Latest / 1 More Paper / Agentic information systems — Holldack et al. (2026)
Transcript
- You’re listening to “Agentic information systems,” by F. Holldack, L. Banh, and G. Strobel. Published in 2026. Holldack, Florian; Banh, Leonardo; Strobel, Gero Article — Published Version Agentic information systems Electronic Markets Provided in Cooperation with: Springer Nature Suggested Citation: Holldack, Florian; Banh, Leonardo; Strobel, Gero (2026): Agentic information systems, Electronic Markets, ISSN 1422-8890, Springer, Berlin, Heidelberg, Vol. 36, Iss. 1, the linked source This Version is available at: the linked source Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. the linked source Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. Agentic information systems Florian Holldack1 · Leonardo Banh1 · Gero Strobel1 Abstract. Recent advancements in artificial intelligence (AI) have catalyzed the emergence of agentic information systems (IS), which exhibit autonomous behavior and advanced cognitive capabilities. Unlike traditional IS, which functioned primarily as reactive tools supporting humans, agentic IS can make decisions independently, act in unstructured environments, and even delegate tasks to humans. This paradigm shift fundamentally transforms the human-IS relationship, questioning the long-standing assumption of human agentic primacy in IS research and practice. In this article, we provide a conceptual overview of agentic IS, delineating their defining characteristics and situating them within the broader evolution of IS. We introduce key archetypes of agentic IS, explore novel patterns of delegation and interaction between humans and machines, and discuss the socio-technical implications of these developments. Furthermore, we highlight the challenges and risks associated with integrating agentic IS from an individual, organizational, and societal perspective, emphasizing the need for nuanced understanding to harness the potential while addressing emerging complexities. JEL Classification M21 · O3 Introduction. “The IT department of every company is going to be the HR department of AI agents in the future” - Jensen Huang, NVIDIA CEO, on AI Agents at CES 2025 As the boundaries between human and technologi-cal capabilities continue to blur, traditional perspectives on information systems (IS) are facing unprecedented challenges. IS research has Responsible Editor: Ricardo Büttner. the email address Leonardo Banh the email address Gero Strobel the email address historically examined how humans interact with techno-logical artifacts to complete tasks, while positioning IS as passive tools that support human decision-making and action. This perspective establishes human agency as the dominant theoretical assumption in IS research. However, recent advancements in artificial intelligence (AI), such as generative agents, have rendered this notion of agentic primacy for human agents increasingly insuffi-cient. Modern IS now demonstrate advanced cognitive capabilities, enabling autonomous decision-making and action under uncertainty. This evolution has given rise to agentic IS, where technological artifacts possess significant agency themselves, thereby fundamentally transforming the traditional understanding of the human-IS artifact relationship. Agentic IS have evolved to dynamic, active entities that are no longer solely dependent on or subordinate to humans. Their agency derives from the ability to make decisions and act autonomously in unstructured environments, without requiring human intervention or initiation. Self-driving vehicles, generative AI-based software developers, and human-like recommender systems exemplify the growing prevalence of agentic IS in various domains. With the ability to autonomously handle tasks previously reserved for humans, they function as independent entities in socio-technical systems, capable of meaningful collaboration and even supervision within the existing human workforces. Furthermore, the transfer of rights and responsibilities concerning task-related decision-making and action is no longer unidirectional; agentic IS artifacts have the capacity to delegate tasks to humans as well. This shift challenges the traditional distinction between IS artifacts and human agents, as both are increasingly regarded as autonomous entities with shared objectives, sophisticated cognitive functions, and a high degree of agency. Consequently, these advancements enable novel forms of collaboration between humans and IS, fostering hybrid teams that work together in various configurations to achieve common goals. Florian Holldack Despite these potential advantages, effectively integrating agentic IS and actualizing their potential present substantial challenges (e.g., ensuring effective delegation between human and IS as well as managing accountability). Moreover, a comprehensive exami-nation of the key concepts, characteristics, and impacts of agentic IS remains to be conducted, leaving major aspects of this new generation of IS undefined. It is therefore imperative to develop an understanding of the distinctive characteristics of agentic information systems compared to traditional systems, as well as their socio-technical implications. This knowledge empowers organizations to effectively integrate agentic IS artifacts into their workforce while anticipating challenges. To develop this understanding, we first conceptualize