Artificial intelligence (AI) for good? Enabling organizational change towards sustainability

Technology has advanced rapidly in recent years, altering the very ways in which businesses work and create value. In recent years, AI’s significant (enabling and inhibiting) influence on both society broadly and businesses in particular (Gupta et al. 2021; Haftor et al. 2021) has made the technology stand out as something with the potential to define the future development of the global economy (Nishant et al. 2020). The term "AI" may be defined as "the use of computational machinery to emulate capabilities inherent in humans, such as doing physical or mechanical tasks, thinking, and feeling" (Huang and Rust 2021). The history of AI can be traced back to the 1950s—the formative years of computer science —but at the start of the twenty-first century, enhanced computational power and greater data quality and accessibility have turned AI into a remarkable tool capable of real-time pattern recognition and learning (Wirtz and Müller 2019). AI today can reduce errors, take over human tasks, and accelerate research (Gupta et al. 2021; Thomas et al. 2024). The surge in AI-related publications over the last five years is indicative of this rapid progress (e.g., Burström et al. 2021; Kanbach et al. 2024; Mariani et al. 2022). This surge presents both challenges and opportunities when it comes to the application of AI across various industries (including critical ones, such as medical solutions and disaster management). Wirtz and Mueller (2019) revealed that AI applications in public management structures have effectively become essential to the exercise of state power and public influence. An AI-driven medical drone overcame critical challenges associated with delivering critical medical supplies during the COVID-19 pandemic and, beyond that crisis, to remote regions (Damoah et al. 2021; Lamptey and Serwaa 2020). In marketing, AI-driven tools like SNAzzy (Social Network Analysis in Telecom) and VOCA (Voice-of-Customer-Analytics) have bolstered the field of customer analytics (Akerkar 2019). Prior studies have noted AI’s influence across multiple sectors as well as its potential benefits in others (e.g., Larsson et al. 2019; Rantala et al. 2023), highlighting the importance of studying AI beyond the confines of specific industries (Gupta et al. 2021).

Although constraints on AI applications remain a topic worthy of investigation, multiple prior studies have highlighted the advantages that AI may offer moving forward in the domain of corporate sustainability (Bickley et al. 2024; Kar et al. 2022). With its power of predictability, AI can be used to identify and predict supply and demand, minimizing production waste, reducing inventory-management overhead costs, and lowering carbon emissions in production processes and throughout supply chains (Arinez et al. 2020; Singh et al. 2022). Moreover, research has shown that AI could enable organizations to achieve up to 79% of the Sustainable Development Goals (SDGs) (Vinuesa et al. 2020).

Over the last decade, the term “corporate sustainability” has become a significant point of focus for organizational leaders, but the challenges associated with fully integrating sustainability-related factors into organizations still constitute a daunting and complex task (Baumgartner and Rauter 2017; Engert et al. 2016). Corporate sustainability refers to the approach that organizations take to ensure they fulfill their responsibilities to both current and future stakeholders (Dyllick and Hockerts 2002). The concept of corporate sustainability is inherently complex, multidimensional, and subject to rapid change. As a result, it represents a significant challenge for organizations to maintain a high level of corporate sustainability over time (Baumgartner and Ebner 2010; Nguyen and Kanbach 2023). To transition into more sustainable businesses, organizations require a wide range of enabling factors—both internal and external elements ranging from organizational capabilities and infrastructure to business partnerships and local engagements (Annunziata et al. 2018; Dangelico et al. 2017; Fernández-Torres et al. 2024). Multiple prior studies have emphasized technology as a critical component of any drive toward corporate sustainability (Baumgartner and Ebner 2010; Xavier et al. 2020). The potential benefits of AI for corporate sustainability have also garnered significant attention from researchers (e.g., van Wynsberghe 2021; Vinuesa et al. 2020).

Nonetheless, integrating AI into corporate sustainability efforts comes with several challenges, including uncertainty, issues, and barriers (Kar et al. 2022). In the field of social sustainability, for example, data is often qualitative, leading to concerns from both employees and customers about bias, fairness, transparency, and accountability (Köchling et al. 2024; Larsson et al. 2019; López-Torres et al. 2019). On the one hand, businesses must get their customers and employees to be willing participants. On the other hand, businesses face numerous technological challenges associated with AI, such as the need for big data resources, insufficient employee capabilities, high energy demand, and high carbon emissions—all obstacles that must be overcome before AI’s benefits can become apparent (Nishant et al. 2020; van Wynsberghe 2021). Furthermore, some have raised concerns regarding the inflation and ambiguity of AI’s benefits for sustainability, noting that the technology could ultimately do more harm than help (e.g., Heilinger et al. 2024). Manifold challenges, such as the need for an operating license, the need to gain customer acceptance, and the need for employees with sufficient skills and knowledge, tend to follow similar patterns whenever new disruptive elements, including sustainability initiatives, are introduced into an industry, driving the costs of AI adoption upward (Hahn et al. 2010).

