An AI assistant joins a project meeting. It summarises customer feedback, recommends priorities and suggests who should take responsibility for each task. The manager accepts its recommendations. Team members begin organising their work around them.

At that point, the organisation has changed more than its software. It has changed how information becomes a decision, how work is allocated and whose judgement carries weight.

The Journal of Organizational Behavior special issue, “Is Our Future Colleague Even Human?”, examines these changes. Its research offers leaders a useful starting point: define the AI’s contribution, establish its authority and evaluate how it affects the people working with it.

Define what “team member” means

Ulfert et al. (2026) distinguish between AI performing tasks, coordinating work and influencing relationships within a group. They also distinguish how much initiative remains with humans, is shared with AI or is delegated to the system.

These distinctions make implementation decisions clearer. An assistant that drafts a report needs different oversight from a system that allocates work. A chatbot’s friendly manner tells leaders little about either system’s authority.

Before introducing AI, specify what it may recommend, what it may execute and which decisions require human approval. Assign a named person responsibility for reviewing its contribution. Revisit these arrangements when the system’s capabilities or role changes.

Make information quality visible

Zercher et al. (2026) studied 81 teams undertaking a decision task. Teams supported by AI holding comprehensive information made more accurate decisions than human-only teams. Teams working with partially informed AI did not outperform them.

The findings come from a controlled student experiment, so they cannot establish that the same results will occur in an executive team. They nevertheless highlight a practical implementation question: does the system have the information needed for the contribution expected of it?

For example, a staffing recommendation may omit knowledge about an employee’s development goals or an upcoming client commitment. Leaders should ask teams to identify what the AI knows, what it lacks and which assumptions require checking. An incomplete recommendation can still sound convincing.

Separate trust from agreement

Employees’ willingness to follow AI advice should not become the main measure of success.

Smith et al. (2026) investigated how people aligned their decisions with AI recommendations. Acceptance depended on factors including agreement between human and AI advice, while the effects of optional advice varied across tasks. These findings explain patterns of reliance; they do not establish that greater reliance always improves decisions.

Trust also develops through colleagues. Erengin et al. (2026) found that another human teammate’s trust in AI influenced participants’ own trust perceptions and behaviour, particularly when that colleague was considered trustworthy.

Leaders should therefore demonstrate how they assess recommendations. Explain why advice was accepted, what evidence was checked and what would justify rejecting it. Employees need permission to challenge an AI-supported decision even when a respected manager endorses it.

Examine the assumptions built into the interface

Names, voices and avatars can shape employees’ expectations before they have tested a system’s performance.

Mirowska and Arsenyan (2026) found that human-likeness, gender cues and workplace context influenced initial reactions to a proposed AI teammate. Their vignette studies measured initial judgements rather than sustained collaboration.

Tu et al. (2026) found that participants generally perceived AI teammates as more competent but less warm than human teammates. Gendered characteristics changed those perceptions in different directions: female-like characteristics could improve perceived warmth while reducing the perceived competence advantage.

These are findings about people’s perceptions, not evidence that a gendered system performs better. Tu et al.’s experiments largely assessed cooperation intentions, while their field study used self-reported cooperation.

Leaders should question design choices that reproduce familiar stereotypes. Test whether an interface helps employees understand the system’s capabilities and limitations. Avoid treating comfort, friendliness or apparent confidence as proof of reliability.

Prepare the whole team to work differently

Individual proficiency does not ensure effective coordination. Employees may each know how to use AI while remaining unclear about who checks its work or resolves conflicting recommendations.

Lancaster et al. (2026) interviewed 22 competitive eSports athletes about training with AI. Participants’ needs differed with expertise, and the findings highlighted communication, role awareness and coordination alongside task skills. The specialised sample limits direct transfer to other workplaces.

A practical response is to rehearse a real team task before deployment. Agree who supplies information, who reviews outputs and how disagreements are resolved. Include a deliberately flawed recommendation so the team can practise detecting and correcting an error.

