When the AI Joins the Team: How AI Teammates Reshape Human Collaboration

We keep asking the wrong question about AI in teams. Not "How good is the AI?" but "What does an AI teammate do to the team around it?" In this Ignite talk, I share findings from studies on the LaLALab TRAIL platform, where we swapped one human seat for a generative-AI teammate and watched what changed. The results reveal a paradox. The AI spoke the most in every team yet contributed the fewest new ideas: talkative, but shallow. Humans pulled back, with lower belonging, lower perceived status, and more AI airtime tied to students feeling less valued (r = –0.52). And yet the same AI teammate helped groups hold harder, more reasoned moral stances under pressure, acting as "moral cover" by voicing socially costly positions humans wouldn't. An AI teammate is never neutral. It reshapes who talks, how connected people feel, and how teams decide, for better and worse at once. The takeaway for the learning sciences: study the team, not just the tool, and design AI teammates for the social fabric they leave behind.