The Future of Work is not Humans in the Loop, It’s Humans in the Lead

As I travel around the world speaking with business leaders from our clients across industries, they all understand AI will take over an increasing share of their business processes. However, they also understand humans must continue to own the purpose, the decisions and the consequences of their business.

For years, we have described responsible AI with a reassuring phrase: human in the loop. It suggests that people will remain present, checking the machine's work and stepping in when needed, but is just “being in the loop” a vision of the future that will inspire our workforce today? Does it underscore the importance of the role we must all play.

For example, a few years ago Amazon’s machine learning team uncovered a big problem with their automated recruiting system.It was trained to screen applicants by observing ten years of hiring and resume data.However, like much of the tech industry, most of the candidates were male. The automated system taught itself male candidates were preferrable and even began penalizing resumes that included any mention of non-male organizations (e.g., graduates of two all-women’s colleges, a women’s chess club champion). 

Of course, this example in no way is meant to single out Amazon since many companies have struggled with similar challenges with their automation tools. At the same time, the impact on candidates was significant, so the role of an experienced talent acquisition team was very important in this example.That is whyI strongly believe we need to shift the conversation from “humans in the loop” to “humans in the lead.”

From oversight to leadership

Being in the lead is not the same as being involved in every action. A leader does not personally complete every task performed by a team. A leader sets direction,establishesboundaries,allocatesresponsibility, andremainsaccountable for the outcome. The same idea should guide how organizations execute strategic workforce transformation. 

AI should be able to do what it does best: process large volumes of information, identify patterns, coordinate repeatable activities and improve speed and consistency. Humans should do what only humans can ultimately own: define the objective, interpret context, exercise judgment, consider ethical consequences, and be involved every time a decision will have a significant impact on the current or future livelihood of another person.

This distinction matters. The central question is not, “Was a human present at every point of the process?” It is, “Who will be held accountable and face the consequences if something goes wrong?” 

Do not automate yesterday

Too many AI conversations begin with a narrow question: Which individual task can we automate? In practice, this runs the risk of simply adding an automation layer to inefficient workflows without adding value. Insights from your team’s experience both managing the processes and client relationships should inform the future of your business. 

This is why AI transformation goes beyond traditional technology deployment including work reinvention. Now that we have this fantastic technology, how can we rethink the work to make it less painful, boring, and more enjoyable for our employees? How can we rethink it in a way that adds more value for our clients?

In recruiting, for example, it is possible to imagine different agents handling distinct parts of a process. One may organize information, another may coordinate scheduling and another may check whether defined requirements have been met. These agents could work together to prepare a recommendation. But the consequential decision should remain with a person who can understand the candidate, the role, the organization and the wider context.

Recent ManpowerGroup research of 40,000+ employers across 41 countries reinforces this important point. We asked them to tell us what adds the most value in their hiring process. Most of the hiring managers (57%) believed “experienced recruiters to review resumes” added the most value to the process.Nearly half of respondents also believed various forms of AI automation within the hiring process were valuable, so these results also underscore the desire for human-led process improvement supplemented by automation. 

Architecture is an ethical choice

Separating agents is not only a technical decision. It can also create clearer lines of responsibility. Each agent can have a defined purpose, controlled access to data and explicit limits. That structure can support explainability because organizations can better understand which component performed which activity and what information it used.

This is essential in any process that affects people's opportunities.Speed cannot come at the expense of fairness.Efficiency cannot override privacy. A recommendation cannot become a decision simply because it arrives with confidence. Responsible AI must be designed into the system from the beginning through access controls, governance, testing, transparency and clear human decision rights.

The more autonomous parts of a process become, the more intentional the human leadership around them must be. Removing people from routine handoffs should not remove human accountability. It should make that accountability more visible.

A challenge to traditional models of leadership

AI also amplifies the challenge for a familiar model of leadership. For years, I’ve spoken about how expertise is increasingly available at people's fingertips. Leaders will not create value simply by holding more information than everyone else. Their role is to show direction, energize teams, build trust and help people make sense of what technology produces. As AI accelerates this trend, the risk is growing for leaders that haven’t shifted 

That requires leaders to involve employees in identifying use cases and redesigning work. Trust is not created by announcing a new system after the decisions have been made.It grows when we involve the people who are actually doing the work in the AI business transformation process since they will know where it can add the most value.

A practical test for every AI process

Before scaling an AI-enabled workflow, leaders should be able to answer five questions:

• What human outcome is this process intended to improve?

• Which activities can agents perform, and what are their defined limits?

• Who has access to which data, and why?

• Which decisions require human judgment and accountability?

• How will we explain, challenge and correct the system's recommendations?

If those answers are unclear, the organization is not ready to scale the process, regardless of how impressive the technology may be.

Keep humanity at the center

The rise of agentic AI does not make people less important. It raises the value of the capabilities that make us human: judgment, creativity, empathy, responsibility and the ability to understand consequences beyond the data in front of us.

There will be processes in which no human remains in every operational loop. That is not necessarily a failure of responsible AI, but we must ensure people are still responsible for the outcomes.

AI can manage parts of a process. It can accelerate analysis. It can connect specialized agents and bring forward recommendations.But it cannot carry moral responsibility for a decision or any outcome for that matter.That responsibility belongs to us.

The organizations that succeed with AI will not be those that keep a person clicking approval at every stage. They will be those that redesign work boldly while drawing an unmistakable line around human dignity. The goal is nota human in every loop. It is humans in the lead.