Your Newest Teammate Is an Algorithm: Rethinking Team Dynamics When AI Joins the Group

For most of the past two years, AI at work meant opening a separate window, typing a prompt, and copying the result back into your real workflow. That era is ending. AI is moving from a tool you consult to a teammate that acts. Agents now appear in Slack and Teams channels through @-mentions, working side by side with people, contributing context, taking action, and helping move conversations forward.

This is a genuinely new management question. When a new human joins a team, we think carefully about trust, communication norms, and how they will fit the group. We rarely apply that same intentionality when the new member is an algorithm sitting inside the same channel. Yet the early evidence suggests the arrival of AI teammates reshapes how people talk to each other, what they delegate, and even how original their collective thinking becomes.

This piece looks at what the research is showing, and the practical design choices leaders face to capture the gains without quietly hollowing out team cohesion.

From Tool to Teammate

The shift is not hypothetical. Major platforms are now positioning AI explicitly as a working member of the group rather than a utility. Human workers are increasingly collaborating with AI agents on everything from sales support to marketing programs to engineering projects. Vendors describe these systems in frankly colleague-like terms, with one "AI teammate" designed to work natively across tools to analyse data, search, and orchestrate work from start to finish.

The potential is real, but so is the adjustment it demands. There is tremendous upside in this new world of digital labour, including productivity gains, time savings, and increased responsiveness, but also challenges, one of the biggest being employee readiness, because the human workforce must learn how to interact with their new agentic colleagues and to trust them.

That word, trust, is where team dynamics start to matter more than the technology itself.

What the Early Evidence Shows

A large field experiment by Harang Ju and Sinan Aral offers one of the clearest looks so far at what changes when an agent joins the work. Using a platform they developed to study human-AI collaboration, the researchers randomly assigned 2,234 participants to human-human and human-AI teams that produced over 11,000 ads for a think tank. The results are both encouraging and cautionary.

On productivity, the gains are substantial. Human-AI teams produced 50% more ads per worker and higher text quality, while human-human teams produced higher image quality, suggesting a jagged frontier of AI agent capability. In other words, AI lifts some dimensions of quality and lags on others, so the mix of what a team produces shifts rather than simply improving across the board.

More interesting for team dynamics is how people worked together differently. Human-AI collaboration was significantly more task-oriented, with 25% more task-oriented messages and 18% fewer interpersonal messages. People also handed over more of the work: participants delegated 17% more work to AI agents than to human partners and performed 62% fewer direct text edits when working with AI.

This is the quiet trade-off leaders need to notice. The same shift that makes teams faster and more focused also strips out some of the interpersonal exchange that builds relationships, surfaces tacit knowledge, and holds a group together. Efficiency and connection are not the same thing, and optimising hard for one can erode the other.

There is a creativity risk too. The study found that human-AI teams produced more homogeneous outputs, a diversity collapse, as their ads were more self-similar. When everyone leans on the same model, the group can converge on similar ideas and lose the friction that produces genuinely original thinking.

The research community is actively probing where this tips from help to harm. One emerging study notes plainly that organisations are increasingly deploying AI agents as members of teams, which can impact team dynamics and influence team creativity, particularly when multiple AI agents are integrated into teams , and sets out to test how the proportion of AI members in hybrid teams affects team creativity and process satisfaction. The honest answer, for now, is that the balance point depends on design choices leaders make.

The Leadership Task

If AI agents reshape communication, delegation, and creativity, then integrating them well is a leadership and culture task, not just an IT rollout. Three priorities stand out.

Set transparency norms. People behave differently when they know a teammate is an agent. In the field experiment, recognition that the collaborator was an AI moderated the effects, as participants who correctly identified they were working with AI were more task-oriented and more likely to delegate work. Ambiguity about who, or what, produced a piece of work erodes trust. Teams should know when they are working with an agent and what it is permitted to do.

Design for guardrails and oversight. Trust in an agent is not blind faith; it is confidence built on clear boundaries. Leading deployments build in safeguards such as audit trails, data masking, and guardrails that clearly define the boundaries in which agents operate. The practical point for team leaders is that accountability must stay human. An agent can draft, suggest, and act within limits, but a person remains answerable for the outcome.

Protect what remains distinctly human. The most credible voices in this space frame agents as augmentation, not replacement. As Salesforce's chief people officer Nathalie Scardino put it, agents are not about replacing people; they are releasing us from the routine, the bureaucracy, and the noise, while augmenting human productivity, creativity, and purpose. That framing only holds if leaders actively reinvest the time saved into the human side of the work, relationships, judgement, mentoring, and original thinking, rather than simply absorbing it as more output.

Actions to Consider

A few practical steps to integrate AI agents without hollowing out cohesion:

1. Name the agent, and name its role. Make it explicit in your channels when an agent is contributing and what tasks it owns. Treat its onboarding with the same clarity you would give a new hire, an idea explored in our work on setting people up for success from day one.
2. Protect human-to-human time deliberately. If agents are absorbing the task-focused chatter, ring-fence moments for the interpersonal connection that builds trust and cohesion within teams. It will not happen by accident.
3. Guard against idea convergence. Build in deliberate divergence, independent thinking before anyone consults the agent, and structured challenge afterwards, to counter the pull toward homogeneous output.
4. Keep accountability human. Be clear about where the agent's remit ends and a person's judgement begins, especially on decisions that carry risk or affect people.
5. Build the skills to collaborate well. Readiness is the real bottleneck. Invest in helping people learn how to brief, question, and sensibly rely on their agent colleagues, in the same spirit as learning in the flow of work.

Final Thoughts

The arrival of AI teammates is not a threat to human-centred teams, but it is a test of them. The evidence suggests agents can make us faster, more focused, and more productive, while also quietly reducing the interpersonal exchange and creative diversity that strong teams depend on. Both things can be true at once. The outcome is not decided by the technology; it is decided by the norms, boundaries, and intent that leaders design around it.

The question worth sitting with is this: as AI takes on more of the task, what are you deliberately doing to protect the distinctly human work, the trust, the judgement, and the original thinking, that your team cannot afford to lose?

Sources & Further Reading

• Collaborating with AI Agents: A Field Experiment on Teamwork, Productivity, and Performance (Ju & Aral, arXiv)
• A New Collaborative Workforce of Humans and Agents (Salesforce)
• Human-AI Partnership: Salesforce's Blueprint for the Future Collaborative Workplace
• Salesforce Publishes the Workforce Innovation Playbook: The 4Rs to Building the Human + Agent Workforce
• Agentforce Coworker: Your Autonomous AI Teammate (Salesforce)
• Agentforce for Employees (Salesforce)
• Human-AI Collaboration in Hybrid Teams: Implications for Team Creativity and Process Satisfaction (AMCIS 2025 Proceedings)