Training should also match experience. New users may need help understanding system behaviour; experienced staff may need opportunities to examine how AI changes established professional judgement.

Protect employees’ ability to influence their work

Klonek and Parker (2026) analysed more than 2,700 social-media posts. Greater expressed AI control was associated with greater expressed stress, while human–AI teamwork processes were associated with less stress. The observational design cannot establish causation.

The study gives leaders a reason to investigate employees’ experience during implementation. Can they override an unsuitable recommendation? Does checking AI output add unrecognised workload? Have faster outputs produced unrealistic delivery expectations?

Ulfert et al. (2026) identify accountability and employee voice as important criteria for designing AI’s organisational role. Involve employees in deciding how the system should operate, and provide a clear route for raising concerns. A nominal approval step offers little protection if employees lack the time, information or authority to question the output.

Evaluate the result at team level

Tsai et al. (2026) propose a theoretical model connecting human–robot collaboration with organisational value. Their argument directs attention to social dynamics and task interdependence, but does not demonstrate a guaranteed financial return from AI adoption.

Start with one defined workflow and establish its baseline performance. During the pilot, examine decision quality, errors, coordination, verification time and employee workload. Check whether people retain the capability to recognise mistakes and complete essential work when the system is unavailable.

Expand the system’s role when the evidence supports doing so. Giving AI greater responsibility should be an explicit leadership decision, with clear ownership of the consequences.

References

Erengin, T., Briker, R., & de Jong, S. B. (2026). You, me, and the AI: The role of third-party human teammates for trust formation toward AI teammates. Journal of Organizational Behavior, 47(8), 1397–1422. https://doi.org/10.1002/job.2857

Klonek, F., & Parker, S. (2026). Does AI at work increase stress? Text mining social media about human–AI team processes and AI control. Journal of Organizational Behavior, 47(8), 1423–1437. https://doi.org/10.1002/job.70000

Lancaster, C. M., Flathmann, C., Hsu, J., McNeese, N. J., O’Neill, T. A., & Salas, E. (2026). Preparing for tomorrow’s teamwork: Insights from eSports on how human expertise shapes training needs for AI-integrated work. Journal of Organizational Behavior, 47(8), 1438–1458. https://doi.org/10.1002/job.70036

Mirowska, A., & Arsenyan, J. (2026). The A(I) team: Effects of human-likeness and conformity to gender stereotypes on initial trust and willingness to work with an AI teammate. Journal of Organizational Behavior, 47(8), 1288–1312. https://doi.org/10.1002/job.70009

Smith, A., van Wagoner, H. P., Keplinger, K., & Celebi, C. (2026). Navigating AI convergence in human–artificial intelligence teams: A signaling theory approach. Journal of Organizational Behavior, 47(8), 1338–1368. https://doi.org/10.1002/job.2856

Tsai, C.-Y., Marshall, J. D., Eckardt, R., Dionne, S. D., Jung, M. F., Hou, Y. T.-Y., & Yang, B. (2026). Human capital robotic integration and value creation for organizations. Journal of Organizational Behavior, 47(8), 1459–1475. https://doi.org/10.1002/job.70044

Tu, Y., Li, J., Chen, J., Li, C.-J., & He, W. (2026). When AI becomes my teammate: Unpacking how employees perceive and collaborate with gendered AI teammates. Journal of Organizational Behavior, 47(8), 1313–1337. https://doi.org/10.1002/job.70038

Ulfert, A.-S., Georganta, E., & Grote, G. (2026). Is our future colleague even human? Advancing human–AI teamwork from an organizational perspective. Journal of Organizational Behavior, 47(8), 1277–1287. https://doi.org/10.1002/job.70125

Zercher, D., Jussupow, E., Benke, I., & Heinzl, A. (2026). How can teams benefit from AI team members? Exploring the effect of generative AI on decision-making processes and decision quality in team–AI collaboration. Journal of Organizational Behavior, 47(8), 1369–1396. https://doi.org/10.1002/job.2898