Human-AI agency in the age of generative AI

Source: Sebastian Maximillian Krakowski; Information and Organization (open access) · 2025

The rapid emergence of generative artificial intelligence (GenAI) is profoundly transforming the nature of work and organizations, challenging prevalent views of AI as primarily enabling prediction and optimization. This paper argues that GenAI represents a qualitative shift that necessitates a fundamental reassessment of AI’s role in management and organizations. By identifying and analyzing four critical dimensions – (i) GenAI’s broad applicability as a general-purpose technology; (ii) its ability to catalyze exploratory and combinatorial innovation; (iii) its capacity to enhance cognitive diversity and decision-making; and (iv) its democratizing effect on AI adoption and value creation – the paper highlights GenAI’s potential to augment and scale human creativity, learning, and innovation. Building on insights from the AI and management literature, as well as on theory of human-AI agency, the paper develops a novel perspective that challenges the dominant efficiency-oriented narrative. It proposes that a human-complementary approach to GenAI development and implementation, leveraging it as a generative catalyst for exploration, can enable radically increased creativity, innovation, and growth. GenAI’s democratizing aspects can amplify these mechanisms, promoting widely shared growth when combined with appropriate policy and managerial choices. Implications for theory, practice, and future research directions are discussed, drawing attention to the need for approaches in GenAI development and deployment that are complementary rather than competitive to human beings. The paper concludes by discussing the theoretical, practical, and policy implications of this transformative technology. It outlines future research directions, emphasizing the critical role of human agency in determining the organizational, societal, and ethical outcomes associated with AI adoption and implementation.

The rapid advancement of artificial intelligence (AI) is profoundly transforming organizations and the nature of work in ways that can be likened to the impact of electrification on factories during the industrial revolution (Agrawal, Gans, & Goldfarb, 2022). Just as electricity empowered people to work more independently and take on more autonomous, value-creating initiatives, AI is expected to enable innovation on a distributed scale by giving employees access to information-processing and reasoning capabilities. Research on AI in management has explored its impact on organizations, focusing on AI adoption approaches and the division of labor between human and AI agents (Choudhary, Marchetti, Shrestha, & Puranam, 2025; Raisch & Krakowski, 2021; Shrestha, Ben-Menahem, & von Krogh, 2019). The emergence of generative AI (GenAI) has however marked a significant leap beyond the capabilities of predictive AI, which primarily enabled automation and analytical decision support, with early findings indicating GenAI performance reaching or even exceeding human performance in tasks involving divergent thinking (Hubert, Awa, & Zabelina, 2024), social intelligence (Sufyan, Fadhel, Alkhathami, & Mukhadi, 2024), and empathic communication (Yin, Jia, & Wakslak, 2024), among others.

Despite these advancements in capabilities, as well as indications that GenAI is being adopted at rates comparable to (or even exceeding) prior emerging transformational technologies like the PC, the internet, and prior AI iterations (Bick, Blandin, & Deming, 2024; Humlum & Vestergaard, 2024; Wu et al., 2023), the dominant discourse in AI and management research still primarily frames AI as a tool for uncertainty mitigation and operational efficiency (Constantinides, Monteiro, & Mathiassen, 2024; Jarrahi, 2018). GenAI challenges this prevalent view, as its value and impact largely relate to its generative nature, widespread accessibility, and associated capacity to generate novel and creative output (Adam, 2023; Davenport & Mittal, 2022; Epstein et al., 2023). Current theory and assumptions relating to human-AI collaboration are based on the capabilities of predictive AI and do not adequately address how humans can be expected to interact and work with GenAI, representing a fundamental and qualitative shift that implies a reassessment of AI’s potential and desired impact in organizations and society.

This paper uses innovation search as an analytical lens to develop four key propositions that suggest a foundational reassessment of AI’s role in management and organizations: (i) GenAI’s broad applicability as well as its generalizable reasoning and learning capabilities distinguish it from predictive AI across both divergent and convergent applications in tasks, effectively contributing to making AI a general-purpose technology (GPT); (ii) building on its capabilities and wide applicability, GenAI’s generative capabilities enable a shift from exploitative to exploratory uses in both problem definition and solution search stages, catalyzing innovation through recombining concepts and ideas; (iii) as such, GenAI can act as a complementary artifact that augments human learning, creativity, and innovation throughout the problem-solving process; and (iv) GenAI’s affordability, accessibility, and conversational interfaces democratize AI adoption across organizational tasks and contexts, amplifying and compounding the speed and impact of these propositions.

