Picture two bank employees sitting at neighboring desks. One follows a predictable script all day, checking boxes on standardized forms and handling the same customer requests in the same order. The other juggles a shifting mix of problems, improvising solutions and pulling from experience that no manual fully captures. Now imagine their company just announced it is rolling out artificial intelligence tools. Which employee is more likely to feel energized at work, and which is more likely to quietly disengage?
That question sits at the heart of a new investigation published in the Journal of Business Research. The research examined how the routine, scripted nature of a person’s job shapes their sense of vitality, learning, and willingness to share knowledge with coworkers, and how anxiety about AI replacement changes that picture. The short answer: employees with highly scripted jobs reported less thriving overall, and fears about AI taking their jobs interrupted the knowledge-sharing behaviors that typically help workers flourish.
The question behind the research
Companies are integrating AI into workplaces at a rapid pace. IBM has publicly discussed replacing thousands of roles with AI tools, and retraining displaced workers could cost roughly $24,800 per person in the United States alone, according to figures from the World Economic Forum cited by the authors. Yet researchers have done relatively little to sort out a practical puzzle: why do some employees seem to flourish when AI arrives in their workplace, while others withdraw?
Mai Nguyen of Griffith University in Australia, working with colleagues Lars-Erik Casper Ferm at the University of Western Australia and Liem Viet Ngo at the University of New South Wales, suspected the answer had something to do with how programmable a person’s tasks are. A programmable task is one that follows clear rules and steps, the kind of work an AI system or algorithm can most easily replicate. Data entry is highly programmable. Leading a creative team through an ambiguous strategic problem is not.
The researchers drew on a framework called the Socially Embedded Model of Thriving, which proposes that people flourish at work when they have room to act independently, learn new things, and interact meaningfully with coworkers. The model defines thriving as a combination of vigor (feeling energetic and alive) and learning (feeling you are getting better at what you do).
Key concepts worth defining
Two ideas drive the research. The first is information asymmetry, which simply means that one person holds more or better information than another. In a workplace, an employee who handles varied, non-routine situations often accumulates unique knowledge that colleagues do not have. An employee following a rigid script tends to build up less of this private expertise because the procedures are standardized.
The second concept is online knowledge sharing, which the researchers broke into three behaviors: donating knowledge (actively giving information to coworkers), collecting knowledge (actively seeking it out), and lurking (passively reading and absorbing information shared on digital platforms without contributing). Lurking gets less attention in workplace research, but the authors argue it matters because roughly 90 percent of social media users are passive consumers of content.
How the researchers tested their ideas
The team ran two experiments with a combined total of 550 participants recruited through Prolific, an online research platform. All participants were required to have worked as frontline employees within the past year.
In the first experiment, 205 participants were randomly assigned to read one of two scenarios. In the high-programmability version, they imagined being a bank frontline employee who performed the same tasks every day and followed clear company-provided steps for every customer case. In the low-programmability version, they imagined doing different tasks each day and handling cases with no preset procedures. Participants then completed surveys measuring their expected vigor, learning orientation, sense of holding unique job knowledge, and likelihood of donating, collecting, or lurking for knowledge online.
The results showed a clear pattern. People imagining the highly scripted job reported lower learning orientation, lower vigor, less unique job knowledge, and less engagement in all three knowledge-sharing behaviors compared to those imagining the varied job. Statistical analysis revealed a chain: highly programmable tasks were linked to lower information asymmetry, which was linked to less online knowledge sharing, which in turn was linked to lower thriving.
The second experiment added a twist. The 345 participants were randomly assigned not only to a high or low programmability scenario but also to a high or low AI-induced job insecurity scenario. Those in the high insecurity condition imagined their company had adopted AI tools and worried about losing their jobs. Those in the low insecurity condition imagined the same AI rollout but felt secure.
This experiment added a wrinkle to the earlier findings. When employees felt low AI-induced job insecurity, the chain connecting task programmability to thriving through information asymmetry and knowledge collecting (or lurking) held up. When employees felt high AI-induced job insecurity, the chain broke down. In other words, worry about AI appeared to interrupt the knowledge-sharing pathway that normally helps employees thrive.
What the findings mean for managers
The authors offer several suggestions for businesses integrating AI. First, they recommend redesigning highly scripted jobs to include some variety, problem-solving, or creative challenge, since routine work appears to drain learning and vigor regardless of AI. Second, they suggest transparent communication about why AI is being introduced and what it will and will not replace. Fear of being made obsolete, the findings suggest, can short-circuit the collaborative behaviors that make teams work well.
Third, they point to reskilling and upskilling programs as a way to reframe AI from a threat into a tool for professional growth. And fourth, they note that monitoring employee well-being through the AI transition matters, since earlier research cited in the paper links workplace AI to mental health strain.
Caveats to keep in mind
A few limitations are worth noting. The experiments used hypothetical scenarios rather than observations of actual workers on the job, and participants were based in the United States, so the findings may not translate directly to other cultures or industries. The data also relied on self-reports from a single point in time, which the authors acknowledge by calling for longitudinal studies that track employees over months or years as AI becomes part of their daily routines. Finally, the studies focused on frontline employees at a bank; healthcare, education, manufacturing, and other sectors may experience AI differently.
Still, the research offers a useful way of thinking about why AI’s arrival lands so differently for different workers. Scripted jobs were already linked to lower thriving. Add anxiety about AI replacement, and the informal knowledge exchanges that might help employees adapt start to falter.