agentic IS by situating it within the evolution of IS toward demonstrating advanced cognitive capabilities and human-like agency. Second, we elaborate on emerging agentic IS archetypes grounded on distinct agency perspectives. Additionally, we identify and conceptualize emerging delegation patterns, thereby presenting novel human–machine interaction mechanisms in the era of agentic IS. Subse-quently, potential obstacles and risks associated with changes in the IS landscape and corresponding future research opportunities are addressed. Finally, we present concluding remarks that emphasize the need to recognize the transition toward agentic IS. Conceptualization of agentic IS The conceptualization of IS as socio-technical systems is centered on the premise that humans employ passive IS artifacts to achieve specific objectives. This traditional IS paradigm is characterized by an unidirec-tional interaction pattern wherein humans delegate tasks while being responsible for decision-making and acting, thereby assuming primary agency. Within this paradigm, IS artifacts are reactive entities handling well-defined tasks. Although capable of surpassing human abilities in specific areas like large-scale data analysis, they inherently lack human-like features, including autonomy, generalizability across tasks and domains, or learning abilities. However, over the past decades, IS have evolved significantly (see Fig. 1), leading to novel types of IS that aim to match or even surpass human abilities, thus questioning human agency primacy. Early IS centered on the aggregation of vast amounts of data, serving as decision support systems for humans. Nevertheless, the responsibility for deriving meaningful insights remained predominantly with the user. Subsequent genera-tions of IS introduced expert systems and knowledge-based systems possessing reasoning abilities, thereby reducing human effort and enhancing the overall value of the decision support. Despite these advancements, the primary function of them was to facilitate task-specific automation or augmentation for narrowly defined objectives, with limited adaptability to dynamic changes. Moreover, they relied on structured information and human users to leverage them as tools for improved task completion and goal attainment. The next major evolution of IS capabilities introduced intelligent agents, implemented through either symbolic knowledge representation or machine learning models. These intelligent agents (e.g., fraud detection in finance) significantly reduced the need for human intervention by autonomously respond-ing to environmental stimuli and executing suitable actions. Although capable of (partly) independent decision-making, they were still developed for task-specific automation, with adaptability constrained by narrow perception requirements for specific data structures. Consequently, these IS functioned as tools for human users to automate well-defined tasks, thereby leaving the traditional IS paradigm largely unaffected. Fig. 1 Evolution of cognitive capabilities in IS The emergence of agentic IS artifacts facilitates the next generation of IS, powered by advances in AI such as deep learning, generative AI, and generative agents. These artifacts can establish and pursue long-term objectives, make and act upon independent decisions under uncertainty, and maintain control over their internal states. They extend the abilities of previous IS toward human-like features such as multimodal perception, complex reasoning, continuous learning, natural language communication, or sophisticated adaptation in evolving environments. These cognitive capabilities enable them to transfer rights and responsibilities for task execution and outcomes from and to human agents. Thus, agentic IS artifacts increasingly represent human-like actors in various domains (e.g., software development or higher education), demonstrating agency comparable to humans. Toward a new IS paradigm The paradigm shift toward agentic IS emerges as artifacts demonstrate human-like agency through advanced cognitive capabilities, challenging the human agency primacy. This development disrupts the foundation of the traditional IS paradigm, creating a new bal-ance of agency and novel role distributions between humans and IS artifacts. The term agency is typically associated with an entity’s ability to act intentionally, which requires the presence of a cer-tain cognitive capacity that facilitates the generation and execution of decisions. However, within IS research, agency is conceptualized in more distinct ways (see Fig. 2), namely into entity-focused and relational perspectives. The entity-focused perspective delineates agency as a property of a specific entity relating to either inherent characteristics (inherent agency) or specific behavior (behavioral agency). The concept of inherent agency is centered around an entity’s mental state and is attributed to those possessing human-like properties such as consciousness, intentionality, or free will. This concep-tion emphasizes human agency while conceptualizing IS agency as a proxy that enacts human intentions. Behavioral agency refers to an entity’s capacity to rationally take actions that influence the environment without requiring inherent mental properties such as consciousness. Therefore, this agency perspective can be attributed to various entities capable of autonomously making decisions