By identifying and analyzing these dimensions, this paper makes three main contributions. First, it challenges the dominant efficiency- and convergence-oriented narrative in AI management literature by demonstrating how GenAI enables both divergent and convergent applications in problem solving. Second, it advances theory on human-AI hybrid agency by examining how GenAI’s distinctive capabilities affect task allocation and interdependencies between human and artificial agents in the problem definition and solution search stages of problem solving. Third, adopting a task-based perspective of the impact of technology on the economy, labor, and organizations, it provides insights for policy and practice. The paper develops these arguments through theoretical background, analyzing the main dimensions and discussing implications, whereafter it concludes with directions for future research on GenAI’s impact on humans, organizations, and society.

GenAI can offer distributed and ubiquitous capabilities in information processing and reasoning to enable enhanced and distributed problem solving (Boussioux et al., 2024; Feuerriegel, Hartmann, Janiesch, & Zschech, 2024). While emerging discussions of GenAI’s potential societal and organizational implications draw comparisons to the rapid and transformative adoption of electrification, personal computers, and predictive AI (Acemoglu, Autor, & Johnson, 2023; Agrawal, Gans, & Goldfarb, 2023a; Bick et al., 2024), there is limited research exploring AI’s creative dimension and its implications for organizations (Amabile, 2020; see Jia, Luo, Fang, & Liao, 2024 for an exception). Building on the preceding section’s innovation search framework, this paper explores how GenAI’s emergence impacts organizations and management along four key and progressively building dimensions: i) its nature as a GPT, which enables it to function as ii) a catalyst for exploratory innovation, allowing it to serve as iii) a complementary cognitive artifact, ultimately acting as iv) a democratizing force in value creation that compounds and amplifies the expected impact of the preceding dimensions.

GenAI is emerging as a transformative GPT – a highly impactful innovation like the steam engine, electricity, and computers that is pervasive, continually improving, and innovation spawning (Bresnahan & Trajtenberg, 1995), marking a significant leap beyond predictive AI’s capabilities. While predictive AI can drive innovation across industries (Brynjolfsson, Rock, & Syverson, 2017; Goldfarb, Taska, & Teodoridis, 2023; Trajtenberg, 2019), GenAI offers broader applicability with more generalizable reasoning and learning capabilities (Bick et al., 2024). Unlike the comparatively task-specific nature of predictive AI, which is typically trained and used for optimization or classification tasks in the convergent stages of problem solving, GenAI exhibits adaptability across a wide range of domains (Berg et al., 2023; Bommasani et al., 2022). This suggests that GenAI potentially surpasses predictive AI by enabling novel transversal applications, effectively constituting a GPT (Eloundou, Manning, Mishkin, & Rock, 2024). For example, frontier models like GPT-4 can generate coherent text, code, and images, illustrating their adaptability across different industries and functions (Berg et al., 2023; Epstein et al., 2023).

Building on GenAI’s general-purpose nature, its generative capabilities enable a significant shift from exploitative to exploratory applications, allowing the development of qualitatively new ideas and solutions (Kulkarni et al., 2024; Raisch & Fomina, 2025). Through “generativity,” defined as “a sociotechnical system where social and technical elements interact to facilitate combinatorial innovation” (Thomas & Tee, 2022, p. 255; Yoo, Boland Jr., Lyytinen, & Majchrzak, 2012), GenAI facilitates the large-scale recombination of ideas and concepts, potentially driving combinatorial innovation.

By leveraging vast volumes of training data and processing new and multimodal data, GenAI can act as a catalyst for innovation by identifying potential combinations of existing ideas or technologies that might not be immediately apparent to human decision makers. This generativity is enabled by GenAI’s quasi-random sampling mechanism, which enables “aleatory creativity” under constraints (Rietzschel & Rus, 2022). The probability distribution of the next iterative output then acts as a creative constraint, such as sequentially sampled tokens in the case of LLMs. Consequently, GenAI can expand the scope of exploratory search through combinatorial and serendipitous innovation (Busch, 2024; Mariani & Dwivedi, 2024; Robertson, Ferreira, Botha, & Oosthuizen, 2024).