Tornatzky et al.’s (1983) Technology-Organization-Environment (TOE) framework is a conceptual model in information systems that explains how various factors influence the adoption and use of new technologies (e.g., AI) in organizations (Hradecky et al. 2022; Schwaeke et al. 2024). This model considers the technology's attributes, the organizational context of its implementation, and the external environment in which the organization functions. The TOE framework comprises the three titular elements: technology, organization, and environment. The technology component examines the characteristics of the technology, including its functionality, complexity, compatibility with existing systems, and user-friendliness (El-Haddadeh 2020). The organization component addresses the internal context of the technology’s implementation, including the organization’s size, structure, culture, and available resources. The environment component conversely addresses the external context, encompassing the market conditions, regulatory requirements, and sociocultural norms that the organization is facing. One of the TOE framework’s key strengths is its comprehensive approach to technology adoption and implementation. Instead of concentrating solely on the technology or its organizational context, it more comprehensively recognizes the importance of both internal and external factors in shaping technology adoption and utilization (Baker 2012). However, this framework also has some limitations. For instance, it may fail to fully capture the intricacies of technology adoption and implementation, especially in rapidly changing environments in which external factors can significantly impact technology-related decisions. Additionally, it lacks emphasis on individual and behavioral factors (Oliveira and Martins 2011).

Nonetheless, we employed the TOE framework in this study to conduct a structured analysis and identify potential areas for future research (El-Haddadeh 2020; Tornatzky and Fleischer 1990). Studying the adoption of digital innovation among organizations using a theoretical foundation requires careful consideration of the factors behind the adoption process. Existing research has confirmed that the TOE framework is suitable for analyzing the dynamics underlying organizational adoption (El-Haddadeh 2020), offering scholars and decision-makers a more comprehensive overview of the factors that drive AI adoption in corporate sustainability efforts (Khan et al. 2021). Other frameworks like the technology acceptance model and the unified theory of acceptance and use of technology focus on individual acceptance decisions, but the intricacy and broad scope of organizational technologies require a more comprehensive framework. The TOE framework has been applied in various ways in previous studies, having been used to investigate factors like cost-setting, security concerns, and management support as well as to assess the integration of organizational factors with theories of innovation diffusion and technology acceptance. These studies have contributed to our current understanding of the factors behind AI adoption in corporate sustainability efforts (Lee and Lee 2014). The TOE framework’s ability to structure and integrate different clusters facilitates a comprehensive analysis of technologies’ adaptability within an organization. Therefore, we contend that examining the influential factors within the three clusters—technology, organization, and environment—will provide scholars and practitioners with a comprehensive understanding of the current state of the literature on the application of AI to corporate sustainability.

To ensure the thoroughness of this qualitative research, we abided by the Gioia methodology of inductive concept development (Magnani and Gioia 2023), which enables researchers to empirically examine and document phenomena (Gioia et al. 2013; Yin 2014)—in this case, AI adoption in corporate sustainability efforts. This method has been used in and proven to be valuable in previous strategic studies on AI integration (Jorzik et al. 2024). Thus, this qualitative approach is particularly appropriate for this study, as it can extend the existing theory to study how organizations employ AI technology to fulfill their sustainability-related responsibilities and challenges. To comprehensively identify relevant data for our research and offer a cohesive perspective on the application of AI in research on corporate sustainability, we employed thematic analysis, following a pattern-inducing technique to generate concepts from the data (Magnani and Gioia 2023). Initially, we conducted a thorough review in which we identified commonalities among interviews to organize emerging topics into first-order concepts according to Gioia et al.’s (2013) guidelines. These concepts were then aggregated into wider dimensions to reflect the data’s overarching themes.