and acting accordingly. For instance, while an AI-based recommender system lacks inherent agency as it is purposefully designed by humans, its autonomous behavior can be attributed to a degree of behavioral agency. In essence, the entity-focused perspective facilitates the attribution of distinct agency levels to specific entities within institutions,1 thereby enabling comparisons between IS artifacts and other entities, such as humans. On the other hand, the relational perspective conceptualizes agency as emerging from the sociomaterial entanglement between humans and IS artifacts. Instead of treating agency as an entity-specific attribute, it defines agency as a systemic phenomenon shaped by human-IS collaboration. Accordingly, agency is materialized through the dynamic interplay between humans, agentic IS artifacts and institutions. Weaker notions still distinguish between social agency (human-related) and material agency (IS-related), while emphasizing their deep interconnection in creating sociomaterial phenomena. Strong relational perspectives are grounded in the ontology of agential realism, thus concentrating on the affordances of the human-IS collaboration without dis-tinguishing between entity-specific agency levels (i.e., non-human and human agency) (Harding et al., 2021; Visser & Davies, 2021). Moreover, the relational perspective avoids imposing human-like agency criteria (i.e., anthropomorphic assumptions) on artificial entities, thereby facilitating a comprehensive understanding of human-agentic IS dynamics within institutions. Agentic IS research predominantly employs the entity-focused perspective to capture and compare varying levels of agency, where agency is broadly defined as “the ability to accept rights and responsibilities for ambiguous tasks and outcomes under uncertainty and to decide and act autonomously”. This conceptualization primarily adopts behavioral agency, without specifically distinguishing between human and non-human agency. Furthermore, it emphasizes an increase in cognitive capabilities that facilitates independent decision-making and acting for open-ended and intricate tasks. Thus, higher levels of agency enable agentic entities to take on rights and responsibilities for more complex tasks and the corresponding outcomes. To describe these entities, the concept of agent is commonly utilized, referring to various entities characterized by some degree of agency. Within agentic IS, an agent is not necessarily an artificial entity, e.g., an AI agent; rather, it can be also represented by other actors that partici-pate inside institutions, such as human agents. Both humans and artificial agents are able to transfer rights and responsibilities regarding task-related decision-making and actions toward each other. Therefore, instead of the traditional IS paradigm of humans leverag-ing IS artifacts as tools toward task completion, the agentic IS paradigm employs a novel perspective on agents with progressively analogous capabilities and degrees of agency operating in diverse constellation. Agentic IS archetypes Depending on the employed conceptualization of agency, there are various agentic IS archetypes that encompass a spectrum of functions, from assisting humans in ambiguous tasks with enhanced capabilities to automating comprehensive work procedures (see Table 1). From an entity-focused agency perspective, these IS artifacts differ in their individual levels of agency, which enables them to assume novel, more complex tasks without human intervention or initiation. From a relational perspective, the degree of human-agentic IS integration differentiates them with higher relational agency representing more close collaboration. Table 1 Overview of core agentic IS archetypes Employing the entity-focused or weak relational perspective, humans as well as IS artifacts can possess some degree of agency, resulting in varying novel human-IS constellations and corresponding archetypes. Considering the conventional distribution of agency (i.e., based on the traditional IS paradigm), the assisting agentic IS archetype emerges, wherein humans possess the agency primacy, while IS artifacts are positioned on the lower ends of the agency continuum. Thus, rights and responsibilities regarding decision-making and actions are primarily attributed to humans. Within this archetype, IS artifacts primarily function as reactive tools requiring human initiation and intervention, making them suitable for assisting humans during task completion and decision-making. For instance, generative AI-based applications (e.g., ChatGPT or Claude) can assist users in optimizing text fluency, generating images for social media, or supporting the identification and correction of coding errors. While these artifacts exhibit some degree of agency, such as interpreting user prompts and generating corresponding outputs, their agency remains substantially limited to human-provided tasks and scopes. However, as agentic IS acquire higher degrees of agency, the autonomous agentic IS archetype manifests, which challenges or even reverses the traditional roles by attributing lower levels of agency to humans. Within the reversed role, human agents become increasingly dependent on agentic IS artifacts' decision-making capabilities (e.g., autonomous trading agents), as human influence and control over the procedure and overall outcome is significantly reduced. These