Building on Krakauer’s (2016) distinction between competitive and complementary cognitive artifacts, GenAI represents a qualitative shift in how technology, particularly AI, can affect and potentially enhance human cognition. This specifically points to augmenting human learning, cognitive representations, and creativity through its generative capabilities. Competitive cognitive artifacts are tools such that “when we are deprived of their use, we are no better than when we started” (e.g., a calculator that provides temporary arithmetic augmentation and yields dependency rather than lasting cognitive benefits), while complementary cognitive artifacts enable learning through sustained skill improvements that persist even when the artifact is no longer available, since “after repeated practice and training, the artifact itself could be set aside and its mental simulacrum deployed in its place” (e.g., an abacus, whose user develops arithmetic capabilities that eventually render the tool redundant) (Krakauer, 2016).

Applied to AI, this conceptualization distinguishes between AI that merely augments performance without promoting learning (competitive artifact) and AI that promotes enduring skill development (complementary artifact). While predictive AI tends to function as a competitive artifact by optimizing tasks without engaging humans in metacognition or learning, GenAI’s interactive and generative nature enables it to be a more impactful complementary artifact. This stems from two key mechanisms: first, GenAI’s vast training data, which exceeds the distribution of individual human experience by orders of magnitude, enables exposure to novel knowledge and combinations thereof; second, its natural language interface and generative responses enable learning through dialogue and exploration that can persist beyond its immediate use.

GenAI’s effectiveness as a complementary cognitive artifact will depend on active human engagement and reflection. Its black-box nature and the inherent opacity of DL, along with ongoing modifications by system designers, can discourage deeper human learning if used passively (van den Broek, Sergeeva, & Huysman, 2021). Further concerns arise from human-computer interaction harms and loss of human autonomy and agency, such as anthropomorphization leading to misplaced trust and excessive dependency (Ray, 2023; Weidinger et al., 2022). Intentional design of human-AI interaction patterns can however promote complementary learning effects that reconfigure expertise and workflows (Barrett, Oborn, Orlikowski, & Yates, 2012), rather than GenAI becoming a competitive artifact characterized by overreliance and skill degradation (Chen & Chan, 2024; Lindebaum & Fleming, 2024).

Overreliance on GenAI could lead to a decline in essential skill development, particularly among junior professionals who may not receive adequate experiential learning opportunities (Beane & Anthony, 2024; Brynjolfsson, Li, & Raymond, 2023). Recent research has however found a performance compression effect where inexperienced or lower-performing workers appear to benefit more from GenAI assistance compared to senior or high-performing individuals (Boussioux et al., 2024; Brynjolfsson et al., 2023), though GenAI might negatively impact the development of new expertise and tacit knowledge critical for innovation. Further, the effectiveness of hybrid agency may be hard to measure (Weidinger et al., 2022), and is likely to depend on matching human-AI interactions to expertise levels, attitudes, and task characteristics, including temporal dynamics (Burton, Stein, & Jensen, 2020; Raisch & Fomina, 2025).

GenAI’s accessible, conversational user interfaces and broad applicability across tasks and domains can democratize AI adoption, enabling a wider range of individuals and organizations to harness its potential for value creation and innovation. This availability and engagement of more human agents at scale can in turn be expected to amplify and compound the suggested impact and ramifications associated in the preceding dimensions.

Together, these four dimensions illustrate how GenAI extends and redefines AI’s role in management and organizations, with Table 1 providing an overview of key takeaways. By acting as a GPT, catalyzing exploratory innovation, serving as a complementary cognitive artifact, and democratizing AI adoption, GenAI challenges AI’s dominant convergent-oriented narrative. As organizations increasingly integrate GenAI, new frameworks and strategies that account for its unique characteristics will likely be needed.

Viewed through the lens of punctuated equilibrium, where periods of stability are disrupted by rapid change (Hanelt et al., 2021; Romanelli & Tushman, 1994), GenAI represents a transformative technological development that challenges existing organizational equilibria. The emerging dynamics between human and AI agents create tensions in balancing automation and augmentation (Agrawal, Gans, & Goldfarb, 2023b; Raisch & Krakowski, 2021). This will likely require a redefinition of roles, workflows, and approaches to human-AI agency (Murray et al., 2021; Raisch & Fomina, 2025). Adjusting to this punctuated equilibrium calls for new approaches to assess and realize GenAI’s potential, while mitigating risks and unintended consequences (Brynjolfsson, 2022; Eloundou et al., 2024).

