Introduction: From Assistance to Autonomy

Information systems (IS) research currently finds itself in a phase of conceptual reorientation: generative AI systems are no longer understood primarily as reactive assistance tools but increasingly as agentic information systems (Agentic IS) that independently pursue goals, initiate actions, and assume responsibility for tasks characterized by high uncertainty (Baird & Maruping, 2021). Agentic systems are thus no longer passive tools but artifacts capable of independently assuming rights and obligations in pursuit of goal attainment (Holldack et al., 2026). For IS research, this opens up a field of inquiry at the intersection of Human-Computer Interaction (HCI), organizational design, and intelligent information systems.

Theoretical Positioning and Conceptual Foundations

The growing prevalence of agentic systems calls for an advancement of established theories of socio-technical systems design. Whereas classical approaches in IS use research grant the human actor analytical primacy and treat technology as a passive tool, Baird and Maruping (2021) argue that this assumption must be revisited for a new generation of "agentic" IS artifacts. To this end, they introduce the concept of delegation as a theoretical lens for explaining the relationship between humans and agentic IS artifacts along the dimensions of endowments, preferences, and roles, as well as the mechanisms of appraisal, distribution, and coordination (Baird & Maruping, 2021).

In addition, the Task-Technology Fit (TTF) model provides an established framework for the fit between task requirements and technological characteristics (Goodhue & Thompson, 1995). For agentic architectures, this model needs to be extended along the following dimensions:

• Degree of autonomy: the extent of independent decision-making and action initiation (Baird & Maruping, 2021).
• Intentionality: the capacity for goal-directed planning across multiple sequential actions (Holldack et al., 2026).
• Fault tolerance and resilience: the ability to manage uncertainty within dynamic environments.
• Transparency: the traceability of decision logics and action pathways.

Methodological Approaches and Empirical Research

Studying agentic AI requires methodological designs that go beyond pure performance metrics and account for socio-technical effects. Multi-method research approaches are particularly well-suited to this task:

• Design Science Research (DSR): the development and evaluation of artifacts in the form of AI agents or multi-agent systems, differentiated by instantiations, models, and methods, using both ex-ante and ex-post evaluation designs (Hevner et al., 2004).
• Theory-driven delegation research: applying the delegation framework to model testable hypotheses about appraisal, distribution, and coordination mechanisms (Baird & Maruping, 2021).
• Qualitative case studies: examining organizational transformation processes and socio-technical interactions during the adoption of agentic systems (Yin, 2014).

Challenges and Key Research Areas

A central tension concerns the traceability of autonomous decision and action chains. Autonomous AI agents violate core assumptions of classical accountability attribution, since actions can no longer be unambiguously traced back to individual human decisions (Rahwan et al., 2019). Accountability must therefore increasingly be situated at the level of organizational governance architecture, which determines which goals an agent pursues and to what extent decision-making authority is delegated (Rahwan et al., 2019).

This gives rise to several key research areas:

• Agentic governance: developing frameworks for steering, controlling, and auditing autonomous systems.
• Delegation theory: empirically testing the delegation mechanisms proposed by Baird and Maruping (2021) in real organizational contexts.
• Shifting roles and competencies: examining the effects of agentic delegation on decision-making structures and managerial accountability.
• Process integration: embedding agentic IS within existing process landscapes.

Implications for Research

Agentic AI does not represent an incremental advancement but requires a fundamental revision of IS use theory, as reflected in the MIS Quarterly Executive Special Issue on Agentic AI (MIS Quarterly Executive, 2026). For IS research, this creates a pressing need to empirically validate models of delegation, governance, and trust. In particular, the interface between human oversight and machine autonomy remains a central focus of future research.

References

• Baird, A., & Maruping, L. M. (2021). The next generation of research on IS use: A theoretical framework of delegation to and from agentic IS artifacts. MIS Quarterly, 45(1), 315–341. https://doi.org/10.25300/MISQ/2021/15882
• Holldack, F., et al. (2026). Agentic information systems. Electronic Markets 36, 5. https://doi.org/10.1007/s12525-025-00861-0
• Rahwan, I., et al. (2019). Machine behaviour. Nature, 568(7753), 477–486.
• Goodhue, D. L., & Thompson, R. L. (1995). Task-technology fit and individual performance. MIS Quarterly, 19(2), 213–236. https://doi.org/10.2307/249689
• Hevner, A. R., March, S. T., Park, J., & Ram, S. (2004). Design science in information systems research. MIS Quarterly, 28(1), 75–105. https://doi.org/10.2307/25148625
• Yin, R. K. (2014). Case study research: Design and methods (5th ed.). Sage.
• MIS Quarterly Executive. (2026). Special issue call for papers: Are organizations ready for autonomous AI agents? https://lnkd.in/dx-wesDZ, accessed August 1, 2026.

