What Happens When AI Becomes Part of the Team?

Retailers increasingly rely on AI tools for marketing, creative work and back office operations. This shift in work structure leads to a pressing question: how do employees react as AI becomes a part of their team? With support from the Lundgren Retail Collaborative, Dr. Agrim Sachdeva is studying what happens when AI agents become visible team members; how the mix of humans and AI shapes whether people still feel like they belong on the team; and when multiple AI capabilities are needed, what can organizations do to support healthy teams.

Q: What was the central question behind this research?

A: We wanted to understand what changes when AI agents stop feeling like background tools and start appearing as named members of a team. Does having more AI agents affect the feeling that “this is my team”?  Does having more AI agents cause existing team members to see a divide between the humans and the AI, and how does it affect the human team members desire to keep working with that team. 

Q: Why does this matter for retailers?

A: Retailers use AI in areas such as creative development, product content, merchandising, customer support and internal operations. As these workflows move from one person using one AI tool to groups working with multiple AI agents, leaders need to think about more than efficiency— they need to consider how these organizational changes will affect employee morale and productivity. Even capable AI systems can create friction if the team structure makes people feel outnumbered, split into “us versus them,” or less connected to their work.

Q: What kind of task did participants work on in the study?

A: The study used a realistic creative-work scenario. The specific task was to create retail ad copy. The team needed to combine text, an image and slide design into one finished piece. That made the setting familiar to retail and marketing work, where teams often combine product information, visuals and brand templates. The study looked at several different team configurations, ranging from all-human teams to balanced teams, to teams with one human and three AI agents.

Q: What were the main findings? 

A: First, when there were the same number of humans and AI agents on a team, people were more likely to see a faultline-a split between the human side and the AI side. Second, when the team had more AI agents than humans, people felt less identification with the team overall. It seems that scaling AI can weaken the sense of shared team identity and undermine the perception that the team worked hard as a collective or invested effort in the task. Importantly, the pattern was not simply “ AI is bad.” A single visible AI agent in a human-majority team did not create the same problem. 

Q: Did AI-heavy teams produce better or worse work?

A: The outcomes were the same for the human vs. AI-heavy teams. But people were less satisfied with both the process and the final outcome when AI agents outnumbered human team members and thus they did not feel like part of a cohesive team. Even though the outcome was similar, participants were not as happy with it when fewer humans were involved.

Q: What should retail leaders take away from this research?

A: Do not design AI workflows only around how much work the AI can take on. Design them around how the team will feel to the humans inside it. Keep visible AI agents in the minority when possible. Avoid leaving one human alone with one AI agent. Avoid human-AI parity if it makes the team feel split. And when multiple AI capabilities are needed, consider a single visible AI interface that connects to specialized agents in the background. 

Background

Previous research on teams and technology has largely focused on support systems, such as decision-support tools that help or augment human work. What is different now is that AI agents are beginning to do more than support people. They can automate tasks, take on specialized roles and exercise agency within the workflow.

As these agents gain autonomy, it becomes more important to understand how interacting with AI agents changes the human experience. We need to consider how people perceive their own role in the process, their relationship to the tools they are using and their connection to the final output being created.

It is important to not lose sight of the human. Because even if the outcomes are going to be the exact same, there's something special about working with people, and we try to answer what that something is. In this case, team identification, the feeling that “this is my team”, is the human factor that helps explain why team design matters. The practical point is that performance and experience are both important. A workflow can produce high level outcomes and still fail if people do not want to stay engaged with it. One promising design solution is to keep a single AI agent visible to the team and let that agent coordinate other AI capabilities behind the scenes. That setup can preserve the benefits of multiple AI specialties without making the team feel crowded with AI agents. However, there was an important exception: a one-human, one-AI pairing still created problems. People need at least one other human peer to help anchor the sense of team membership.