We employed purposive sampling—a strategy that is particularly well-suited for exploratory studies aimed at gathering rich, context-specific insights—to capture a broad range of perspectives on and facilitate the development of AI adoption theory (Patton 2014). Given our focus on how AI facilitates corporate sustainability, our sample selection covered a range of organizational sizes (from those with fewer than 10 to those with over 50,000 employees) and maturity levels (from start-ups to those with over 100 years of operational experience) to achieve a comprehensive understanding of AI use in organizational sustainability efforts. We first constructed a comprehensive list of corporations with a demonstrable interest in AI by consulting various sources, including company websites, press releases, and industry reports. To address the research question fully, we considered firms from a wide range of industries and countries, thereby capturing an equally wide range of approaches to AI-driven corporate sustainability. We then identified and approached managers, senior managers, and executives within the considered firms who were responsible for and directly involved in AI-related sustainability initiatives (see Table 1 for a summary of the interviewees). In-depth exploratory semi-structured interviews were conducted with each of these officials, centered on four key themes: use cases of AI (including applied tools); approach to AI implementation; AI's contribution to corporate sustainability; and personal reflections on likely future trends of the discussed AI applications. This structure enabled us to gather respondents' insights regarding entrepreneurs’ adoption of AI to manage uncertainty. To expand the scope of our data, we employed the snowball sampling technique, whereby interviewees were asked to recommend or invite additional participants who meet the requisite criteria and could contribute valuable information (Biernacki and Waldorf 1981; Bell and Bryman 2007). This approach is consistent with this study’s exploratory nature, which was designed to generate novel insights that contribute to theoretical development rather than to achieve a representative sample (Eisenhardt 1989). A total of 24 interviews were conducted and recorded between November 2023 and August 2024, with data collection continuing until theoretical saturation—the point at which no new theoretical insights were gained from additional interviews (Corbin and Strauss 1990)—was achieved. Each interview was conducted in line with a structured guide that enabled the interviewers to respond promptly to participants' remarks (Eisenhardt 2007; Guest et al. 2006). The interviews were conducted via digital videoconferencing tools, namely Zoom and MS Teams, and lasted an average of 30 min.

The Gioia method (Gioia et al. 2013) was instrumental in structuring our coding process, enabling a systematic progression from raw data to refined theoretical constructs. As outlined by Corbin and Strauss (1990), we employed open coding, axial coding, and theoretical coding to distill the data into a cohesive structure. During open coding, we assigned descriptive labels to notable data segments (i.e., interview transcripts) in order to capture the main idea of each piece of information. The interviews were transcribed verbatim, and distinct linguistic and phonetic features were excluded to allow for a sole focus on the substance. Our investigation involved multiple iterations of cumulative coding. Each researcher independently transcribed and coded the data. To enhance the reliability and validity of our findings, the research team held regular discussions on coding and data analysis. We iteratively analyzed the data until recurring patterns emerged and were confirmed through feedback loops. Axial coding facilitated the identification of underlying structures, enabling us to reduce the number of codes and focus on core themes. Finally, theoretical coding allowed for the emergence of higher-level theoretical constructs, giving way to a robust data structure progressing from first-order concepts to second-order themes and aggregate dimensions, as depicted in Fig. 1. We used MAXQDA to obtain a comprehensive overview of the theoretical relationships between the codes, and we used the Gioia method (Gioia et al. 2013) to infer patterns inductively. This iterative comparison strengthened our findings’ internal validity, enabling us to refine our theoretical framework based on real-world insights, as recommended by Eisenhardt (1989). This step-by-step interaction between theoretical expectations and empirical observations not only anchored our findings in established theories but also contributed novel insights into how AI enables corporate sustainability.

Full size image

When examining the application of AI with a focus on environmental sustainability, we uncovered evidence to distinguish between its direct and indirect impacts on sustainability efforts. Direct impacts are the immediate and observable effects of AI on environmental sustainability. Applications of AI reduce environmental harm and optimize resource usage. Moreover, they enhance energy efficiency, manage smart grids, and integrate renewables. AI advances climate monitoring by improving forecasts and prediction patterns. In waste management, AI can improve sorting and recycling. It also aids pollution control through real-time monitoring and prediction. Additionally, AI supports biodiversity conservation by tracking wildlife and detecting poaching using cameras or satellite data. Indirect impacts include the less visible ways in which AI contributes to environmental sustainability. Such impacts are often realized through changes in systems, behaviors, or long-term outcomes. For example, AI enhances decision-making by offering advanced analytics and simulations that improve environmental planning and policymaking. By providing insights into the potential impacts of various strategies, AI aids in the development of more effective and sustainable environmental policies. Additionally, AI-driven applications can influence consumer behaviors toward more sustainable choices; recommendation systems on e-commerce platforms can promote products with lower environmental footprints, while apps can encourage individuals to reduce their energy consumption through personalized tips and feedback. AI also fosters innovation and research in environmental technologies by accelerating the development of new materials that can reduce carbon emissions or optimize sustainable agriculture processes. These innovations often lead to cascading effects on sustainability. Supply chain optimization is another area to which AI contributes indirectly. By predicting demand and optimizing logistics, AI enhances supply chain efficiency, which translates to reduced emissions from transportation and manufacturing. Lastly, AI solutions can significantly enhance transparency, boosting compliance with sustainability regulations (e.g., European Sustainability Reporting Standards) by leveraging advanced data-analysis and -reporting capabilities while reducing the manual effort required for compliance documentation, minimizing the risk of errors, and ensuring that reports align with regulatory standards. By improving data accuracy and report-generation efficiency, AI helps organizations to more effectively meet regulatory requirements while enhancing overall operational efficiency.

