Short answer
When an AI agent joins the team, the manager's job shifts from handing out work to describing it: defining the agent's task, its permission limits and the steps that need human approval, and owning the outcome. The agent prepares and a person decides; above all, decisions that affect people's lives stay with people.
Highlights
- The agent prepares, a person decides: steps that spend money, go outside the company or affect someone's life need human approval.
- The manager's new duties: writing the task definition, setting permission limits and owning the agent's work.
- The skill that matters most in the age of agents is not using the tool but questioning its output.

5 min readAuthor: UNIT Journal EditorEditor-in-chief: Uğur Deniz İlhan
Not long ago, AI was a tool the team used. Today it is increasingly becoming a "member" working inside the team: it prepares reports, monitors campaigns, gives the first answer to customer questions and even completes purchases on its own. This shift is less about technology than about management. One of us manages AI agents, the other works on people and leadership; in this piece we bring the two perspectives together.
What does the agent take on, and what stays with people?
An AI agent breaks a goal into steps, uses tools and tries to carry the work through to the end. It is very fast at work that repeats, has clear rules and whose mistakes can be undone. But it does not carry context, relationships or responsibility.
Our line is simple: the agent prepares, a person decides. Every step that spends money, goes outside the company or affects someone's life passes a person's approval. In HR the line is even sharper. An agent can pre-screen large numbers of applications, organise interview notes and report training needs. But the hiring decision, the performance conversation and difficult feedback belong to someone who can look the other person in the eye. Employees want to know there is a human behind decisions about them; that trust cannot be handed to an agent.
What are the manager's new responsibilities?
The first thing that changes in a manager's job is the move from "handing out work" to "describing work". For years, many managers got away with incomplete instructions by relying on an experienced employee's intuition. An agent fills that gap with guesses, not intuition. So new responsibilities land on the manager's list:
- Writing the task definition: what will the agent do, what will it not do, and when must it stop and ask?
- Setting permission limits: which data, accounts and budget can it reach? Start with the least privilege and widen it as trust builds.
- Owning the responsibility: "the AI did it" is not an answer. The agent's work needs an owner, and that role is the AI Agent Manager.
These are really what a good manager already does with people: clear expectations, the right authority, open accountability. Managers who can set expectations in writing, measurably and with clear limits get better results from both their agents and their teams.
How does the team work with an agent?
The hardest part of the change is not the technology, it is anxiety. "Will I lose my job?" is in the air even when nobody asks it. The way to manage it is not silence but talking early and openly: explaining from the start what the agent will do, what it will not do and how the team's work will change.
- Run the agent in suggestion-only mode first, and let the team compare its results with their own work.
- Log every step the agent takes, so that when something goes wrong you look for the cause, not a culprit.
- Make someone on the team the agent's teammate: the person who gives it feedback and updates its instructions.
In a well-built setup the team spends less time collecting reports and filling in checklists, and more time on interpretation, customers and decisions. Bringing employees in as the people who design and improve the agent, rather than leaving them outside the process, turns resistance into ownership.
Which skills matter in the age of agents?
In hiring and development, the focus is moving from "can they use the tool?" to "can they question the tool's output?" Describing work clearly, answering "did it get better?" with data, curiosity, speed of learning and taking responsibility outlast technical knowledge. For managers there is a new one on top: organising people and agents in the same team so that each completes the other's work.
While AI transforms some jobs, it is also opening entirely new roles: AI Agent Manager, AI product owner, automation and data governance specialists. Companies that define these roles well and fill them with the right people will go through the change faster and with less pain.
Final word
An AI agent starts as a software project and continues as a management matter. The companies that win will not be those that build agents well, but those that manage them well. That takes the people who know the technology and the people who know people sitting at the same table. This piece is an example of that table.
About the authors
Uğur Deniz İlhan is co-founder of UNIT İstanbul. As AI Agent Manager and Growth Hacker, he leads AI agent projects for brands in Turkey and internationally. He designs each agent's tasks, permissions, human-approval points and success criteria, and with the experiment-and-measurement discipline of growth hacking treats every agent as a hypothesis and every result as a measurement. Agents are built through Unit Software's AI agent development service. Uğur Deniz İlhan also provides AI Agent Manager training and consulting to companies and managers.
Erman Kavlak is the founder of Yönetici İnsan. Yönetici İnsan closely follows the new positions AI is creating, defines the skills companies will need and builds candidate pools for these roles. It works with companies preparing to bring AI agents into their teams and looking for the right people to lead that change.
Links
Sources
- TechCrunch ·
Follow UNIT Journal
What's new in search, AI and technology, in your feed every day.