Problem Statement and Scholarly Relevance

The increasing integration of AI systems into organizational decision-making processes marks a profound shift in the design of sociotechnical systems. While technological capabilities continue to advance, the scholarly grounding of governance structures and algorithmic responsibility (accountability) remains underdeveloped in many respects. In this context, algorithmic responsibility is understood as the traceability, assignability, and controllability of algorithmic decisions.

For the information systems discipline, this gives rise to a central research question: How can human oversight and algorithmic autonomy be configured in hybrid decision architectures such that efficiency gains are realized without violating regulatory and ethical requirements? From a behavioral IS perspective in particular, this entails examining how individual and organizational behavior patterns, attitudes, and trust shape the use and control of AI-based decision systems.

The disciplinary relevance is substantial, as information systems research has traditionally occupied the interface between technological innovation and organizational value creation. Whereas prior work has primarily focused on efficiency gains, the debate is increasingly shifting toward transparency, explainability (Explainable AI, XAI), and risk control in sensitive application domains such as financial services and public administration Rai, 2020.

Theoretical Concepts and Current Debates

At the core of behavioral IS research on AI systems lies the tension between algorithmic efficiency and human agency. Established adoption and use models such as the Technology Acceptance Model (TAM) and its extensions (e.g., UTAUT2) provide a structured lens for analyzing perceived usefulness, expected use benefits, trust, and habit as drivers of system use Davis, 1989; Venkatesh et al., 2012. For AI-based decision systems, however, a purely acceptance-oriented perspective is insufficient, as perceptions of control, opportunities to intervene, and attributions of responsibility must also be modeled.

An increasingly influential guiding concept in this context is meaningful human control, which specifies the conditions under which autonomous systems remain subject to accountable human control Santoni de Sio & van den Hoven, 2018. The notions of tracking and tracing enable normative requirements regarding control and responsibility to be integrated with behavioral constructs such as perceived control, willingness to assume responsibility, and trust.

In parallel, XAI research emphasizes that interpretability is not only a technical property but also a behaviorally relevant feature that shapes attitudes and usage patterns Rai, 2020; Doshi-Velez & Kim, 2017. For behavioral IS, this creates the challenge of empirically linking constructs such as trust, perceived risk and fairness, and responsibility attribution to concrete XAI features and governance mechanisms.

Methodological Approaches and Empirical Research

Investigating algorithmic responsibility from a behavioral IS perspective requires a methodological repertoire that captures both individual and organizational behavior patterns and technical artifacts:

• Quantitative surveys and experiments: Causal and correlational analyses of the effects of XAI features and governance mechanisms on trust, intention to use, perceived control, and responsibility attribution.
• Qualitative studies: In-depth analyses of how organizations engage with AI systems through interviews and observations, in order to uncover informal practices of oversight, delegation, and responsibility attribution.
• Design Science Research (DSR): Systematic development and evaluation of artifacts (e.g., governance dashboards, decision logs, audit mechanisms) that operationalize “accountability by design.”

Within IS research, DSR is understood as a research paradigm that generates knowledge through the construction and evaluation of innovative artifacts; such artifacts may take the form of constructs, models, methods, or implemented instantiations Hevner et al., 2004. For doctoral projects on algorithmic responsibility, this implies that a governance dashboard, an XAI explanation module, or an accountability framework should be conceived as explicitly theory-based artifacts whose design decisions can be traced back to behavioral IS theories (e.g., TAM/UTAUT, trust, risk, control) and normative requirements (e.g., meaningful human control, AI governance principles).

The design science research methodology proposed by Peffers et al. provides a procedural framework that systematically links problem identification, objective definition, design and development, demonstration, evaluation, and communication Peffers et al., 2007. A methodologically rigorous DSR study on algorithmic responsibility should therefore specify which type of artifact is being developed, which requirements are derived from behavioral IS theory and practice, and which evaluation forms—such as experiments, case studies, simulations, or analytical assessments—are used to judge the artifact’s quality Hevner et al., 2004; Peffers et al., 2007.

Mixed-methods approaches, which combine data science techniques (e.g., logfile analyses of actual usage patterns) with social science methods (surveys, experiments, qualitative analyses), are particularly promising for empirically capturing the gap between technical system logic and observed behavior in organizations Doshi-Velez & Kim, 2017.

Challenges and Implications for Research

A persistent core challenge is the “black box” nature of many AI models. Despite advances in XAI, the implementation of legally robust yet practically feasible transparency requirements remains limited, with direct consequences for trust, responsibility attribution, and usage patterns Rai, 2020; Doshi-Velez & Kim, 2017. At the same time, regulatory developments—most notably the EU AI Act—increase pressure on organizations to implement robust governance and documentation mechanisms European Union, 2024.

Against this backdrop, several key research questions arise that are closely aligned with behavioral IS and DSR:

• How do governance structures and XAI features shape trust, perceived control, and responsibility attribution across different user groups?
• Which artifacts (e.g., dashboards, explanation interfaces, audit trails) are suitable for making algorithmic responsibility operational in organizational contexts, and how can they be developed and evaluated in line with DSR principles?
• How can ethical principles and regulatory requirements be translated into behaviorally grounded design requirements and embedded in technically implementable decision logics?