artifacts are increasingly capable of autonomously handling ambiguous, long-term task within intricate environments, thereby reducing or even removing the necessity for human initiation and intervention. Moreover, decisions made by agentic IS artifacts can lead to social actions that actively influence the behavioral space of human agents, with some artifacts even prescribing specific actions, potentially resulting in a complete reversal of traditional supervision and subordi-nate roles. Thus, agentic IS artifacts can be employed to direct other agents (including humans) by prescribing tasks, tracking and evalu-ating overall progress and even disciplining them through rewards or replacement. The assisting and autonomous archetypes both primarily describe situations in which one entity holds superiority in agency. Nevertheless, within the agentic IS paradigm, the agency of human agents and IS artifacts can converge, with both entities possessing high degrees of agency, leading to the collaborative agentic IS archetype. In these instances, agentic IS can function as teammates or colleagues working in collaboration with human agents in order to achieve shared objectives. Thus, both agents can take on rights and responsibilities for tasks and problems in a similar manner while actively com-municating with each other (e.g., generative agents in software engineering). This transforms agency primacy into a dynamically shifting property, preventing clear distinction and attribution to specific entities. These boundaries blur further with advanced human-IS interfaces that facilitate dense integration into human decision-making and acting (e.g., AI-based augmented reality glasses or even futuristic brain-computer interfaces), referring to the hybrid agentic IS archetype. Instead of maintaining distinct and separate levels of agency, these heterogeneous teams exhibit conjoined agency, i.e., a shared capacity to act intention-ally. However, the entity-focused and weak relational perspectives assume both the possibility of attributing agency to specific entities and the suitability of this approach for capturing the effects and implications of the emerging agentic IS paradigm. Consequently, these perspectives become increasingly inadequate for understanding the agency dynamics and implications within highly integrated human-agentic IS teams, thereby calling for analysis through the theoretical lens of sociomateriality. The strong relational perspective builds on these limita-tions by focusing on the collaboration between human and IS artifacts and thus facilitates a more comprehensive perspective on the collaborative and hybrid agentic IS archetype. It defines agency by the actions that are afforded or constrained through their relationship. This shifts attention to the bilateral interplay between humans and agentic IS artifacts and their shared capacity to act within organizations leading to co-creation, framing agency as a systemic phenomenon rather than an entity-based attribute. Therefore, this agency influences which entities are able to perform what actions and how those actions are conducted. Hence, more sophisticated human-IS collaborations (e.g., collaborative risk assessment in courts) influence the sociomaterial entanglement by affording new action poten-tials. Moreover, the relational perspective becomes increasingly relevant when considering networked businesses in which human agents collaborate with multiple artificial agents. In these environments, multi-agent systems demonstrate high collective agency even when individual entities exhibit low agency, requiring system-level analysis to capture their com-bined effects. However, the strong relational perspective does not provide a framework for differentiating how agency is distributed across individual entities, thereby blurring responsibility boundaries and hindering account-able assignments of control. Nascent collaboration patterns In light of agentic IS assuming novel roles, the traditional IS paradigm becomes insufficient for conceptual-izing the multifaceted interaction and collaboration patterns between human and IS artifacts. The theoretical framework of delegation has emerged as the prevailing alternative for describing and analyzing the relationships within the agentic IS paradigm. Delegation is defined as the transfer of rights and responsibilities for making decisions and per-forming tasks to another agent. The objective is aimed at enhancing productivity and flexibility by recognizing that a single agent may lack the resources, capabilities, or knowledge necessary to achieve complex, multi-step goals indepen-dently. This process requires metaknowledge, i.e., a comprehensive understanding of an individual’s capabilities, those of other agents, and the overarching objectives. While delegation has always been cen-tral within the IS paradigm, agentic IS artifacts enable extended forms of delegation and even reversed patterns toward humans (see Fig. 3). These patterns of delegation build on and extend concepts from related fields like HCI’s mixed-initiative interaction by specifically foregrounding the transfer of organizational rights and responsibilities. User-invoked delegation is the most common form of delegation where human agents transfer rights and responsibilities to IS artifacts or other agents while retaining delegation ownership. The delegation procedure involves temporarily