The success of AI-driven sustainability initiatives hinges not only on innovative algorithms and data insights but also on robust and efficient IT infrastructure. As organizations integrate AI into their sustainability strategies, they must address the unique challenges and demands associated with AI-based technologies.

Drawing on the insights gathered from the interviews in the previous section and concepts from Stouten et al. (2018) and Tornatzky and Fleischer (1990), we have developed a framework with which to leverage AI to enhance organizational sustainability. This framework explores the central themes of AI integration in corporate sustainability and proposes a structured approach to overcoming barriers identified by prior research.

Figure 2 visualizes this framework. Using the TOE framework as its foundation, it considers external, technological, and organizational factors behind successful implementation. The environmental aspect of the TOE framework influences the identification of potential AI use cases in corporate sustainability efforts. In this context, the environmental aspect (which consists of government regulations and market structure and characteristics) has been expanded to include external stakeholder management (Tornatzky and Fleischer 1990). With regard to sustainability, many companies have underrated the value of public opinion (Parguel et al. 2011). In the capabilities step, an organization assesses its operational enablement and technical capacity, addressing deficiencies as needed. Accordingly, the organizational component of the TOE framework (comprising organizational characteristics like size, industry, and available resources) has been augmented by the inclusion of leadership and management support. The results of our data analysis have demonstrated that this is of great consequence when it comes to the adoption of AI for the purpose of driving corporate sustainability. The technology aspect of the TOE framework has been augmented through the incorporation of second-order themes (namely IT infrastructure, data structure, and system integration). It is incumbent upon organizations to address the challenges and demands that AI-driven technology places upon them. To maintain a robust and efficient IT infrastructure, organizations must ensure the cleanliness of the data structure and processes within software applications as well as the seamless exchange of data across the company-wide IT landscape through fully integrated systems. The framework broadens the existing literature by incorporating impact measurement as a dimension behind technology adoption. Moreover, it expands the strategic steps, as proposed by Stouten et al. (2018), available to organizations looking to adopt AI to improve their sustainability. Organizations start by identifying the opportunities, risks, and regulatory requirements associated with aligning AI with their sustainability goals. Key capabilities include the prioritization of sustainability, cross-functional collaboration, and stakeholder management for upskilling and partnerships. The next phase involves developing infrastructure for AI adoption, focusing on risk management, and fostering an innovative culture. An additional step that this study showcased was impact measurement, which includes monitoring progress, achieving short-term wins, and fostering resilience. Overall, the framework stresses the need for responsible AI practices to ensure that the technology has positive impacts on the climate, in line with evolving environmental and regulatory demands. By integrating TOE principles, this framework guides organizations in deploying AI strategically to bolster their long-term sustainability.

Full size image

This study’s findings validate and extend the established literature, its theoretical enhancements summarized in Table 2. The insights derived from this research offer a comprehensive understanding of organizational capabilities, leadership, technical infrastructure, and performance measurement.

Integrating AI into sustainability efforts necessitates the fostering of an operational culture that embraces innovation and technological advancements. Managers must promote a mindset shift that compels employees to perceive AI not merely as a tool for raising efficiency but as a critical enabler of sustainable practices. Encouraging a culture of ongoing learning and adaptability ensures that the workforce is prepared to leverage AI solutions in pursuit of long-term sustainability goals. Leadership plays a pivotal role in aligning AI initiatives with sustainability objectives. Managers must develop clear, forward-thinking strategies that integrate AI into the organization’s broader environmental, social, and governance (ESG) goals. Effective leadership should not only champion AI adoption but also guide the organization through change-management processes, ensuring that AI-driven sustainability becomes a core element of the company’s strategic direction. AI-driven sustainability initiatives require collaboration across multiple stakeholders (both internally and externally) through strategic partnerships. Managers must engage stakeholders in a transparent manner to showcase how AI contributes to their organization’s sustainability goals. Building trust through clear communication and collaboration helps to ensure that stakeholders are on board with AI implementation and, in turn, to facilitate a smooth adoption.

