Virtual assistants as support in knowledge work - the case of marketers

Artificial intelligence is increasingly appearing in everyday work, helping to write texts, organize information, plan activities, and support decision-making. However, at the time of research within the grant it remained unclear when it facilitates work and when it might complicate it. The aim of the project was to investigate how human collaboration with an intelligent assistant affects productivity and job satisfaction in knowledge work. Marketing was chosen as an example because it is an area that combines creativity, data analysis, communication, and time pressure. From the outset, the project had an interdisciplinary character, combining management, occupational psychology, technology research, and organizational neuroscience.

The research was carried out in stages. First, the team conducted a literature review to synthesize existing knowledge about human-AI collaboration. Next, interviews and surveys were conducted to better understand what employees expect from virtual assistants and which system features they consider useful. The subsequent step was the creation of a proprietary research tool - a chatbot - and a digital environment for conducting experiments. During the project, the interface was developed, system functions were tested, and conversation logs were analyzed. Later, experiments comparing work with and without AI were conducted alongside a laboratory study using psychophysiological measurements. As part of this, a research facility equipped with specialized psychophysiological apparatus was established at Kozminski University.

The primary conclusion of the project is that AI can realistically support humans at work, but it does not do so automatically or under all conditions. The results showed that generative AI can improve efficiency in tasks related to automation, support, content creation, and innovative processes. The studies also revealed that the user's characteristics and their relationship with the technology are of paramount importance. The mere "pleasure of use" is insufficient to motivate continued utilization - trust plays a pivotal role. The team also determined that modern marketers do not expect AI to be merely a simple tool, but rather a reliable, communicative partner that fits seamlessly into a fast-paced work environment.

The results of the psychophysiological study were particularly innovative. They demonstrated that the way AI "behaves" toward a human—for example, whether it is supportive or more dominant and antagonistic - affects not only user behavior but also their physiological arousal and the quality of their work. Importantly, participants were not always aware of this effect. This suggests that evaluating AI systems relying solely on self-reported surveys may be inadequate. This represents an important contribution to the field, shifting the focus from the question "Does AI work?" to "Under what conditions does it work well and safely for humans?"

The significance of the project extends beyond academia. The findings can help organizations better implement AI tools, design them in a more human-centric manner, and avoid solutions that ostensibly increase convenience but in practice elevate stress or degrade work quality. The project also expanded the current understanding of human collaboration with generative AI, resulting in scientific publications and new theoretical frameworks. Its outcome is not just theoretical knowledge but also a permanent research infrastructure, which is already being utilized for subsequent research and the training of future study participants.