transferring (sub)tasks to agentic IS artifacts with or without direct human supervision. Thus, humans commonly oversee task progression and maintain accountability for outcomes. Within this paradigm, there exists a spectrum of human supervision levels, thereby providing human agents with varying levels of control. The most restrictive is the “human-in-the-loop” approach, where a human agent manually approves all decisions or at least those of a critical nature made by the agentic IS artifact prior to execution. This approach significantly limits the agency of the IS artifact and thus restricts the transfer of rights and responsibilities regarding autonomous decision-making. A less restrictive approach is referred to as “human-on-the-loop,” where humans actively monitor decisions and actions made by agentic IS artifacts but intervene only when necessary. Current self-driving vehicles exemplify this pattern: while they operate autonomously, drivers must remain alert to intervene in dangerous situations. The most advanced form of user-invoked delegation entails the complete transfer of rights and responsibilities for a specific task or even long-term objectives to an agentic IS artifact, Fig. 3 Agentic IS delegation patterns without requiring human supervision. The bidirectional delegation pattern involves multiple entities, such as human-agentic IS teams, closely working together toward shared objectives. This collaborative approach incorpo-rates dynamic delegation patterns, where ownership transi-tions between humans and IS artifacts in accordance with situational demands and comparative advantages. It aims to leverage the complementary strengths of both entities as they work toward common goals that neither could achieve indepen-dently. However, this collaboration can be categorized into two forms: augmentation and assemblage, which represent different degrees of collaborative integration. Augmentation manifests in human-IS teams function as independent co-workers who support each other by exchanging rights and responsibilities for tasks. In this form, leadership transi-tions in a fluid manner, with both parties having the capacity to initiate tasks, delegate responsibilities, and oversee processes. Assem-blage represents a more integrated approach to human-IS teams, where humans and agentic IS do not have to explicitly conduct delegation but instead merge into seamless, nearly indistinguishable units. Thus, they are contextually and tempo-rally assembled, functioning as integrated entities (e.g., AI-powered robots in medical surgery) that collaborate and interact continuously. This collaborative mode exhibits a reduced reliance on explicit delegation between integrated units, as they inher-ently share common objectives, tasks, and responsibilities. In scenarios where groups of humans and agentic IS artifacts collaborate, delegation ownership evolves into a collective and distributed attribute. It therefore represents a dynamic continuum of changing rights and responsibilities toward task completion rather than a clear transition of authority to any single entity or unit of two. IS-invoked delegation emerges as a novel perspective regarding the collaboration between humans and IS. This pattern contrasts with exclusive human agency by enabling agentic IS artifacts to anticipate delegation toward IS or to directly delegate to human agents. Weak IS-invoked delegation occurs when agentic IS artifacts proactively suggest delegation to human agents or even from humans to artificial entities, such as software engineering agents anticipating that a human developer requires task-support. Accordingly, while the IS agency experiences an increase, human agents maintain delegation ownership. In contrast, with strong IS-invoked delegation, both delegation ownership and task leadership transfer to agentic IS artifacts. This type of IS-to-human delegation involves the transfer of responsibility for task and process outcomes to agentic IS artifacts, thereby fundamentally challenging the suit-ability of the predominant IS paradigm. Thus, agentic IS artifacts have the capacity to pursue long-term, multi-objective tasks and to delegate problems that they are unable to solve indepen-dently to other agents, including humans. In other scenarios, agentic IS artifacts may function as coordinators, supervis-ing and delegating tasks to other agents without actively par-ticipating in problem-solving (e.g., supporting the software development cycle by delegating responsibilities to other entities). While this direction of delegation has the potential to reduce unnecessary cognitive load on humans by allowing them to concentrate on more complex tasks, its success is contingent on the capabilities of the IS artifact and requires a high level of trust from humans to work effectively. Challenges Although agentic IS artifacts pose potential benefits for productivity and efficiency through novel collaboration and automation approaches (e.g., AI teammates or autonomous vehicles), the agentic IS paradigm introduces several new challenges. They range from facilitating effective delegation between humans and agentic artifacts, to managing new distributions of accountability, especially in the case of fully autonomous artifacts. Metaknowledge Effective collaboration between humans and agentic IS artifacts depends on sufficient levels of metaknowledge, referring to the ability to