Hafezieh, N. and Eshraghian, F. 2025. From Copilots to Autonomous Agents: Designing Responsible Agentic AI for Intelligent Enterprise Futures. 31st Americas Conference on Information Systems (AMCIS). Montréal, Canada 14 - 16 Aug 2025
| Title | From Copilots to Autonomous Agents: Designing Responsible Agentic AI for Intelligent Enterprise Futures |
|---|---|
| Authors | Hafezieh, N. and Eshraghian, F. |
| Type | Conference paper |
| Abstract | The emergence of agentic AI systems capable of autonomous action, adaptive planning, and context-aware interaction is shifting the boundaries of enterprise automation. Unlike conventional AI assistants, these systems can independently manage workflows, invoke external tools, and learn from feedback loops, enabling a new era of proactive, goal-oriented automation (Ng, 2024). However, this technological leap presents not only opportunities but also urgent challenges around trust, transparency, and system accountability. This research aims to investigate how organisations can strategically adopt agentic AI while embedding robust mechanisms for responsibility, resilience, and quality assurance. The first objective is to explore the organisational impact of agentic workflows such as self-critiquing agents, multi-agent collaboration, and iterative planning (Coshow, 2024) on operational agility, productivity, and innovation. The second is to define practical strategies for ensuring these systems operate ethically and transparently, including mechanisms for monitoring, human oversight, and the management of unintended consequences. Drawing on qualitative approach, the study will combine interviews and case analysis. A key focus will be the shift from traditional “software-as-a-service” models to “service-as-a-software” frameworks, where AI agents autonomously deliver business outcomes, from customer service resolution to predictive maintenance (Kamal, Ansari, & Chapaneri, 2024). While such models offer significant efficiency gains, they also demand rethinking how quality assurance is designed—moving from static testing to continuous, real-time evaluation and auditing mechanisms. Four agentic design strategies, reflection, planning, tool use, and multi-agent orchestration, will be analysed for their implications on organisational structure, AI governance, and user accountability (Li, 2024; Ng, 2024). The research will also examine how early adopters are addressing emerging risks, including data bias, lack of system resilience, and over-dependence on autonomous agents. By integrating insights from both AI engineering and organisational transformation, this study aims to develop a framework for responsible agentic AI adoption and design. It will contribute to the broader conversation on AI safety, offering a roadmap for enterprises aiming to leverage the capabilities of agentic AI without sacrificing quality, ethics, or human trust. |
| Keywords | Copilots; Autonomous Agents; Responsible Agentic AI; Intelligent Enterprise |
| Year | 2025 |
| Conference | 31st Americas Conference on Information Systems (AMCIS) |
| Accepted author manuscript | File Access Level Controlled (open metadata, closed files) |
| Web address (URL) of conference proceedings | https://aisel.aisnet.org/treos_amcis2025/125/ |