In our previous articles, we have discussed how artificial intelligence (AI) is being integrated into modern businesses, the demands it creates, and the importance of fostering an AI mindset so that businesses and organizations can take advantage of the opportunities it offers.
In this article, we focus on a dimension that has received relatively little attention, despite its pivotal importance for the future of work: what happens in work teams when AI is integrated into their daily routines. The arguments we present are drawn from our ongoing research project.
From the individual to the team: why it matters
Most discussions about AI in the business world focus on what it can do for each individual employee: how it will increase productivity, how it will automate tasks, how it will help write a text faster or provide useful predictions.
But this picture is limited. Modern work is largely collaborative. Teams of surgeons, teams on construction sites, sales teams, innovation teams, teams in human resources departments.
Their value stems from what organizational scientists call collective intelligence: the ability of members to share their knowledge, coordinate with one another, and adapt to situations that no single individual could handle alone.
This collective intelligence is built gradually through two distinct processes. The first is called the transactive memory system and refers to the group’s knowledge of who knows what: who is the expert in which field, whom we turn to when facing a specific challenge, and on whom we rely when we need a second opinion.
The second process is called diligent interaction (heedful interrelating). This refers to the ability of team members to understand how their own work fits into the whole, to anticipate their colleagues’ actions, and to coordinate with them.
Teams in intensive care units, flight crews, and high-performing family businesses all rely on these unspoken forms of coordination. Both processes are built over time, through shared experiences and daily interaction.
Why Teams Embrace TN
Modern teams face increasingly intense pressures. First, they deal with information overload: the volume of data they must process exceeds human capacity. Second, they face coordination challenges when working remotely, across different countries, or in hybrid work models. Third, they experience socio-emotional pressure related to building trust and psychological safety among their members.
Fourth, in physically and sensorially demanding environments such as mines, construction sites, or agricultural facilities, they face physical and sensory overload. In all these cases, AI comes to the rescue.
The integration of AI tools is an expression of what we call digital resilience, the ongoing process through which teams use digital technologies to reshape what they can do under pressure.
AI integrated into teams currently takes two dominant forms. The first is that ofAI agents, autonomous digital machine learning systems. These systems often take the form of AI coaches in startups that support investment decisions in venture capital funds, or even AI agents that process clinical data and suggest diagnoses to medical teams.
The second form is what is often referred to as Physical AI. Here we are talking about robotic systems and autonomous machines that perform physical tasks in manufacturing environments. Autonomous vehicles on construction sites, robots milking cows in dairy farms, automated systems in mines, sensor-equipped equipment on farms.
Embedded AI removes workers from dangerous tasks and extends sensory monitoring beyond human limits.
The phenomenon of progressive encasement
The integration of AI into teams, however, is not a neutral process. It creates a phenomenon that Hinds and von Krogh (2024) call progressive encapsulation (progressive encapsulation). As AI absorbs functions traditionally performed by team members among themselves, the opportunities through which collective intelligence was built are also absorbed. The team’s transactional memory—that invaluable “who knows what”—begins to shift from people to algorithms.
The careful interaction, the silent coordination born of daily interactions, rarely has the chance to develop. Recent studies of teams incorporating AI systems report reduced communication among members, lower collective knowledge, and homogenization of decisions. Interaction among people decreases, while each person’s interaction with AI increases.
Progressive encapsulation is progressive because AI systems are constantly learning and updating themselves, performing more and more tasks with minimal human supervision.
The way in which they reach their conclusions, in turn, becomes increasingly opaque. In business terms, this means that a team may gradually lose its ability to function as a team, without the managers themselves realizing the transition. The team continues to appear productive—sometimes even more productive—as long as the environment remains stable.
However, when an unforeseen crisis arises—a situation outside the algorithm’s training data—the team may be unable to respond collectively, as relationships and mutual understanding among members have weakened.
The Necessity of Anchoring Procedures
In our research framework, we argue that a team’s digital resilience in the age of AI requires two parallel processes. The first is the process of resourcing work. These are the processes through which the team uses the digital environment as a computational and memory resource to meet the demands of its work. The second and equally important process is what we refer to as the anchoring (anchoring work).
This refers to the relational processes through which the group maintains and readjusts its identity as a group, despite the progressive encasement caused by the very use of TN. In the teams we have studied, we see that anchoring can be achieved through various forms and processes under different conditions.
Therefore, we urge managers to establish specific anchoring procedures, such as regular reflection meetings where members discuss TN proposals, voice disagreements, and negotiate joint decisions.
What this means for managers and businesses
The transfer of functions to the TN is often presented by technology providers as a clear benefit: cost reduction, speed, and accuracy. Our research suggests that managers need to broaden their perspective and ask themselves: what opportunities for collective learning are lost when AI takes over a function?
Which relationships weaken when team members turn to the algorithm instead of their colleagues? What are the new rituals, roles, and routines that must be cultivated to maintain the team’s identity?
These questions are practical management issues with long-term consequences. Organizations across all sectors are adopting TM systems faster than research can explain the consequences, and many of the related decisions are difficult to reverse. When a transaction memory system is lost, it takes years to rebuild.
When team members cease to know one another, rebuilding relationships requires a substantial investment. Our call to managers is to view the integration of AI as an opportunity to reshape collective work practices, with people and interpersonal relationships at the center.
* Konstantinos Zopounidis (photo, left) Professor, Academic, Knight of the Order of the Academic Palms, Honorary Doctorate, Aristotle University of Thessaloniki, Technical University of Crete
** Angelos Kostis, Associate Professor, Umeå University, Sweden