Evaluating AI-Assisted Work: How Managers Judge Generative AI Use in Everyday Work
This thesis explores how managers in a regulated Swiss financial organization evaluate colleagues’ use of generative AI. It shows that managers focus less on whether AI was used and more on whether employees can explain, validate, and stand behind the final result.
Topic
This Master thesis explores how managers in non-technical roles within a regulated Swiss financial organization perceive and evaluate colleagues’ use of generative AI. Prior research suggests that people may face social penalties when disclosing AI use, for example being seen as less competent or less diligent. This study examines how such perceptions appear in everyday managerial evaluation, where managers know the employee, understand the task context, and work in an organization that actively encourages GenAI use.
Relevance
The topic is relevant because GenAI is increasingly encouraged in organizations, while employees may still worry about how AI use is perceived by others. For managers, this creates a practical challenge: they need to evaluate AI-assisted work while maintaining standards of quality, ownership, confidentiality, and accountability. By focusing on one regulated financial organization, the study shows how perceptions of GenAI use are shaped in a real work setting. The findings are especially relevant for similar knowledge-intensive and regulated organizations.
Results
The findings show that managers did not generally interpret GenAI use as a sign of laziness or lower competence. Instead, they focused on whether the person delivering the work could explain, validate, and stand behind the result. Three patterns emerged. First, existing evaluation standards remain, but expectations rise. Second, AI engagement increasingly signals openness to change and willingness to learn. Third, context determines when AI use matters, especially where sensitive data, communication, or human judgement are involved.
Implications for Practitioners
- Employees remain responsible for the final result, regardless of whether GenAI was used.
- Managers should evaluate AI-assisted work through quality, judgement, ownership, and task context rather than tool use alone.
- AI training should address validation, task fit, sensitive data, and responsible use.
- Teams should clarify when AI use should be disclosed and when disclosure is unnecessary.
- Organizations should support slower adopters early, before not using AI starts to be perceived as a negative signal.
Methods
This thesis used a qualitative single-case study design. Fifteen semi-structured interviews were conducted with managers in non-technical roles within a regulated Swiss financial organization. The interviews explored how managers notice, interpret, and evaluate colleagues’ GenAI use in everyday work contexts. The sample focused on managers who assess, review, or approve the work of others. The data were analyzed using the Gioia methodology, moving from first-order concepts close to the interview material to second-order themes and aggregate dimensions. This approach allowed the study to develop an inductive understanding of how GenAI use is socially interpreted in managerial evaluation.