to adopt intelligent technologies in certain areas, such as caring for children and the elderly, providing health care services, or autonomous driving, they might embrace similar technologies in different contexts, such as parcel delivery or more generally when intelligent technologies are socially communicative. Many of these studies focus on the adoption of specific technologies, such as AI devices or robots, and on the perspective of consumers or end users. Some exceptions focused on different populations include the work of Greiner, Jovy-Klein, and Peisl (2021), who found that employees are more open to using AI at work if AI systems are transparent. Ochmann and Laumer (2020) revealed factors fostering the acceptance of AI among job seekers, such as performance expectancy of the AI or their propensity to use AI. Xu and Wang (2021) also identified tasks in which robot lawyers can replace humans, such as data collection and case analysis. Despite the tremendous (potential) impact of AI on employees and organisations, research on the acceptance of AI in organisations is limited. Given that technology adoption is context-specific and that attitudes towards smart technologies in organisations have become more negative, exploring the adoption of AI in organisations is imperative. More recently, GenAI has emerged as a new form of AI. The rapid rise of large language models like ChatGPT highlights the diverse use cases for which GenAI can be beneficial, such as software development, writing reports, and learning, where GenAI can produce ‘human-like’ output. As interactions with GenAI models resemble those with humans, individuals might perceive them as a threat and feel uneasy. The uncanny valley theory suggests that as technologies become more human-like, they elicit positive emotional responses up to the point at which they seem eerie and unsettling. This dip in comfort, or the ‘uncanny valley’, is overcome when the likeness becomes indistinguishably human. Understanding when individuals feel comfortable to embrace such digital transformations is thus a complicated issue, though one that nonetheless needs further examining in the context of emerging GenAI. 3. Study 1. 3.1. Theoretical background: Theory of Planned. Behavior in predicting GenAI adoption The Theory of Planned Behavior (TPB) is a widely used model to assess technology adoption. It builds on the Theory of Reasoned Action (TRA) adding perceived behavioural control to attitudes and subjective norms as predictors of behavioural intention. Originally developed as a general model of behaviour, it is now widely used for technology adoption. Another popular model to predict technology adoption is the Technology Acceptance Model (TAM) developed by Davis (1989). While both TAM and TPB aim to predict behavioural intention, they do so through different ways: While TPB includes a broad set of constructs that also capture social and motivational dimensions of behaviour, TAM focuses only on perceived usefulness and ease of use, neglecting social and motivational predictors of behaviour. In addition, the Unified Theory of Acceptance and Use of Technology (UTAUT) offers an integrative framework, combining elements from TAM, TPB, and other models. UTAUT emphasises performance-related constructs such as effort expectancy and facilitating conditions, and has often been applied in structured, organisational contexts involving information system implementation. However, GenAI adoption is rather unstandardised and emerges across varied tasks and functions. The TPB is well suited to modelling such voluntary, individual-level behaviour and allows for context-specific measurement and integration of psychological constructs like mindsets. We thus draw on the TPB, as it captures voluntary, context-dependent technology adoption and is particularly suited to modelling unstructured and emerging behaviours such as GenAI use. The TPB also enables the integration of psychological mechanisms such as mindset. By organising diverse findings on technology adoption around attitudes, subjective norms, and perceived behavioural control, the TPB provides a coherent theoretical structure for integrating prior research and for guiding the selection of key variables in the context of GenAI. Although past research has revealed the significant explanatory power of the TPB to predict technology adoption, the model needs to be verified for each target behaviour and target population, and the relative importance of each TPB construct needs to be assessed. Below we argue how the TPB framework can be applied to understand adoption of GenAI in an organisational context and introduce a conceptual model depicting the hypothesised relationships (see Figure 1). Specifically, the model conceptualises GenAI adoption as a function of employees’ attitudinal evaluations, perceived social expectations, and perceived control over using the technology, which together shape intention and subsequent usage. Attitudes. Attitudes describe the overall evaluation of a behaviour as positive or negative. According to TPB, attitudes affect behavioural intentions in general and technology adoption specifically. However, attitudes towards intelligent technologies can be ambivalent – simultaneously perceived as convenient, challenging or even threatening. The Uncanny Valley Theory might explain this by suggesting that smart technologies elicit negative emotional reactions as they become more human-like, creating a feeling of unease known as the ‘uncanny valley’. We argue that GenAI might cross this ‘uncanny valley’, due to its high human-likeness, which could positively influence attitudes towards smart technologies and in turn strengthen adoption intentions. Previous evidence on traditional AI substantiates this, finding a positive association between attitudes and intention. Within the Theory of Planned Behavior, attitudes represent the core evaluative component driving behavioural intention. Based on this core TPB assumption and prior findings on intelligent technologies, we therefore expect that more positive attitudes toward GenAI are associated with stronger intentions to use it. Hypothesis 1: Positive attitudes towards GenAI are associated with increased intention to use GenAI. Subjective Norms. Subjective norms describe the social pressure to perform or avoid a particular behaviour, typically shaped by important referents such as family, friends or peers. Individuals tend to adopt the norms and culture of their social network, which increases their sense of belonging. Following the Computers as Social Actors Theory, we attribute human-like attributes to computers, making social norms a relevant construct when assessing technology acceptance. The extent to which social norms influence behavioural intention also depends on the context, such that norms exert more influence when individuals have