Generative AI (GenAI) tools are already widely used in many organizations for drafting, summarizing, analysis, and decision support. The next step many organizations anticipate is agentic AI: AI systems that can not only generate content, but also plan, execute multi-step tasks, interact with tools and systems, and operate with a degree of autonomy (e.g., initiating workflows, retrieving and updating information, coordinating actions across systems, and escalating to humans when needed).

While the technical capabilities of AI agents are evolving quickly, organizational readiness often lags behind. Broad adoption of agents is not just a tooling decision—it raises questions about process redesign, roles and responsibilities, governance, risk management, data access, compliance, and change management. Organizations will likely need to adapt how work is structured, how accountability is defined, and how quality, security, and control mechanisms are implemented.

This thesis is intentionally framed as a topic area rather than a fixed topic: it investigates what organizational change (e.g., in processes, governance, and ways of working) is needed to prepare the broad usage of AI agents in the workplace. The exact focus, industry context, and target organizations will be defined later together with the supervisor, depending on the student’s interests and access to interview partners.

The thesis may explore one or more of the following directions (non-exhaustive):

• Process readiness and redesign: Which existing processes are suitable for agent support, which need redesign, and how “human-in-the-loop” checkpoints should be integrated
• Roles, responsibilities, and accountability: How decision rights shift when agents act; who is responsible for outcomes; escalation and exception handling models
• Governance and controls: Guardrails, approval mechanisms, auditability, logging, monitoring, and performance management of agents in operational workflows
• Change management and adoption: Training and enablement, communication strategies, resistance and trust issues, and how organizations build agent literacy
• Measuring impact: How organizations define and measure “success” (productivity, quality, speed, employee experience, risk exposure) and manage trade-offs
• Pathways to scale: From pilots to broad rollout—patterns that help or hinder scaling (center of excellence, champions, operating model, procurement/vendor strategy)

To develop a qualitative, practice-grounded understanding of what organizations need to change to enable the broad, responsible, and effective use of AI agents at work.

The thesis should follow a qualitative research design, centered on:

• Semi-structured interviews with relevant stakeholders, e.g. process owners, team leads, operations managers, IT, risk/compliance, HR/change management, and employees who would work with agents
• Qualitative analysis (e.g., thematic analysis, coding, or a grounded-theory-inspired approach—exact method to be agreed)

Candidate profile

• Interest in digital transformation, information systems, organizational change, and empirical research
• Comfortable communicating with practitioners and conducting interviews in English (or local language, depending on context)
• Basic familiarity with qualitative methods is beneficial (but not required)
• Beneficial: background in business/information systems, management, industrial engineering, sociology/organizational studies, HCI, or related fields
   

Notes

This is intentionally framed as a topic area. The final research question(s), scope (e.g., specific processes or departments), and target population will be defined jointly during the initial phase of the thesis, based on the student’s interests and access to interview partners.

Relevant literature in this context is dependent on the exact unit of analysis and will be defined later on. An inclusion of whitepapers is possible due to the novelty of the topic.

If this thesis topic description sounds interesting to you, please reach out to the topic owner Leon Müller (leon.mueller@unisg.ch).