accurately assess one’s own capabilities and those of collaborative partners. This awareness enables agents (both humans and artificial entities) to make informed decisions about task allocation and responsibility distribu-tion, including when to collaborate and how to interpret, accept, or challenge one another’s outputs. Humans tend to lack the metaknowledge necessary for effective collaboration with agentic IS artifacts, leading to misjudgments of capabilities and task complexity, particu-larly in uncertain situations. In contrast, IS-initiated delegation tends to achieve better outcomes, as agentic IS artifacts are observed to delegate tasks more effectively, thereby avoiding some of the metaknowledge limitations seen in humans. However, this approach creates a fundamental tension: While IS-initiated delegation can optimize collaborative performance by addressing the metaknowledge gap, it reduces human agency and control. This reduction potentially conflicts with accountability requirements and limits applicability in domains where human supervision is man-dated due to the limited cognitive capabilities of the artificial entity. Alternative approaches, such as weak IS-initiated delegation where agentic IS artifacts provide suggestions rather than direct task assignments, can partially address the control issue while still addressing the lack of human metaknowledge. These intermediate solutions cannot fully capture the collaborative potential of bidirectional delegation patterns found in more sophisticated agentic IS (i.e., collaborative agentic IS or hybrid agentic IS) architectures. Moreover, they do not resolve humans’ difficulty in critically evaluating system outputs when metaknowledge is lacking, increasing the risk of overreliance or unwarranted rejection, which can undermine collaboration. Consequently, developing new mechanisms to enhance and coordinate metaknowledge in heterogeneous human-agentic IS teams represents a critical challenge for realizing the full potential of agentic IS collaboration. Delegation To facilitate human-agentic IS collaboration within organizations, effective delegation is imperative. However, several factors influence humans’ willingness to delegate rights and responsibilities to agentic artifacts and their acceptance of IS as a delegating entity, including perceived control, trust, accountability, and cognitive effort. Although these factors are present in traditional IS relationships, they become more prominent as IS artifacts develop increased agency, and human agency diminishes. As agentic IS artifacts serve as team members (collaborative agentic IS) or even supervisors (autonomous agentic IS), they increasingly challenge human control and willingness to collaborate. As a result, trust emerges as a pivotal factor in effective delegation performance and human-agent IS collaboration. Within the agentic IS paradigm, trust is influenced by various factors including artifact design elements (e.g., transparency and anthro-pomorphism), human characteristics (e.g., personality and knowledge level), and task features, such as whether outcomes affect oneself or others. While low levels of trust can hinder effective delegation between human and agentic IS artifacts, excessive trust, such as unquestioningly accepting IS behavior, can result in irresponsible collaboration. Although user-initiated delegation enhances trust and the sense of control compared to IS-initiated delegation, it can reduce delegation effectiveness. Thus, effective human-agentic IS collaboration demands novel approaches to balance agency and trust relationships between humans and artificial agents to facilitate effective delegation. Human centricity With enhanced cognitive capabilities and human-like agency, agentic IS artifacts are increasingly challenging traditional notions of human-centricity within the workforce, thereby suggesting potential disruptions to employment patterns (e.g., fully automated warehouses). However, with humans maintaining exclusive inherent agency (e.g., consciousness or intentionality) and possessing unique capabilities (e.g., empathy, adaptability, unstructured problem-solving, or moral reasoning), a human-centered perspective is necessary. This is par-ticularly evident in high-risk contexts, such as clinical or discretionary judgments, where empathetic communication, adaptable reasoning, and value trade-offs are paramount. Moreover, combining the comple-mentary capabilities of humans and agentic IS artifacts has the potential to yield performance beyond what either could achieve alone. Therefore, instead of focusing solely on human replacement, the agentic IS paradigm has to concentrate on the development of hybrid workforces that leverage the unique strengths of both agentic IS artifacts and human agents. This hybrid workforce perspective requires a fundamental theoretical shift away from sole entity-focused agency, which views humans and IS artifacts as separate and potentially converging entities, toward relational agency, which conceptualizes agentic IS as a systemic phenomenon. This perspective highlights what is afforded through the sociomaterial entanglement of humans and agentic IS artifacts collaborating closely. Nevertheless, adopting relational agency as a theoretical lens for human-AI collaboration introduces several challenges. It requires reconceptualizing constructs such as controllability and