only limited experience with an issue or behaviour. According to the TPB, subjective norms influence behavioural intention by shaping individuals’ motivation to comply with perceived social expectations regarding a behaviour. Empirical research in technology adoption contexts consistently finds that subjective norms positively predict usage intentions, particularly under conditions of novelty and uncertainty. Accordingly, if employees sense that their peers embrace GenAI, they are more likely to form stronger intentions to use it. Hypothesis 2: Positive subjective norms regarding GenAI are associated with increased intention to use GenAI. Perceived Behavioral Control. Perceived Behavioral Control (PBC) refers to individuals’ perception about how easy or difficult it is to perform a behaviour, shaped by factors such as self-efficacy and facilitating conditions. According to the TPB, PBC not only affects behavioural intention but also behaviour directly, as individuals are more likely to engage in a specific behaviour if they believe they have control over it. We argue that PBC is particularly relevant for GenAI adoption for two reasons. First, resistance to change can stem from a feeling of losing control, a sentiment often experienced with intelligent technologies. Employees who feel ‘in control’ might be more likely to use GenAI. Second, having the necessary resources, such as adequate IT infrastructure, is crucial for successful implementation. Therefore, if individuals perceive that their organisations provide these resources, they might be more likely to feel ‘in control’ and thus engage with the technology. Taken together, this suggests that employees who perceive GenAI use as manageable and adequately supported – through both personal capabilities and organisational resources – are more likely to form intentions to use the technology and, potentially, to enact these intentions. Hypothesis 3: Perceived behavioral control over using GenAI is associated with (3a) greater intention to use GenAI and with (3b) greater usage of GenAI. Behavioral Intention. Behavioural intention is a crucial determinant of real behaviour. The stronger the intention, which is predicted by attitudes, subjective norms, and perceived behavioural control, the more likely a particular behaviour will be enacted. Indeed, research on technology acceptance reveals that intention to use a technology predicts its usage. We thus argue that the intention to use GenAI explains self-reported GenAI usage, as behavioural intention occupies a central position in the TPB, serving as the most proximal determinant of actual behaviour. Hypothesis 4: Intention to use GenAI is positively associated with usage of GenAI. In sum, these hypotheses describe an integrated theoretical model in which attitudes, subjective norms, and perceived behavioural control jointly shape employees’ intentions to use GenAI and, in turn, their actual usage behaviour. This model extends prior technology adoption research by applying the TPB to the specific context of generative AI in organisational settings. 3.2. Method. 3.2.1. Research design. To test our conceptual model, we first conducted a pilot study (n = 34) to develop a TPB questionnaire tailored to our target population and behaviour. We then used a structural equation model to assess the extent to which the TPB explains GenAI adoption (Hypothesis 1-4), using a representative sample of German white-collar employees (n = 159). 3.2.2. Participants and sampling. The sample included 159 employees from Germany, currently in part-time or full-time employment, with 55.3% of the sample being female and aged from 19 to 65 (Mage = 42.77, SDage = 13.59). We decided to focus on white-collar employees as their attitudes towards technology are generally more favourable and they are more likely to encounter AI at work. On average, participants had 19.37 years of 1German Federal Statistical Office (2022). work experience (SDexperience = 13.08), with 42% holding a university degree (n = 66) and 32% holding a leadership role (n = 51). Table 1 summarises the sample and population demographics, showing that the sample represents German white-collar employees in terms of age, gender, and industry. Demographic details were self-reported. We recruited the sample via the online panel provider Bilendi, who offered financial compensation equivalent to an hourly rate of around six euros. Online populations are often utilised in research focused on work contexts, as results are often comparable to those of lab or field studies and such samples have been used to successfully replicate past research. More specifically, mindset as well as technology adoption research frequently leverages online populations to study employees in work-related contexts, as these populations are readily accessible online and produce results comparable to lab or field studies. 3.2.3. Data collection. After receiving ethics approval from the German Association for Experimental Economic Research (No. XWzwNGG5), we recruited eligible participants to take part in a survey on ‘Technology at the Workplace’ from March 13 to March 26, 2024. Using screening questions, we verified that participants were full – or part-time white-collar employees. For data quality assurance, we excluded participants who failed quality, attention, and integrity checks; whose corresponding demographic quota was already full; whose data quality was insufficient; or who dropped out voluntarily. Participants answered the questionnaire either in English or German, which was the preferred choice. We also included a definition of GenAI, as individuals’ interpretation of AI varies, which can impact their reaction to a technology. These procedures ensured that only eligible participants with sufficient data quality and an adequate understanding of generative AI were included in the final sample. 