LYQOR™ Founder Dr. Natalie Callis Examines the Human Side of AI Oversight
AI governance conversations often get technical quickly. Organizations talk about models, data, controls, accuracy, security, compliance, and the rules that should determine how artificial intelligence is used.
Those conversations are necessary, but Dr. Natalie Callis sees a second layer that’s easier to overlook. At some point, a person has to decide what to do with what the technology produces. They may accept it, question it, investigate further, override it, or decide that someone else needs to become involved.
“The more intelligence we place into machines, the more intentional we have to become about the human judgment surrounding them,” says Dr. Callis. Her concern is not whether organizations should govern AI. It is whether governance is complete if it focuses on the machine without giving equal attention to the people making decisions around it.
AI Can Recommend, but People Still Respond
Suppose an AI system flags a potential problem. What happens next depends on more than the alert’s quality.
The person reviewing it has to decide how seriously to take it. They may compare it with other information, question how the system reached the result, ask a colleague for another opinion, or decide immediate action is necessary. If something about the recommendation feels wrong, they also have to decide whether that concern is strong enough to challenge the technology.
No single human response fits that situation. One person may trust the system because it has been reliable in the past, while another may be skeptical because the output conflicts with what they are seeing in front of them.
“AI may produce the information. The human still has to decide what to do with it,” says Dr. Callis. That part of the process can be easy to underestimate when attention is concentrated on what the technology can do.
The Same Output Can Meet Very Different Conditions
How someone responds to AI can also depend on what is happening when the information arrives. Reviewing a recommendation with plenty of time and few distractions is one thing. Reviewing it when a deadline is closing in, several other problems need attention, and a wrong decision could have serious consequences is another.
In both cases, the system could produce exactly the same result. The human conditions surrounding that result are what changed.
A heavy workload might make the quickest answer more appealing. Uncertainty may make a technology-generated recommendation seem reassuring because it appears definitive, while fear of getting something wrong can make it tempting to defer to a system rather than rely on one’s own judgment.
People can move in the opposite direction too. Someone who has already formed a strong opinion may dismiss useful AI-generated information because accepting it would require reconsidering a decision they have already made.
Dr. Callis’ work on performance under pressure focuses on precisely these kinds of moments. The issue is not simply whether someone trusts AI too much or too little, but what may be shaping that response at the time.
Experience Does Not Make Someone Immune to Pressure
Putting an experienced professional in charge of oversight is important, but experience alone cannot account for every condition in which judgment will be required. A person can be highly knowledgeable and still find it harder to question, slow down, or change course in a particular moment.
Organizational expectations can add another complication. If an organization has invested significant time and money in an AI system, an employee may question its output differently than they would if the same recommendation came from an ordinary source.
The technology’s perceived authority matters as well. When a system is regarded as sophisticated, highly accurate, or more objective than a person, challenging it may require a level of confidence that is easy to assume people will have until the moment actually arrives.
None of those examples necessarily indicates a problem with the AI itself. They point instead to the human environment in which oversight is taking place, and that environment can influence whether expertise is fully used.
Looking Beyond the Machine
This human layer is the focus of Human Intelligence Governance™, which Dr. Callis is advancing through LYQOR™. The work considers judgment, accountability, regulatory access, decision integrity, and human responsibility alongside the technical structures used to govern intelligent systems.
“AI governance cannot stop at governing the machine. We also have to understand the human being making decisions around the machine,” says Dr. Callis. The distinction matters because technical controls and human oversight address different parts of the problem.
An organization can establish detailed procedures for how an AI system should be used. It can define thresholds, create escalation pathways, monitor outputs, and assign responsibilities. Yet those structures still reach a point where someone has to notice that a situation deserves a closer look and decide to do something about it.
That is where accountability becomes practical rather than theoretical. It is one thing to assign responsibility on paper and another to understand whether the person holding that responsibility can exercise sound judgment when the circumstances become uncertain or uncomfortable.
What Does Human Oversight Look Like in Practice?
Human oversight is sometimes discussed as though keeping a person involved automatically solves the problem. Dr. Callis’ perspective raises a harder question: what conditions does that person need in order to provide meaningful oversight?
An employee who is technically authorized to override an AI recommendation may still be reluctant to do so. A manager responsible for escalation may have the right procedure available but feel pressure not to delay a decision. Another professional may recognize that something deserves further review but be unsure whether their concern is enough to challenge a system that others trust.
These situations are not solved simply by declaring that a human remains “in the loop.” Organizations also have to consider whether people understand their responsibility, can exercise it in practice, and can still access judgment when pressure changes the situation.
This is where Human Intelligence Governance™ connects with Dr. Callis’ broader Perform Under Pressure™ work. Both focus on what remains available to people when a consequential moment becomes harder than expected.
Governing AI Means Preparing the People Around It
As AI becomes part of more decisions, organizations will continue improving technical governance. Dr. Callis argues that preparing the people around those systems deserves attention alongside it.
That preparation includes more than teaching employees how an AI tool works. People also need clarity about when they are expected to question it, what happens when human judgment and an AI recommendation disagree, where responsibility ultimately sits, and whether the environment genuinely supports escalation when concerns arise.
AI may become faster, more capable, and more deeply embedded in everyday work. Those advances do not remove the moment when a person has to decide what happens next.
For Dr. Callis, that is why the human side of oversight cannot be treated as an afterthought. Governing intelligent technology also means understanding the judgment of the people entrusted to govern it.
About Dr. Natalie Callis
Dr. Natalie Callis, DNP, NEA-BC, is a PhD candidate and doctoral researcher examining human performance under pressure. She is the founder of LYQOR™ and a Human Intelligence Strategist whose work examines human performance, judgment, and decision-making under pressure. With more than 25 years of experience in healthcare leadership and high-stakes environments, she develops research-informed frameworks that explore what happens internally when pressure begins to affect performance. Her work spans leadership, healthcare, sports, behavioral risk, education, and the governance of emerging technologies.
About LYQOR™
LYQOR™ is the company and intellectual architecture behind Dr. Natalie Callis’ work on human performance and judgment under pressure. Its research-informed frameworks have applications across leadership, sports, artificial intelligence, behavioral risk, healthcare, education, and other consequential environments. Human Intelligence Governance™ extends this body of work to the human judgment, oversight, accountability, and decision integrity required to govern intelligent technologies responsibly.
Connect with Dr. Natalie Callis:
Instagram: instagram.com/dr_ncallis
LinkedIn: linkedin.com/in/dr-natalie-callis
X: x.com/Drncallis1
Disclaimer: This article is for informational purposes only and explores the human aspects of AI oversight. The views expressed are those of the author and do not constitute professional or legal advice.