Implications for Research Practice

Algorithmic responsibility is emerging as a core research area in information systems, in which behavioral IS and DSR offer complementary perspectives. Behavioral IS approaches enable theory-driven analyses and explanations of attitudes, perceptions, and usage patterns in relation to AI systems, while DSR advances the design and evaluation of artifacts that operationalize accountability, transparency, and control in practice Hevner et al., 2004; Peffers et al., 2007.

For doctoral researchers, this implies designing research projects that systematically link behavioral theories (e.g., TAM/UTAUT, trust, risk, control) with the development of governance and XAI artifacts. It is crucial not only to analyze algorithmic models themselves, but also to critically investigate their sociotechnical embedding and the associated behavioral and governance processes, and to actively shape these through the design and evaluation of artifacts Santoni de Sio & van den Hoven, 2018; European Union, 2024.

References

• Davis, F. D. (1989). Perceived usefulness, perceived ease of use, and user acceptance of information technology. MIS Quarterly, 13(3), 319–340.
• Venkatesh, V., Thong, J. Y. L., & Xu, X. (2012). Consumer acceptance and use of information technology: Extending the unified theory of acceptance and use of technology. MIS Quarterly, 36(1), 157–178.
• Rai, A. (2020). Explainable AI: From black box to glass box. Journal of the Academy of Marketing Science, 48(1), 137–141.
• Doshi-Velez, F., & Kim, B. (2017). Towards a rigorous science of interpretable machine learning. arXiv preprint arXiv:1702.08608.
• Santoni de Sio, F., & van den Hoven, J. (2018). Meaningful human control over autonomous systems: A philosophical account. Frontiers in Robotics and AI, 5, 15.
• Hevner, A. R., March, S. T., Park, J., & Ram, S. (2004). Design science in information systems research. MIS Quarterly, 28(1), 75–105.
• Peffers, K., Tuunanen, T., Rothenberger, M. A., & Chatterjee, S. (2007). A design science research methodology for information systems research. Journal of Management Information Systems, 24(3), 45–77.
• European Union. (2024). Regulation (EU) 2024/1689 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act). Official Journal of the European Union.

The Evolution of Autonomous Agents

Business informatics stands at the threshold of a fundamental transformation: while generative language models (LLMs) were primarily perceived as dialogue-based assistants over the past year, the current discourse is shifting toward agentic systems. These systems are characterized by their ability not only to generate text, but also to independently execute complex business processes through tool use, planning, and autonomous decision-making. For doctoral students, this shift opens up a highly relevant field of research that goes beyond mere implementation and raises fundamental questions about the structure of work and organizations.

Theoretical Classification and Concepts

The concept of agency—that is, a system’s ability to autonomously operate within defined goals—forms the theoretical core of this transformation. In contrast to last year, when research focused heavily on user acceptance of chatbots (based on TAM or UTAUT), attention is now returning to socio-technical systems theory. The key question is how to recalibrate the coupling of human expertise and algorithmic autonomy. Central theoretical reference points include concepts such as human-in-the-loop architectures and principal-agent theory in a digitized environment where the agent is no longer exclusively human but can also be a software entity.

Methodological Approaches and Research Designs

Studying agentic systems requires methodological innovation. While quantitative experiments can help quantify efficiency gains in standardized processes, understanding collaborative dynamics increasingly calls for qualitative longitudinal studies or design-oriented approaches (Design Science Research). Particularly promising for doctoral researchers are:

• Design Science Research: Development and evaluation of artifacts that act as intermediaries between AI agents and human decision-makers.
• Ethnographic studies: Examination of the gradual transformation of work processes in organizations that have already implemented initial agentic workflows.
• Comparative case studies: Analysis of different governance models in the introduction of autonomously acting AI systems in highly regulated industries.

Challenges and Tensions

Current research within the AIS Senior Scholars' Basket of Eight indicates that technological feasibility is currently outpacing theoretical discourse. Key areas of tension that offer opportunities for doctoral research include:

• Accountability: Who bears responsibility for erroneous decisions made by an autonomous agent operating within a chain of partial decisions?
• Transparency and explainability: How can the decision paths of complex agent workflows be made understandable to human actors without compromising performance?
• Skill erosion vs. empowerment: Does outsourcing complex cognitive tasks to agents lead to a loss of human competencies or an enhancement of job roles?

Implications for Practice and Future Research

For research practice, the rise of agentic systems implies a shift away from a purely output-oriented focus (text generation) toward process orientation. Practitioners face the challenge of not merely implementing individual tools but designing agentic ecosystems. This creates compelling entry points for doctoral research: for example, investigating the governance of AI agents in multinational corporations or modeling the changing work roles in departments such as controlling or procurement through the use of autonomous systems. The current debate invites a redefinition of the boundaries of human decision-making in the context of an algorithmically shaped world of work.