3.2.4. Measures. We developed the TPB questionnaire following the suggested process from Ajzen, which encourages researchers to construct behaviour-specific and context-relevant measures rather than relying on generic items. First, we defined the target behaviour as ‘Regularly using GenAI tools and technologies in daily work tasks over the next six months’ and the research population (i.e. white-collar employees in Germany). To construct context-relevant items, we conducted a pilot study in which participants responded to both open- and close-ended questions designed to capture salient behavioural, normative, and control beliefs. We analyzed these responses to identify key themes and simultaneously tested the initial items quantitatively. This process yielded an initial set of six items per construct, which we refined to four per construct based on psychometric validation. This integrated approach ensured that our final TPB measures were both theoretically grounded and empirically adapted to our specific research context. Attitudes towards GenAI were assessed with four items, using a seven-point scale. Sample items include ‘I feel confident that GenAI can lead to (Unsatisfactory – Outstanding) results in my job’ and ‘I find GenAI technology to be (Unexciting – Exciting)’. Cronbach’s Alpha was 0.87. Subjective norms regarding GenAI were also evaluated using four items on a seven-point scale. Sample items include ‘If I don’t use GenAI, my colleagues would (Approve – Disapprove)’ and ‘My friends and family believe that using GenAI is (Unimportant – Important)’. Cronbach’s Alpha was 0.77. Perceived behavioural control towards GenAI was assessed with four items, using a seven-point scale. Sample items include ‘My technical skills (Hinder – Enable) me to use GenAI for my work tasks’ and ‘GenAI tools which are helpful for my work tasks are (Not accessible – Accessible)’. Cronbach’s Alpha was 0.86. Intention to use GenAI was also measured using four items on a seven-point scale. Sample items include ‘I (Do not intend – Intend) to regularly use GenAI in my work over the next six months’ and ‘My goal is to integrate GenAI technology into (Limited tasks – All tasks)’. Cronbach’s Alpha was 0.90. Usage of GenAI was evaluated using three items on a seven-point Likert scale. While most studies exclude measures of usage, relying solely on behavioural intention, we utilised items on past usage, which is often a close proxy of real behaviour. However, it should be noted that this measure captures self-reported usage rather than directly observed behaviour. Sample items include ‘In the past three months, I have used GenAI at work’ and ‘In the past three months, GenAI significantly supported my daily tasks at work’. Cronbach’s Alpha was 0.96. Controls. We also included demographic and employment control variables, as past research indicates that participants’ demographic characteristics influence the acceptance of technology. 3.2.5. Analytical approach. Given that our TPB-based model specifies a pre-defined measurement structure, we followed Anderson and Gerbing’s (1988) recommended two-step SEM approach and evaluated the measurement model using confirmatory factor analysis (CFA), which is intended for situations where the factor structure is known a priori. Thus, we first conducted a CFA to test the measurement model, assessing the reliability as well as convergent and discriminant validity. Subsequently, we estimated the hypothesised relationships using covariance-based structural equation modelling (CB-SEM). CB SEM is particularly suitable for theory testing with reflective constructs and pre-defined models, as it focuses on reproducing the observed covariance structure and allows for a rigorous evaluation of model fit and parameter estimates. STATA (version 16.1) was used for data analysis. 3.3. Results. Table 2 displays the descriptive and correlational statistics of Study 1. 3.3.1. Confirmatory factor analysis for the. measurement model We first tested the measurement model with a CFA, assessing the internal consistency reliability, convergent validity, and discriminant validity of the items and constructs. Results in Table 3 demonstrate that Cronbach’s Alpha values for all assessed constructs were equal to or above 0.77, exceeding the cut-off value of 0.70 stated by Hair, Ringle, and Sarstedt (2011). The composite reliability (CR) should also be above 0.70, which was the case for all constructs. Both the Cronbach’s Alpha and the composite reliability values thus offered evidence of internal consistency. To assess the convergent validity, the average variance extracted (AVE) should be greater than 0.50, which we observed for all constructs. Additionally, all factor loadings were significant (p > 0.001), further supporting convergent validity. As the square roots of the AVE values for all constructs were larger than the correlations with other constructs, we can assume discriminant validity. Considering the fit statistics, our initial model indicated a mediocre fit (CFA: χ2 = 373.556, χ2/df = 2.621, SRMR = 0.068, RMSEA = 0.102, CFI = 0.905, TLI = 0.886). While some test statistics met the recommended threshold, such as the SRMR being smaller than the suggested threshold of.08, the CFI being above 0.9 and the ratio of Chi-square to degrees of freedom being below 3, some other statistics did not meet the criteria for good fit. While the significance of the Chi-square test is often biased by the sample size, the TLI should exceed.9 and the RMESA should be <.08, which was not the case in this model. However, goodness of fit is essential because a model that inaccurately reflects the covariance structure cannot provide reliable parameter estimates for the relationships between variables. Therefore, we considered modification indices and accordingly added covariances between error terms, a type of model re-specification that is frequently required in CFA. Such covariances were introduced where items shared similar wording or content, leading to additional measurement-specific variance beyond the focal latent constructs. For example, items capturing closely related aspects of perceived capability or social expectations may exhibit correlated residuals due to overlapping phrasing or reference points. In line with established SEM practice, these modifications were applied conservatively and without altering the theoretical structure of the TPB measurement model. Additionally, we observed that our model did not meet the normality assumptions, as the skewness exceeded the acceptable range. Therefore, we conducted a Satorra-Bentler adjustment, which corrects the chi-square statistic and standard errors for non-normality to compensate for potential inflation caused by non-normal distributions. The updated model indicated good fit (CFA: χ2 = 217.185, χ2/df = 1.632, SRMR = 0.063, RMSEASB = 0.063, CFISB = 0.961, TLISB = 0.950), with a Chi-square to degrees of freedom ratio below < 3, SRMR and RMSEA <.08, and TLI and CFI respectively matching the higher threshold of.95. To summarise, this measurement model exhibited enhanced fit, as well as reliability, convergent validity, and discriminant validity. 3.3.2. Structural equation modelling. In the second step we tested Hypotheses 1–4 using covariance-based structural equation modelling (CB-SEM) with a Sattora Bentler correction. The structural model indicated excellent fit (SEM: χ2 = 218.796, χ2/df = 1.620, SRMR = 0.064, RMSEASB = 0.063, CFISB = 0.961, TLISB = 0.951). All goodness of fit statistics exceeded the suggested thresholds: Ratio of Chi-square to degrees of freedom < 3; SRMR <.08, TLI >.95; CFI >.95; RMSEA <.08. Consistent with Hypothesis 1, attitudes towards GenAI were positively associated with intention to use it (ß = 0.72, p < 0.001). Hypothesis 2, that subjective norms regarding GenAI are associated with increased usage intention, was also supported (ß = 0.36, p < 0.01). Similarly, perceived behavioural control to use GenAI was associated with increased usage intention (ß = 0.33, p < 0.05), though not increased usage (ß = 0.33, p > 0.05), supporting Hypothesis 3a, but not Hypothesis 3b. Lastly, usage intention was positively associated with usage, supporting Hypothesis 4 (ß = 0.94, p < 0.001). Overall, Hypotheses 1, 2, 3a, and 4 were supported, whereas p < 0.001, p < 0.01, p < 0.05; Square roots of average variance extracted (AVEs, in bold); n = 159. Hypothesis 3b was not supported. Figure 2 and Table 4 summarise the results. 