How and when do employees rely on generative AI chatbots to handle mental tasks at work?

Host: Welcome to A.I.S. Insights — powered by Living Knowledge. I'm Anna Ivy Summers.
Expert: And I'm Alex Ian Sutherland.
Host: Today, we are discussing a study titled "Thinking with Chatbots: Cognitive Offloading in Human–GenAI Interaction at Work," presented at the International Conference on Information Systems in 2026 by researchers Jacopo Valentini, Haiat Perozzo, and Aurelio Ravarini. The study investigates how professionals integrate generative AI chatbots into their daily cognitive work and decision-making.
Expert: Historically, digital tools like calculators or databases handled discrete, lower-level mental tasks, such as storing records or executing arithmetic. Generative AI, however, can structure arguments, synthesize qualitative information, and recommend courses of action.
Host: Which changes the nature of delegation. Yet organizations often lack clarity on the conditions under which workers hand over cognitive effort versus exercising active judgment.
Expert: That is the central question the authors explored. To do so, they conducted an inductive, qualitative thematic analysis based on twenty semi-structured interviews with professionals across diverse organizational roles and functions.
Host: Let us look at what the study uncovered. What are the primary ways employees utilize these chatbots on the job?
Expert: The researchers categorized four main workplace uses: preparing written outputs, organizing task-related information, retrieving or checking information, and comparing alternatives before taking action.
Host: And the study observed that this delegation is not uniform across tasks.
Expert: Correct. The authors developed a framework demonstrating that cognitive offloading is shaped by two core dimensions: perceived mental effort and perceived knowledge sufficiency. Depending on where a task falls along those axes, workers adopt one of four offloading modes.
Host: How do those four modes function in practice?
Expert: The first is routine output offloading, which occurs when perceived knowledge is high and required mental effort is low, such as generating standard correspondence. The second is knowledge-gap offloading, where both effort and knowledge are low, often used for quick informational lookups. The third is ambiguity-resolution offloading, where required effort is high but knowledge is low; workers use the tool to make sense of unfamiliar, complex problems.
Host: And the fourth?
Expert: Augmentation-oriented offloading, occurring when both effort and knowledge are high. In this mode, skilled workers use the chatbot as a sounding board to explore edge cases or refine complex strategies.
Host: That suggests employees are doing more than just shaving minutes off specific chores. They are reallocating cognitive energy across broader tasks, including evaluation and decision-making.
Expert: They are. However, the study identifies a significant organizational risk when offloading occurs under conditions of low knowledge sufficiency. When workers have limited baseline knowledge about a domain, they are structurally less equipped to spot subtle inaccuracies or verify outputs critically.
Host: That is a crucial consideration for management. If teams use generative systems to bypass knowledge building, errors can propagate without detection.
Expert: Exactly. For practitioners, the takeaway is clear: deployment initiatives cannot rely solely on the tool's capability. Organizations must assess employee domain expertise alongside task complexity. High-stakes tasks require that the user possesses sufficient knowledge to independently evaluate the AI's suggestions.
Host: In other words, successful integration requires balancing cognitive assistance with verified competence.
Expert: That summarizes the practical implication well.
Host: Thank you for breaking down the study, Alex, and thank you to our listeners for joining us on A.I.S. Insights — powered by Living Knowledge. Until next time.