Forming, Storming, Norming, Performing: Leading Teams When One Member Is an Algorithm

23 hours ago

5 min read

For six decades, Bruce Tuckman's model of group development has given leaders a reliable map for a messy process: turning a collection of individuals into a functioning team. Teams form, they storm, they eventually norm, and if they are fortunate, they perform (Tuckman, 1965). The model has proven durable because it describes something true about human groups regardless of industry or era. What has changed is the composition of the team itself. A growing share of leaders now manage groups that include people as well as AI agents capable of planning, adapting, and contributing to shared goals (Lou et al., 2025). Applying Tuckman's stages to this new configuration offers a practical way to think about what changes, and what does not, when a machine joins the roster.

A Model Built for People, Now Tested by Machines

Tuckman's original 1965 paper synthesized fifty studies of group behavior into four stages: orientation, conflict, cohesion, and functional role-relatedness, more memorably known as forming, storming, norming, and performing (Tuckman, 1965). A 1977 revision added a fifth stage, adjourning, to account for how groups disband once their purpose is fulfilled (Tuckman & Jensen, 1977). The model's staying power comes from its simplicity and its accuracy, and its trajectory from an HR practitioner favorite into a widely cited academic framework is itself well documented (Bonebright, 2010). Teams really do need time to establish trust, work through friction, and settle into shared norms before they reach full productivity. Nothing about the introduction of AI into a team changes that underlying psychology. What changes is who, or what, the team is establishing trust with.

A Fifth Teammate Who Never Sleeps

Researchers studying human-AI teaming describe a shift already underway in many organizations. AI systems are no longer confined to the role of tool, quietly executing instructions in the background. They are increasingly functioning as active collaborators that learn, adapt, and operate with a degree of autonomy inside shared workflows (Lou et al., 2025). This shift requires new thinking about interaction protocols, delegation, and how responsibility gets distributed between human and machine contributors (Lou et al., 2025). For a leader applying Tuckman's framework, that means treating the AI system as a team member whose onboarding, integration, and performance deserve the same deliberate attention given to a new hire.

Forming: Defining the Agent's Role Before the Work Begins

The forming stage has always been about orientation. Team members test the boundaries of the group, look to the leader for direction, and try to understand how they fit. When an AI system joins a team, that same orientation needs to happen deliberately, not by default. In practice, that means leaders deciding up front which decisions the AI supports, which decisions it should never make unilaterally, and how its output will be verified before it becomes team knowledge. Skipping this step moves the confusion downstream, typically into the storming stage, where it is more expensive to resolve.

Storming: Conflict Now Includes a Trust Problem

Every team hits a period where politeness gives way to friction. With a human-AI team, that friction often centers on a different kind of conflict: whether the AI's output can be trusted, and how much oversight it actually requires. Current research on human-AI teaming identifies exactly this tension as one of the field's central gaps, pointing to the ongoing difficulty of aligning AI systems with human values and objectives, and the tendency of teams to underutilize the AI's genuine capabilities even once it has proven itself (Lou et al., 2025). Left unmanaged, this friction shows up as a new source of workplace strain. A recent Harvard Business Review study found that intensive oversight of AI tools drives measurable cognitive fatigue among employees, a condition researchers call AI brain fry (Bedard et al., 2026). Leaders who recognize storming as a normal, necessary stage, rather than a sign that the AI integration has failed, are better positioned to work through it deliberately.

Norming: Building a Shared Mental Model

Teams reach the norming stage once they develop shared expectations about how work gets done. For human-AI teams, this means building what researchers call a shared mental model: a common understanding, held by both the human and the AI system, of goals, roles, and the standards output must meet (Lou et al., 2025). In practice, this looks like documented prompts, agreed-upon review checkpoints, and a clear record of which tasks the AI now owns outright versus which remain under close human review. Norming is also where trust gets calibrated rather than assumed. Teams that skip this step tend to oscillate between over-trusting AI output and second-guessing every line it produces, neither of which reflects a mature working relationship.

Performing: When the Combination Outperforms Either Alone

The payoff for working through the earlier stages is a team that performs at a level neither its human nor its AI members could reach independently. This is the central promise of human-AI joint cognitive systems: an approach that raises joint performance beyond what either humans or AI achieve on their own, while working around the known limitations of each (Xu & Gao, 2024). Reaching that point is the product of a team that has genuinely formed, stormed, and normed around its newest member, rather than adding a chatbot to an existing workflow and calling it a transformation.

What Leaders Should Take From This

Tuckman's model was never meant to be applied once and forgotten. Teams reform every time membership changes, and adding an AI agent to a workflow counts as a membership change, even if no one updates the org chart. Leaders who name the stage their team is actually in, rather than assuming everyone has already arrived at performing, will spend less time firefighting avoidable conflict and more time building the kind of shared understanding that lets human judgment and machine capability genuinely compound.

References

Bedard, J., Kropp, M., Hsu, M., Karaman, O. T., Hawes, J., & Kellerman, G. R. (2026). When using AI leads to “brain fry.” Harvard Business Review Digital Articles.

Bonebright, D. A. (2010). 40 years of storming: A historical review of Tuckman's model of small group development. Human Resource Development International, 13(1), 111–120. https://doi.org/10.1080/13678861003589099

Lou, B., Lu, T., Raghu, T. S., & Zhang, Y. (2025). Unraveling human-AI teaming: A review and outlook. arXiv. https://doi.org/10.48550/arXiv.2504.05755

Tuckman, B. W. (1965). Developmental sequence in small groups. Psychological Bulletin, 63(6), 384–399.

Tuckman, B. W., & Jensen, M. A. C. (1977). Stages of small-group development revisited. Group & Organization Studies, 2(4), 419–427.

Xu, W., & Gao, Z. (2024). Applying HCAI in developing effective human-AI teaming: A perspective from human-AI joint cognitive systems. Interactions, 31(1).