3.4. Discussion. By developing and testing a Theory of Planned Behavior model of GenAI usage, we established a framework for explaining adoption intentions and self-reported usage. Consistent with our hypotheses, the three TPB constructs – attitudes (H1), subjective norms (H2), and perceived behavioural control (H3a), predicted behavioural intentions to use GenAI. An unexpected finding was the absence of a direct effect of perceived behavioural control on self-reported GenAI usage (H3b). While TPB posits that perceived behavioural control can influence behaviour directly, this depends on how accurately perceived control reflects actual control. In the context of GenAI adoption, where usage may be constrained by organisational policies, task requirements, or access to approved tools, perceived behavioural control may play a more prominent role in shaping intentions than in directly predicting actual usage, suggesting a more indirect role via intention in such constrained and uncertain environments. This is consistent with the TPB, which suggests that perceived control is more likely to influence behaviour indirectly via intention when it does not fully reflect actual control. However, behavioural intention was associated with usage (H4). While the model helps to explain adoption intentions and usage, the TPB constructs may not be the ultimate cause of behaviour. Indeed, they could be shaped by underlying psychological mechanisms. We thereby introduce mindset as one such upstream factor influencing attitudes, subjective norms and perceived behavioural control. We conducted a controlled experiment in Study 2 to explore whether mindsets causally influence these TPB constructs. This approach allows us to move beyond merely explaining GenAI adoption and toward identifying ways of promoting it. 4. Study 2. 4.1. Theoretical background: mindsets as a lever. to foster GenAI adoption While the TPB is effective in explaining behaviour, its constructs are influenced by background factors, such as demographic, psychological, social or environmental variables. This prompts a need to identify these factors and alter those that have the capacity to be changed. Among these, we focus on psychological mechanisms and argue that mindset represents a particularly relevant and malleable factor shaping how individuals form attitudes, perceive social norms, and assess their capability to engage with GenAI. Individuals with a growth mindset are more open to challenges and change. Initial evidence also suggests that mindsets about minds – the belief about whether mental capacities to experience and act are malleable – affect humans’ responses to robots. Further research is nonetheless needed to understand if this effect extends to GenAI, especially following the rapid rise of tools like ChatGPT, and whether it can be replicated for mindsets of intelligence (i.e. the most researched mindset construct; Rammstedt, Grüning, and Lechner 2022). Individuals with a growth mindset believe that their personal attributes (e.g. intelligence or personality) and abilities (e.g. IT skills) can be developed, while those with a fixed mindset presume they are static. Mindsets do not correlate with actual level of abilities and are only weakly correlated with personality. Although mindsets tend to be relatively stable, they can be modified with specific interventions. Given that these mindset interventions require relatively little effort and their effects are strongest when a challenge is present, inducing growth mindsets might be an effective way to increase GenAI adoption. In the following section, we argue how growth mindsets can influence the attitudes, subjective norms, and perceived behavioural control of GenAI, summarised by Figure 3. Attitudes. Attitudes towards intelligent technologies are often ambivalent, being simultaneously perceived as helpful and as a challenge or threat. We argue that individuals with a growth mindset are more likely to have favourable attitudes towards GenAI for two reasons. First, adopting new technologies necessitates change, to which individuals with a growth mindset respond more positively. Research by Chiu et al. (1997) revealed that individuals holding a growth mindset provide more constructive feedback following unexpected changes. Similarly, Heslin, Latham, and VandeWalle (2005) demonstrated that individuals with a growth mindset are less likely to rigidly adhere to their initial impression, but rather update their judgments in response to further input. Second, using GenAI may appear challenging and individuals with a growth mindset are more inclined to embrace challenges. When presented with the choice between an easy task devoid of learning opportunities, and a more difficult task offering a chance to learn but also risking potential failure, individuals with a growth mindset tend to opt for the challenging option. In the face of challenges, individuals with a growth mindset are also more persistent, seek out different strategies, and are more likely to learn from mistakes. Given that GenAI similarly represents a novel challenge, we hypothesise that growth mindsets prompt more positive attitudes towards GenAI. Hypothesis 5: Growth relative to fixed mindsets are associated with more positive attitudes towards GenAI. Subjective Norms. Subjective norms represent the social pressure to perform or avoid a specific behaviour, often resulting in individuals extrapolating and applying the norms of their social network to their own situation. These norms are based on individuals’ perceptions and are thus subjective. Building on the false consensus effect, individuals tend to assume that others are more similar to themselves than they actually are. We thus argue that if individuals have a growth mindset and are thus more open toward GenAI, they will tend to believe that their social network shares this openness, leading to a more positive perception of subjective norms regarding GenAI. Hypothesis 6: Growth relative to fixed mindsets are associated with perceiving social norms regarding GenAI as more favorable. Perceived Behavioral Control. Perceived behavioural control (PBC) is influenced by factors like self-efficacy and facilitating conditions. Self-efficacy describes individuals’ belief in their own ability