AI adoption continues at breakneck speed, but a major gains gap is beginning to take shape: research summarized in Microsoft’s New Future of Work Report 2025 suggests AI consistently boosts individual productivity, while team-level gains lag.
The reason is simple: Teams aren’t just tasks, human interactions take turns, and have shared context, varying motivations, and status across their shared goals. The future of AI design is in collaboration so AI can help teams get better together. Through Microsoft’s analysis, several key themes impacting the advancement of collaborative AI emerge:
Adoption & usage: Gaps are related to trust, psychological safety, and worker voice in design
- Adoption hinges as much on social norms as executive mandates: employees watch what leaders and peers do, and they hesitate when tools feel like surveillance or “efficiency over quality.”
Impact on work
- Beware of “workslop:” Generative AI is associated with meaningful time savings, but sometimes produces polished-looking, low-substance output which can cancel gains when it flows through teams and requires rework.
Human–AI collaboration: from one-shot answers to co-creation. People increasingly use AI in multiturn, iterative ways.
- AI can sound confident without actually understanding. Better collaboration comes from AI that asks clarifying questions, exposes limits, and supports selective delegation (routine to AI; accountability stays human).
AI for Teamwork
- Process beats persona. Researchers see two routes: AI that supports specific team processes (like information sharing) versus end-to-end systems trained on team outcomes. Both are active frontiers.
- Facilitation is the early win. AI meeting facilitators can increase information sharing (one study found +22%) and are often perceived positively, yet they may not change final decisions. Shaping outcomes requires targeting decision processes with structured deliberation.
- AI roles should flex by scenario. Teams benefit from different “AI teammates”: coordinator, creator, perfectionist, doer—or a devil’s advocate to improve reliance and amplify minority viewpoints; or a mediator to synthesize perspectives and reduce polarization.
A proactive AI model changes everything: When AI is developed to be more proactive, it can become more like a teammate, capable of initiating conversation, suggesting next steps and adapting based on the conversation flow. When AI gathers missing info and chooses the right moment to speak, it feels more like a teammate—but it raises new needs for timing, turn-taking, and social acceptance.