For AI to be relevant in sustainability management, its impact must be measurable. Managers should implement systems to track and assess how AI technologies contribute to sustainability outcomes (e.g., energy efficiency, waste reduction, carbon footprint). Data-driven impact measurement empowers managers to make informed decisions, optimize AI applications, and highlight tangible improvements to stakeholders. Notably, the success of any AI-driven sustainability initiative hinges on the organization’s technical capacity. Managers must ensure that their teams possess the necessary technical expertise to both implement and maintain AI solutions. Investing in infrastructure, upskilling employees, and fostering collaboration between data scientists and sustainability experts are crucial when it comes to building the technical capacity necessary to fully harness AI in the promotion of sustainability. By focusing on the five aforementioned areas, managers can strategically align AI initiatives with sustainability goals, driving innovation, operational efficiency, and long-term environmental impact. Additionally, overcoming a lack of expertise through strategic partnerships and upskilling, addressing data challenges through collaborative data ecosystems, mitigating high initial costs with scalable pilots, building trust through transparent AI models, leveraging AI for circular economy-related practices, incorporating AI into carbon footprint-monitoring and -reduction efforts, and promoting cross-functional AI adoption in decision-making are practical strategies that may be used to bolster the managerial relevance of AI in sustainability. Real-world examples, such as Walmart's AI-driven energy-management systems (Goswami et al., 2024), Tomra's AI-driven waste-management sensors (Jogarao et al. 2024), and PepsiCo's integration of AI throughout its supply chain (Goswami et al., 2022), illustrate actionable steps companies can take to effectively harness AI for sustainability, providing a roadmap for managers to address specific challenges in AI adoption while driving positive sustainability outcomes.

Three areas for further research are particularly compelling and warrant further investigation. First, one of the primary concerns associated with the rising utilization of AI in decision-making processes is the potential for a loss of human emotional capabilities, including empathy, humanity, and conscientiousness. Management decisions pertaining to sustainability frequently require managers to take these human qualities into account in order to understand the broader social, ethical, and emotional implications of any considered actions. As posited by Cullen‐Knox et al. (2017), human emotions exert a pivotal influence on corporate decision-making processes and the subsequent evaluations. The lack of such emotions in AI could lead to decisions that are solely data-driven and fail to consider human or moral aspects that are generally non-quantifiable. Future research should investigate the potential negative impact of this lack on the corporate decision-making process and, in turn, on sustainability outcomes. Second, as AI technologies have evolved, their energy consumption and carbon footprint have become subjects of significant scrutiny. Van Wynsberghe (2021) drew attention to AI’s considerable environmental impact, particularly when it comes to the energy consumption necessitated by the training and operation of complex models. It would be beneficial for future studies to investigate the long-term environmental effects of AI’s widespread adoption in the field of sustainability and to ascertain whether its benefits outweigh its environmental costs. These trade-offs must be understood if strategies are to be developed that ensure the maximization of its positive impacts and the minimization of its negative impacts. Third, a company’s ownership structure (e.g., private, public, cooperative) can significantly influence the priorities and strategies that organizations adopt when integrating AI into their sustainability efforts. Thus, future research should ascertain the impact of such structural variance on the integration of AI in sustainability efforts in order to determine which ownership types are most conducive to ethical and effective AI use. Such an understanding would enable policymakers and business leaders to establish conditions that facilitate optimal (at least with regard to sustainability) AI utilization.

This study is subject to two principal limitations. First, although our reporting has exclusively focused on capabilities that have received the highest amount of research attention, it remains plausible that these capabilities may not necessarily possess the strongest correlation with performance outcomes. To ascertain which capabilities demonstrate the most robust relationships with performance, future research endeavors (e.g., meta-analyses) should be dedicated to each category of capability. Second, in the interviews, participants may have provided responses that they perceived as socially desirable to the interviewer; moreover, discrepancies in how participants comprehended questions or articulated their responses could have introduced variability into the data. These inconsistencies may not be directly tied to the research question, which can complicate efforts to derive accurate conclusions from the results. This limitation highlights the importance of careful design and interpretation for reducing bias and improving data quality.