to complete a certain task or perform a particular behaviour. We argue that self-efficacy is bolstered by growth mindsets, as individuals with a growth compared to a fixed mindset are more persistent after setbacks, have a more positive view of effort, are more receptive to critical feedback and tend to have more favourable judgments of complexity. Further, we suggest that growth mindsets influence the perception of available resources, because individuals with a growth mindset are less likely to feel ‘helpless’ when facing obstacles, but rather more methodically adjust their strategies to overcome them. We thus argue that mindsets will similarly influence perceived behavioural control. Hypothesis 7: Growth relative to fixed mindsets are associated with higher levels of perceived behavioral control over GenAI usage. Control Variables. While mindsets may be a particularly promising lever to foster GenAI adoption, other demographic, psychological, social, or environmental variables might also influence the Theory of Planned Behavior constructs. In addition to demographic and employment control variables, we examine the influence of prior knowledge of GenAI, which has been identified as a predictor of technology adoption. Furthermore, Dang and Liu (2022) found that growth mindsets are associated with individuals’ general feelings toward smart technologies, and Hansen, Saridakis, and Benson (2018) have identified risk preferences as a predictor of technology adoption, highlighting these constructs as relevant control variables. 4.2. Method. 4.2.1. Research design. We conducted a controlled experiment to test whether an induced growth mindset can influence attitudes, norms, and perceived control and thus foster the adoption of GenAI (Hypothesis 5-7) among German white-collar employees (n = 389), using hierarchical regression modelling. 4.2.2. Participants and sampling. In Study 2 we recruited a different representative sample of 389 employees in Germany (51.16% female, age 18– 72, Mage = 42.97, SDage = 12.70). Participants had, on average, 19.92 years of work experience (SDexperience = 13.13), 47% held a university degree (n = 181) and 34% were in leadership positions (n = 133). Table 5 summarises population and sample demographics. We recruited our sample via the online panel provider Bilendi, who compensated participants financially at an hourly rate of around six euros. This sampling strategy is appropriate for Study 2, as white-collar employees are likely to encounter generative AI in their daily work and are therefore a relevant population for testing interventions aimed at fostering GenAI adoption. 4.2.3. Data collection. As in Study 1, we invited our sample to participate in a survey on ‘Technology at the Workplace’. After 1German Federal Statistical Office (2022). receiving ethics approval from the German Association for Experimental Economic Research (No. XWzwNGG5), we collected data from March 13 to April 6, 2024. Participants first completed screening questions and read a brief definition of GenAI. They then responded to questions about their chronic (i.e. baseline) mindset, followed by a random allocation to a growth or fixed mindset treatment group. Manipulation checks ensured that the treatments were successful. In the second part, participants answered questions related to the TPB constructs, followed by measures gauging the control variables. We applied the same data exclusion criteria as in Study 1. The questionnaire was again available in both English and German, with most participants choosing German. These procedures ensured high data quality, participant eligibility, and adequate comprehension of generative AI, and allowed for methodological consistency across both studies. 4.2.4. Measures. Chronic mindsets. We used four fixed-oriented items from the mindset of intelligence scale proposed by Dweck (1999), the most prevalent scale applied in mindset research. Participants rated their level of agreement on a six-point Likert scale from 1 (Strongly Disagree) to 6 (Strongly Agree). A sample item is ‘You can learn new things, but you can’t really change your basic intelligence.’ Mindset treatment. We developed an online intervention drawing upon previous research effectively inducing mindsets. While initial mindset interventions were conducted offline, today many interventions are online and self-administered. We adjusted the design from Yeager et al. (2016) to be appropriate for an adult population. Participants, who were blind to the treatments and the objectives of the study, were randomly allocated to either a growth or fixed mindset group. In the growth mindset group, participants first read a two-page scientific testimonial about how intelligence can be developed. Below is an excerpt of the growth mindset testimonial: [‘... New research shows that the brain works like a muscle – it changes and gets stronger as you use it. Scientists have shown how the brain grows and gets stronger as you learn... ’] In contrast, the fixed mindset group read a testimonial stating that intelligence is fixed: [‘... Human intelligence is largely fixed and develops along a predetermined path. A study shows that the intelligence profile solidifies early and, despite all efforts, rarely changes significantly... ’] After the testimonial, participants answered quality checks to ensure they had engaged with and understood the testimonial. Next, we employed ‘Saying-Is-Believing’ exercises using open-ended questions. In the growth mindset group, participants reflected on a situation where they had accomplished something they once judged unattainable and described the actions taken to reach this achievement. Participants in the growth mindset group then shared advice with a friend facing difficulties in progressing in a particular situation, summarising the scientific testimonial and outlining concrete action steps. In comparison, participants in the fixed mindset group reflected on why it is important to diagnose an individual’s level of intelligence. We employed quality checks to ensure comprehension and engagement in the reflection exercises. We debriefed participants about the mindset treatments at the end of the experiment. Given that growth relative to fixed mindsets are associated with more adaptive outcomes, in this debrief, we provided participants with strategies for enhancing their growth mindset. Mindset manipulation check. To evaluate the effectiveness of the intervention, we asked participants four growth-oriented items from the mindset of intelligence scale on a six-point Likert scale from 1 (Strongly Disagree) to 6 (Strongly Agree). This measure differs from the baseline mindset measure, which was assessed using fixed-oriented items. A sample item is ‘No matter how much intelligence you have, you can always change it quite a bit.’ We used a validated German translation from Rammstedt, Grüning, and Lechner (2022). Dependent variables. We used the same dependent variables as in Study I. However, as we aimed to measure the effect of the experimental treatment, we omitted past usage as a proxy for real behaviour. All Cronbach’s Alpha coefficients were above 0.70. Negative attitudes towards AI. As Dang and Liu (2022) found that general negative feelings toward robots are associated with mindsets, we adopted the Negative Attitudes Towards Robots (NATR) scale to focus on AI rather than robots. Participants answered 14 items on a seven-point Likert scale from 1 (Strongly Disagree) to 7 (Strongly Agree). Sample items include ‘I feel that if I depend on AI too much, something bad might happen.’ (reverse coded) and ‘I am concerned that an AI would be a bad influence on children.’ (reverse coded). Cronbach’s Alpha was 0.89. Knowledge about GenAI. We adopted one item about knowledge of GenAI from Clarke (2019) and Dang and Liu (2022). Participants responded to the statement ‘Please indicate the extent to which you have heard, read, or seen anything about generative AI’ on a seven-point scale from 1 (‘To a very small extent’) to 7 (‘To a very high extent’). Risk preferences. We asked participants to rate their risk preference by assessing the statement ‘How willing are you to take risks in general?’ on a seven-point scale from 1 (‘Not willing to take risks at all’) to 7 (‘Very willing to take risks’). Demographic and employment controls. We assessed our hypotheses with and without relevant control variables. Our main control variables were age, gender, education (ranging from ‘Did not complete compulsory school’ to ‘PhD’), and job level (ranging from ‘Supporting position’ to ‘Top management’). As education and job level are ordinal variables with more than five categories, we included it in our model as continuous variables. 4.2.5. Analytical approach. To test whether the mindset intervention influences attitudes, subjective norms and perceived behavioural control towards GenAI, we employed hierarchical regression analysis. This approach is appropriate for experimental designs with observed outcome variables, as it allows for the estimation of the direct effect of the treatment while sequentially accounting for relevant control variables. By entering predictors in steps, hierarchical regression enables an assessment of whether the intervention effects remain robust when demographic, socio-economic, and technology-related covariates are included. We analyzed the data using STATA, version 16.1. 4.3. Results. Study 2 examined whether a mindset intervention influenced attitudes, subjective norms, and perceived behavioural control towards GenAI (Hypotheses 5–7). We first report the manipulation check, followed by regression analyses testing the hypotheses and their robustness to the inclusion of control variables. The descriptive and correlational statistics of Study 2 are presented in Table 6. 4.3.1. Manipulation check. Pre-intervention, we did not find a significant difference in the baseline mindset measure between the fixed (M = 3.59, SD = 1.20, n = 182) and the growth mindset treatment (M = 3.72, SD = 1.23, n = 207); t = −1.05, p = 0.29. Post-intervention, the mindset treatment measure was significantly lower for the fixed (M = 2.88, SD = 1.24, n = 182) compared to the growth mindset treatment (M = 4.42, SD = 0.91, n = 207); t = −14.16, p < 0.001, indicating that the treatment was effective. 4.3.2. Hypothesis testing. First, we assessed the effect of the treatment on attitudes, subjective norms and perceived behavioural control in a simple linear regression. The results supported all three hypotheses: Growth relative to fixed mindsets were associated with more positive attitudes towards GenAI (ß = 0.28, p < 0.05), consistent with Hypothesis 5; with more favourable subjective norms (ß = 0.31, p < 0.01), supporting Hypothesis 6; and higher levels of perceived behavioural control (ß = 0.28, p < 0.05), as suggested in Hypothesis 7. Table 7 summarises the simple regression results. To ensure the robustness of the results, we built a hierarchical regression model introducing three different sets of control variables, detailed in Table 8. Model I included only the effect of the treatment on the dependent variables, which were all significant as described above. In Model II, we incorporated the demographic variables ‘Age’ and ‘Gender’. The effect of the treatment on the dependent variables remained significant. Additionally, age significantly negatively predicted attitudes and perceived behavioural control. Gender influenced all constructs, with men having more favourable attitudes, subjective norms and perceived behavioural control compared to women. Model III also assessed the effect of the socio-economic factors ‘Education’ and ‘Job level’. Again, all dependent variables remained significant, supporting Hypotheses 5–7 in the initial models. Socio-economic factors had no significant effect, except for higher levels of education being associated with increased perceived behavioural control. In Model IV, we introduced factors related to technology and risk preferences, specifically ‘Negative attitudes towards AI’, ‘GenAI knowledge’, and ‘Risk seeking’. While attitudes and subjective norms remained significant at the 5 percent level, consistent with Hypotheses 5 and 6, perceived behavioural control was not significant in Model IV, indicating that Hypothesis 7 was not robust to the inclusion of these controls. This suggests that the effect of the mindset intervention on perceived behavioural control should be interpreted with caution, as it appears sensitive to the inclusion of domain-specific factors such as prior knowledge and attitudes toward AI. However, the control variables were associated with the dependent variables, suggesting that having an overall positive outlook on AI, having prior knowledge about GenAI, and being willing to take risks are also positively associated with employees’ attitudes toward GenAI, perceived subjective norms, and perceived behavioural control. 