accountability at the information system level rather than assigning them to individual entities and developing frameworks to distinguish between varying degrees of relational agency (e.g., assisting agentic IS vs. hybrid agentic IS). Addressing these challenges is essen-tial for the widespread adoption of relational agency as a productive framework for understanding and optimizing the collaboration between humans and agentic IS artifacts, without solely considering agentic IS as means of automatization and replacement of human agents. Accountability With IS artifacts taking on delegation ownership, they con-currently assume responsibility for the outcomes of tasks or processes. This is especially pertinent when agentic IS artifacts autonomously handle tasks rather than delegating them to human agents. However, regulations such as the EU AI Act prohibit the transfer of legal accountability to IS artifacts, mandating human involvement. Thus, while rights and responsibilities related to decision-making and action are transferred to agentic IS artifacts, legal accountability must remain with human agents as the governing entity. Current regulatory frameworks address this discrepancy by attributing accountability to external entities, such as devel-opers or providers of agentic IS artifacts. Alternative approaches neces-sitate direct human supervision (e.g., human-in-the-loop), which separates legal accountability but potentially reduces the efficiency gained through increased IS agency. Thus, addressing legal accountability is particularly challenging when supervisory roles are dynamic (e.g., collaborative agentic IS) or reversed (e.g., autonomous agentic IS). Additionally, the increasing delegation of cognitively demanding tasks to agentic IS artifacts creates a paradoxical situation in which human agents remain accountable yet may progressively diminish their capacity to effectively supervise agentic IS, e.g., through skill erosion. This growing capability gap limits effective human oversight and complicates responsibility allocation within the agentic IS paradigm. Contrary to the entity-based assumptions that accountability can be allocated to specific entities through structural or formal mechanisms, the relational perspective conceptualizes accountability as an emer-gent property of dynamic attribution processes. Within this perspective, accountability is not merely a property to be assigned to an entity but rather an ongo-ing practice that is enacted through the entanglement of algorithms, interfaces, organizational routines, and human actions. Additionally, psychological dynamics further influence accountability. When humans attribute agency to IS artifacts (referring to sense of agency), they experience a reduced sense of control and responsibility, lowering their perceived accountability despite retained legal accountability. This psychological dynamic can undermine human supervision methods, high-lighting the need for new mechanisms that manage varying senses of agency and perceived accountability. As a result, novel governance mechanisms are necessary to ensure human legal accountability while still facilitating efficient and effective integration of agentic IS artifacts into hybrid teams. With agentic IS artifacts becoming increasingly preva-lent in daily life (e.g., artificial sales representatives, AI-based supervisors, autonomous vehicles), addressing these fundamental issues is imperative for effective integration. They impact individuals, organizations and society through changes in collaboration and interaction patterns and by affecting the role of humans within the digital economy. To address these challenges, we provide research questions from an individual, organizational, and societal perspectives as a starting point for future research (see Table 2). Conclusion. This fundamentals article introduces the emerging paradigm of agentic information systems. By examining the increasingly blurred boundaries between the capabilities of humans and IS artifacts, we conceptualize the agentic IS paradigm and distinguish it from the traditionally employed IS paradigm. Thus, we examine the progression from static and reactive tools to agentic entities that possess advanced cognitive capabilities and agency comparable to humans. Furthermore, we elaborate on the concept of agency as a theoretical foundation for understanding agentic IS. The paradigm shift enables new forms of interaction between humans and IS artifacts, thereby challenging the traditional primacy of human agency. Consequently, the nature of collaboration within IS undergoes a transformation, with the emergence of bidirectional delegation patterns leading to the identifi-cation of promising opportunities across various domains. The article also identified four major challenges coordinating metaknowledge, ensuring effective delegation, maintaining human-centricity, and managing accountability and derived corresponding research questions to address them at the individual, organizational, and societal level. In order to success-fully navigate the transition toward agentic IS, researchers and practitioners must understand these new dynamics and emerging challenges. This knowledge is imperative for the effective management and governance of agentic IS. Table 2 Future research questions regarding the challenges of agentic IS Funding Open Access funding enabled and organized by Projekt DEAL.