A department head recently described her team as "eleven people and four agents". The agents handle intake, first-line customer queries, reporting and scheduling. The people handle judgement, relationships, exceptions and improvement. Her question to Oxford Management Centre was simple: nobody trained me for this. She is not alone. Managing teams that combine people and AI agents is becoming an ordinary part of management, and it requires skills that traditional leadership development has not yet caught up with. This article sets out what those skills are and how managers can build them.
Managing human and AI agent teams is the practice of leading a workgroup in which some tasks are carried out by autonomous AI agents and others by people. It requires managers to design roles across both, set the limits and oversight for agents, maintain human judgement and accountability, and lead people through a change in the nature of their work.
What changes when AI agents join a team?
- The work is redistributed. Repeatable, multi-step tasks move to agents. Human work concentrates on judgement, relationships, exceptions and improvement.
- Oversight becomes a core task. Someone must monitor agent output, handle escalations and adjust agent instructions. That is management work, and it needs to be assigned.
- Accountability must be explicit. An agent cannot be accountable. The manager, or a named person in the team, owns each agent's outcomes.
- Performance management expands. Managers now assess people, agents and the combination, and the metrics differ.
- Trust must be managed in both directions. People need to trust that agents are reliable and that their own roles are secure; the organisation needs to trust that people are still exercising judgement.
What leadership skills matter most?
- Work and role design: Deciding which tasks go to agents, which stay with people and where the handovers sit. Poor design produces frustration and error; good design produces roles that people find more rewarding.
- Instruction and oversight: Writing the operating rules for agents, setting approval thresholds and reviewing output. Managers who cannot do this will depend on others to run their own team.
- Judgement preservation: Ensuring people continue to think critically rather than accept agent output by default. This is a cultural skill, built through questioning, review routines and explicit expectations.
- Change leadership: Communicating honestly about how roles will change, involving people in redesign and supporting retraining. Teams that experience agents as something done to them resist; teams that shape the change adopt it.
- Measurement: Defining what good looks like for the combined team: throughput, quality, exception rates, customer outcomes and staff engagement, not just cost.
KEY TAKEAWAY
Managing teams of people and AI agents is a leadership skill, not a technical one. It rests on role design, oversight, preserving human judgement, leading change honestly and measuring the combined team's performance. Managers who build these skills early will shape how their organisations adopt agentic AI.
How should organisations prepare their managers?
Three steps are proving effective:
- Give managers a grounding in how AI agents work and where they fail, in management rather than technical terms.
- Train them in work redesign and oversight before agents are deployed in their teams, not afterwards.
- Build agent supervision into management role descriptions, performance expectations and leadership development.
Oxford Management Centre addresses these through its leadership and management courses and its Agentic AI Governance and Control course, and increasingly through customised in-house programmes for organisations introducing agents across several teams.
Frequently Asked Questions
-
What does it mean to manage a human and AI agent team?
It means leading a workgroup where AI agents carry out some tasks autonomously and people carry out others, and taking responsibility for role design, agent oversight, accountability, judgement and change leadership across both.
-
Can an AI agent be held accountable?
No. Accountability sits with a named person, usually the manager or a designated team member who owns the agent's outcomes and monitors its performance.
-
What skills do managers need to lead teams with AI agents?
Work and role design, writing and adjusting agent instructions, oversight and escalation handling, preserving human judgement, change leadership and outcome measurement.
-
How do you keep people engaged when agents take over parts of their work?
By involving them in redesigning roles, communicating honestly about impact, investing in retraining and ensuring the remaining human work is meaningful and recognised.
-
How do you measure the performance of a team that includes AI agents?
Use combined metrics covering throughput, quality, exception rates, customer outcomes and staff engagement, and review agent performance as part of regular team performance management.