4.4. Discussion. Study 2 investigates how growth mindsets can foster the TPB constructs. In line with our hypotheses, a growth compared to a fixed mindset intervention initially predicted attitudes towards GenAI, favourable subjective norms, as well as perceived behavioural control. Further, a stepwise hierarchical regression model highlighted the robustness of these results, but also demonstrated that risk preferences, perceived knowledge about GenAI, and general feelings towards smart technologies are associated with attitudes, subjective norms and perceived behavioural control. Notably, once these technology- and risk-related variables were included, the effect of the growth mindset intervention on perceived behavioural control was no longer significant. This suggests that perceptions of control over GenAI use may be shaped more strongly by domain-specific experience and beliefs than by more general motivational orientations such as mindsets. n = 389 5. General discussion. GenAI is expected to fundamentally reshape organisational work, making employee adoption critical for realising its potential. However, individuals are often resistant to change and to adopting new technologies, undermining innovation and efficiency in organisations. While a substantial body of research exists on the adoption of AI, research focused on employee adoption is limited. Additionally, most studies to date focus on ‘traditional AI’ rather than GenAI. To address these issues, we established a Theory of Planned Behavior (TPB) model to explain why some individuals form stronger intentions to adopt GenAI and report higher levels of usage than their peers (Study 1). We also tested a mindset intervention that could help organisations foster more favourable adoption-related beliefs regarding GenAI (Study 2). In doing so, our contribution lies in applying and integrating established theoretical frameworks – namely TPB and mindset theory – to the novel context of GenAI adoption in organisations. Study 1 explains variation in GenAI adoption intentions and self-reported usage using a TPB model, showing that attitudes, subjective norms, and perceived behavioural control influence adoption intention. The findings of Study 1 are also in line with past research drawing on the TPB, such that attitudes had the strongest relative impact on technology adoption intention, and that the influence of social norms on adoption intention was particularly strong when individuals had limited experience with such technology. As GenAI is a relatively novel development involving considerable uncertainty, individuals may rely particularly strongly on social cues when deciding to adopt new technology. Consistent with this reasoning, Study 1 shows that subjective norms play an important role in shaping employees’ intentions to use GenAI. Perceived behavioural control is often related variables. p < 0.001, p < 0.01, p < 0.05; Cronbach’s Alpha coefficients are reported in brackets where applicable; n = 365. assumed to counterbalance social influence within the TPB by reflecting individuals’ sense of agency and capability. Our findings suggest a more nuanced pattern. While perceived behavioural control contributed to adoption intentions, it did not directly predict self-reported GenAI usage. Instead, behavioural intention emerged as the strongest predictor of usage. This pattern indicates that, in the context of GenAI adoption, employees’ perceptions and social evaluations primarily translate into behaviour through intention rather than through direct enactment. These findings should be interpreted in light of characteristics specific to GenAI, such as high uncertainty regarding its capabilities and appropriate use, rapid development, and human-like outputs. These features are likely to amplify ambivalent attitudes, increase reliance on social cues, and heighten concerns about loss of control. In this sense, attitudes and subjective norms may play a particularly central role in shaping adoption-related responses to GenAI. Nevertheless, researchers should exercise caution in interpreting this intention-usage correlation, as it may be inflated by common method bias, as both intention and usage were measured via self-report in the present research. Overall, our Study 1 findings contribute to the literature on technology resistance by demonstrating how TPB constructs jointly shape employee responses to a highly novel and uncertain technology such as GenAI. Study 1 thereby offers a robust framework, grounded in the TPB, for examining variation in employees’ adoption of GenAI. Study 2 focused instead on a particular lever for GenAI adoption. Specifically, we conducted a controlled experiment to examine how promoting a growth relative to a fixed mindset affects the primary components of the TPB. Building on the Study 1 TPB model, we randomly allocated participants to a growth or fixed mindset treatment and observed that inducing a growth mindset fosters positive attitudes towards GenAI and favourable subjective norms, as well as strengthens perceived behavioural control. A stepwise hierarchical regression model demonstrated that most results were robust, even when adding multiple sets of control variables. This is important because other variables, such as age and gender, initially appeared to be strong predictors, but lost significance when adding other controls, such as general attitudes towards intelligent technologies, risk preferences, and prior knowledge. While these controls reduced the effect of the treatment on perceived behavioural control, the impact on attitudes and subjective norms remained significant, demonstrating that growth mindset interventions can promote such attitudes and norms that are important for supporting GenAI adoption. It is worth noting that, while the effect sizes observed in Study 2 may seem modest, the implications for real-world behaviour are nonetheless meaningful, given the simplicity with which mindset interventions can be implemented. Indeed, these findings bridge the mindsets and technology acceptance literatures by revealing that growth mindset interventions can influence social-cognitive constructs central to technology acceptance. The findings of Study 2 also lend support to earlier mindsets research that similarly underscores how growth mindsets enhance openness to challenges, help individuals reinterpret difficulty as an opportunity for learning, and foster a greater sense of efficacy and control. In sum, mindsets appear to be a promising lever through which GenAI, and potentially other innovative technologies, could be adopted in the workplace. 5.1. Practical implications. Successfully incorporating GenAI will be an important competitive edge for organisations. Yet, employees may resist such transformations, which poses a threat to organisations insofar as they might lag behind in the GenAI race. Organisational leaders thus need to understand why some employees are more or less open to GenAI, and how they can foster its usage among employees. Given the role of attitudes and subjective norms on adoption intention, as a first step, organisations could shape how GenAI is evaluated and discussed within teams. For example, leaders can be role models for the constructive use of GenAI, sharing success stories and implementing regular ‘GenAI use case sessions’ where employees exchange concrete applications and lessons learned. In addition, targeted training and support from the organisation can help enhance employees’ perceived behavioural control by providing clear user guides, access to GenAI tools, or technical assistance, as well as through tailored GenAI training programmes or seamless integration of GenAI