Organizations will need to assess how to integrate GenAI into their workflows to complement rather than replace human capabilities. Unlike previous waves of AI adoption focused on convergent logic, GenAI’s broad applicability requires a more nuanced implementation approach. From a task-based perspective, organizations must analyze which tasks suit automation versus augmentation through human-AI collaboration and understand how these dynamics evolve (Raisch & Fomina, 2025; Raisch & Krakowski, 2021), suggesting that traditional conceptions attributing innovation solely to human agency will require revision. GenAI democratizes knowledge retrieval and recombination, expanding innovation participation by mitigating expertise bias through its emphasis on feasibility over novelty (Eapen et al., 2023), implying a potential for radically more distributed, autonomous, and pervasive value creation in the workforce. For example, GenAI enables tackling complex architectural challenges in software development (Nguyen-Duc et al., 2023), while it transforms ideation through rapid iteration in creative industries (Eapen et al., 2023).

Effective organizational implementation will require attention to task interdependencies and investment in complementary assets including employee training, culture, and infrastructure (Kemp, 2024). Organizations will have to develop new approaches to performance measurement and incentives that align with changing work patterns and skill development (Tamayo et al., 2023). This includes enabling workers to distinguish between convergent and divergent tasks, choosing between predictive or generative AI approaches, and managing actual and perceived relative advantages in abilities between human and AI agents (Dell’Acqua et al., 2023; SimanTov-Nachlieli, 2025). Necessary skills will likely include data management, AI literacy, and AI deployment decisions (Raisch & Krakowski, 2021; Verganti et al., 2020), leveraging GenAI as a complementary artifact for learning at scale, yet at an individual worker level. Understanding contingencies in employee expertise, personality, and AI attitudes (Chen & Chan, 2024; Park & Woo, 2022), along with contextual factors like decision context and time dynamics (Burton et al., 2020; Raisch & Fomina, 2025), will likely be decisive for successful integration.

Widespread GenAI adoption may entail several challenges for organizations. Overreliance on GenAI could lead to passivity and deskilling if employees become too dependent on AI-generated output without maintaining their own expertise and critical thinking skills. To mitigate this, approaches to managing knowledge and expertise may need to balance GenAI use with traditional mentoring and deliberate practice. They will have to ensure that employees continue to develop and retain essential skills, particularly in areas where human judgment and creativity are essential (Beane & Anthony, 2024; Dell’Acqua et al., 2023). This includes developing practices that integrate both algorithmic and traditional approaches into learning and development, maintaining engagement and motivation across hierarchies, especially if senior workers and higher performers see their skill-based comparative advantages erode.

This paper argues that GenAI represents a qualitative shift in AI, requiring a foundational reassessment of its role in management and organizations. Moving beyond the conventional view of AI as primarily a convergent technology focused on uncertainty reduction and optimization (see Amabile, 2020), GenAI’s generative capabilities in problem-solving tasks challenge theoretical frameworks that emphasize automation and efficiency (see Cabantous & Gond, 2011). Through four key dimensions – GenAI’s i) broad applicability as a GPT, ii) catalytic role in exploratory innovation, iii) function as a complementary cognitive artifact, and iv) democratizing effect – this paper proposes an expanded paradigm that integrates divergent perspectives essential for supporting creativity, innovation, and human potential. This technological phase shift, characterized by emergent capabilities that enable use in both convergent and divergent tasks, reinforces rather than diminishes the importance of human agency and imagination. Ultimately, human agency rather than technological determinism will shape how these technologies are developed and applied to serve organizational and societal objectives (Brusoni & Vaccaro, 2017; Krakowski, 2021).

Realizing GenAI’s potential requires a human-centered approach that places human welfare and agency at the core of design, implementation, and use (Faik et al., 2024; Krakowski, Haftor, Luger, Pashkevich, & Raisch, 2019; Shneiderman, 2022). GenAI’s democratizing aspects can amplify these mechanisms, promoting creativity, innovation, and broadly shared growth when coupled with appropriate policy and managerial practices (Brynjolfsson, 2022). To achieve these benefits, careful attention to complementary human capabilities and organizational practices is central to promote effective hybrid agency that preserves human engagement, reflection, and ethical responsibility. As organizations and society adjust to this profound technological transition, deliberate choices around the enabling implementation of AI for innovation and economic growth, equitable distribution of the benefits, and human-complementary development that preserves core human values and agency will determine whether GenAI’s transformative potential results in widely shared prosperity. The future of human skill development and flourishing may depend not primarily on AI’s expanding capabilities, but on the ability to develop and implement approaches that leverage these capabilities in service of human-centered objectives and outcomes.