Skip to content
Fast Company Executive Board

The Fast Company Executive Board is a private, fee-based network of influential leaders, experts, executives, and entrepreneurs who share their insights with our audience.

Beyond humans in the AI loop

A person at the end of an automated process is not the same as humanity in charge.

[Photo: Getty Images]

Profile photo of Cheryl Contee

Co-Founder · BrightWorksAI.com | Change Agent AI

Concern over artificial intelligence is on everyone’s radar. It recently even became the subject of papal consternation, with Pope Leo XIV calling for AI to be “disarmed.”

I don’t believe the pope meant we should march into Silicon Valley with pitchforks and unplug the servers, as fulfilling as that might sound to some these days. His point is not that we should abandon technology. Instead, Leo, like many others, is arguing that AI should be stripped of its most dangerous weapons: secrecy, concentrated power, manipulation, and dehumanization. It’s also a caution against the idea that just because you can do something with AI, doesn’t mean there should be no guardrails whatsoever.

The pontiff calls us to be “magnificent humans.” Many organizations are embracing the reassuring technology term from AI: “human in the loop.” The idea is that a real person reviews, approves, or corrects what an AI system produces.

WHAT HUMAN-LED AI LOOKS LIKE

A human in the loop is certainly better than letting the robots run loose through the organization like digital bulls in a china shop. But it is not enough.

It’s time we think beyond humans in the AI loop. Instead, humans should take the AI lead. Here’s what that looks like: A human in the loop checks whether the machine did the job correctly. A human in the lead decides whether the machine should be doing that job at all.

The reviewer asks: Is this answer accurate? Too often, organizations place people at the end of an automated process and call that oversight or supervision. But by then, the bot has already framed the problem, narrowed the options, set the pace, and created pressure to approve its recommendation. The employee may technically be in the loop, but a machine without context or lived experience is driving the bus.

Human-led AI starts earlier. People define the purpose, values, boundaries, and acceptable risks before the technology is deployed. They retain the power to question the tech, change it, pause it, or reject it entirely. The leader of AI asks:

  • Is this the right question?
  • Who could be harmed?
  • Is this secure?
  • What assumptions or biases might be baked into the system?
  • What happens to the person who gets mislabeled, rejected, deprioritized, or ignored?
  • Who will be accountable if or when the machine gets it wrong?

The human governs. The machine assists.

THE ETHICS OF AI

Ensuring that your organization has the internal decision-making and values-based infrastructure ahead of AI use is critical for successful, ethical impact. AI efficiency is not justice. Speed is not wisdom. Personalization is not care. And a confident answer is not necessarily a truthful one.

Leadership in the AI age means holding multiple values at once. Productivity matters, but so does dignity, creativity, accessibility, development, and belonging. Speed matters, but so does reflection and due process. Data matters, but so does lived experience.

Religious and philosophical traditions have spent centuries struggling with the very questions the technology sector now urgently needs to confront: What does power owe the vulnerable? What is the difference between knowledge and wisdom? Where do compassion, integrity, and truth fit? Are there things we should refuse to do, even when they are possible?

These questions remind us that human beings are not merely clicks, data points, productivity units, customer segments, or bundles of risk. We have also begun speaking about AI as though it acts on its own: “The algorithm rejected the applicant.” “The model flagged the employee.” Or “The system denied the request.”

But AI does not descend from the cloud in a shimmer of neutral mathematical innocence. Humans build these systems. Humans buy them. Humans choose the data, establish the objectives, and decide where they will be deployed.

When AI causes harm, it’s not okay for everyone involved to point toward the machine and say, “Wasn’t me,” to quote the immortal words of Shaggy. Somebody did indeed do it.

TOWARD ACCOUNTABILITY

Human-led AI requires clear accountability. A named person or team should own the decision to use the system, monitor its impact, and have the authority to intervene. People affected by consequential AI-assisted decisions should know when AI was involved and have a meaningful way to challenge the outcome. An appeal process that leads directly back to the same automated system is not an appeal process. It is a haunted carousel full of scary mirrors.

A rubber stamp is not oversight. And a human shield is not governance. None of this means we should fear or reject AI.

As a senior advisor to Posterchild.AI and Change Agent AI, I have seen how mission-driven AI tools can help small nonprofits translate resources, analyze public data, improve accessibility, accelerate research, and communicate with communities in ways once available only to large institutions.

Using AI for positive change is real progress. But the goal of using AI should be to remove toil—not our essential humanity—from the work. AI can draft a message. It cannot create the trust that makes someone believe it. It can summarize community feedback. It cannot replace active listening to a community in need. It can generate compassionate language. It cannot accept the obligations or experience the good karma that comes with compassion.

WHICH WILL YOU CHOOSE?

The pope’s call to disarm AI is ultimately a call to disarm our own infatuation with technological inevitability. We are not helpless passengers on the AI train to dystopia.

We can use AI tools to promote social progress like expanding access, creativity, opportunity, and civic participation. Or we can use AI to consolidate power in the hands of the few rather than the many, disguise discrimination, manipulate facts, and make consequential decisions while evading responsibility.

A human in the loop can catch a machine’s mistake. A human in the lead decides where we are going, who gets to come with us, and whether the machine has any business choosing the destination. Which will you choose to be today?

Cheryl Contee is co-founder of BrightWorksAI.