tools in their existing IT infrastructure. As a next step, organisations could provide short workshops wherein employees engage with scientific testimonials promoting growth mindsets and apply ‘Saying is believing’ exercises to real work situations. This might involve an employee reflecting on how a growth mindset could benefit them as they master the challenges of GenAI integration in their work. Such initiatives could be reinforced through internal communications emphasising the value of learning, experimentation, and skill development. In practice, these interventions can be delivered as short, low-cost formats – for example, 1–2 h workshops or brief digital modules – that can be scaled across the organisation. Scaling such training, however, may only be effective insofar as organisations provide clear cues for the utility of GenAI in employees’ work, have the infrastructure to provide additional technology support, and can adapt the training to meet the various skill levels and needs of employees or business units. Finally, growth-oriented messaging around GenAI adoption may also be integrated into existing HR processes, such as onboarding, leadership development, or continuous learning programmes. Yet mindset interventions should be viewed as a complement to, rather than a substitute for, organisational measures that support employees’ effective use of GenAI, particularly when it comes to strengthening perceived control, such as clear AI governance policies and guidance on appropriate use. 5.2. Limitations and future research. The present research provides insight into why GenAI adoption can vary among employees, as well as how such adoption can be encouraged. Our research nonetheless has limitations that raise interesting questions for future research. First, both the correlational design of Study 1 and the experiment in Study 2 rely on cross-sectional data in a controlled setting. Replicating the current findings in a longitudinal field setting would thereby strengthen this line of research. Although previous research has demonstrated that controlled studies can be generalised to workplace behaviour and employing proxies for usage behaviour is a well-established practice in technology acceptance research, future research ought to also explore the influence of adoption intention on ‘real-life’ usage behaviour. Moreover, usage behaviour was self-reported in Study 1 and not measured in Study 2, which limits conclusions about actual GenAI use. Because predictors and outcomes in Study 1 were collected via self-report at a single point in time, common method variance may have inflated the observed relationships among constructs. Together, these design and measurement constraints suggest caution in interpreting the findings as reflecting actual behaviour. Longitudinal field research could also be employed to verify and extend the findings observed in Study 2. Specifically, do the effects of mindset interventions on fostering technology adoption in organisations endure over time, as previously observed in other domains, such as academic achievement, performance appraisals, and social interactions? Second, we found that growth mindsets are an impactful lever to foster GenAI adoption, but other background factors, likely also influence the TPB constructs and, in turn, GenAI adoption. Previous studies have examined various factors influencing individual reactions to traditional AI, including age, trust, and threat perception. In addition, organisational AI maturity may shape how employees perceive and use GenAI, as access to tools, governance structures, and prior exposure to AI technologies can vary substantially across organisations. For a more comprehensive understanding of GenAI adoption, future research could thereby examine how factors such as trust in AI, perceived threat, risk preference, and organisational AI maturity influence the relationships between attitudes, subjective norms, perceived behavioural control, and GenAI adoption. Finally, this study focused on white-collar employees in Germany, who are likely to be among the first to interact with GenAI in their daily work, given the increasing relevance of these technologies for knowledge work. GenAI is nonetheless expected to influence a much broader range of roles and fundamentally reshape how work is performed across sectors and job types. Future research might therefore examine whether similar mindset effects emerge in occupational settings beyond white-collar knowledge work, such as healthcare support roles (e.g. nursing or medical administration), manufacturing and technical trades (e.g. maintenance or quality control), or customer-facing service work (e.g. call centres or logistics coordination). Future studies could also explore whether the effectiveness of mindset interventions varies across occupations with different levels of autonomy, technological exposure, presence of routine tasks, as well as across cultural contexts. Moreover, as our findings are based on a German sample, cultural factors may influence attitudes toward AI and openness to technological change, potentially limiting generalizability to other national contexts. 6. Conclusion. The present research set out to explore why some employees are open to adopting GenAI while others remain resistant. Using the Theory of Planned Behavior, we established a model explaining variation in adoption intentions and self-reported usage levels of this rapidly evolving technology in organisational settings, highlighting the central role of attitudes, subjective norms and behavioural intentions. Building on these explanatory insights, we further show that adoption-related evaluations are not fixed: even when individuals are initially hesitant to adopt intelligent technologies, a growth mindset intervention can foster more positive attitudes and favourable subjective norms, and – under certain conditions – enhance perceived behavioural control toward GenAI. Taken together, the findings suggest that GenAI adoption is best understood by combining explanatory models of employee behaviour with interventions targeting underlying psychological mechanisms, with attitudes and social norms appearing more malleable than perceived control. This research highlights mindsets in particular as a promising lever for facilitating organisational efforts to foster GenAI adoption. CRediT: Jennifer Wieland: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Software, Validation, Visualization, Writing – original draft, Writing – review & editing; Lauren Keating: Conceptualization, Funding acquisition, Methodology, Resources, Supervision, Validation, Writing – review & editing; Alwine Mohnen: Conceptualization, Supervision, Writing – review & editing. Disclosure statement No potential conflict of interest was reported by the author(s). Funding This work was supported by Emlyon Business School.