GenAI Ethics for Leaders #7: Humanizing AI Agents Through Human-in-the-Loop Strategies

Why Keeping Humans in the Loop is Essential for Ethical AI, Organizational Success, and Future-Proofing Your GenAI Strategy

Introduction

The conversation around AI agents is evolving rapidly, and if you’re a leader integrating Generative AI (GenAI) into your organization, you need to pay attention. Last week, OpenAI’s Sam Altman stated that AI agents will soon be potent, and everyone will have one. With OpenAI’s latest product releases, Operator and Deep Research, the race to develop highly autonomous AI is accelerating.

At the same time, the counter-narrative to Big Tech’s AI dominance is growing. Startups like DeepSeek have shown that competent GenAI models can be built using open-source frameworks and at a fraction of the cost—challenging the idea that only companies with billions of dollars can create the next generation of AI. Stanford researchers took this further, training an AI model for just $50, proving that access to powerful AI is becoming democratized.

The question for leaders isn’t just about whether to adopt AI agents—but how to ensure they remain human-centered. The key to responsible AI adoption is Human-in-the-Loop (HITL) strategies—where humans guide, intervene, and refine AI decision-making processes to ensure alignment with ethics, agency, and societal values. For example, to the extent you have your AI agent hotel or dinner reservation while you are on a business trip, you may want to have the option for the agent to ask you before paying for such services. In a more mission-critical example in healthcare, the stakes are higher, and humans, whether patients or caregivers, want to ensure they are in the loop; this is especially true given the nascent stage of GenAI technology and healthcare challenges.

This article explores how leaders can humanize GenAI agents, ensuring they empower users rather than replace human oversight with a focus on healthcare, higher education, and financial services.

Recap of Previous Articles

Building the Ethical Foundation for GenAI Agents. As we leaders build their GenAI ethical frameworks, below will recap the key learnings from previous installments and connect them to this article’s discussion:

  1. Article #1: Navigating Bias in AI Systems

Bias in GenAI leads to unintentional outcomes, which may alienate some of your stakeholders (students, patients, or clients) and impact your organization’s profitability and growth. When designing  AI agents, leaders must ensure that model bias checks and human oversight prevent toxic algorithmic harm that negatively affects your organization and you as a leader.

  1. Article #2: Ensuring Data Privacy in AI Applications

GenAI systems process large amounts of personal data. Human-in-the-loop strategies must optimize for the proper use of data, aligned with your privacy and confidentiality policies. Proper guardrail will ensure AI agents access your data in accordance with your regulators’ frameworks (HIPAA, FERPA, et al.), ethical principles, and your organization’s brand, goals, and mission.

  1. Article #3: AI Accountability and Traceability

It is key to define who is responsible for which AI decisions. AI agents must have transparent decision-making processes with clear accountability structures so that humans remain in control, especially on critical decisions that impact their human users.

  1. Article #4: AI’s Impact on Workforce and Skills

AI agents can lead to deskilling. AI agents should be designed as enhancers, not replacers, keeping humans engaged in critical decision-making.

  1. Article #5: Intellectual Property and AI-Generated Content

IP ownership must be defined and guarded as AI agents generate more content. AI agents should support human creativity, not replace it, ensuring that machines do what they are good at and humans as they are.

  1. Article #6: Balancing Organizational Cost-Cutting with Human-Centered AI

AI seems appropriate for reducing costs in some specific administrative tasks. Leaders must ensure that AI agents don’t relocate ethical responsibilities and decisions to machines and that human oversight remains at the center of AI-driven critical decisions.

Now, let’s explore GenAI agents, Human-in-the-loop strategies, and how leaders can design AI systems that enhance—not erode—human agency.

Machine-Human Partnership: GenAI agents with Human-in-the-Loop

GenAI agents are becoming more autonomous and capable of handling complex decision-making tasks. However, GenAI agents can make flawed or ethically questionable decisions without human oversight. HITL ensures that AI:

  • Aligns with ethical frameworks
  • Respect the user’s agency, intent, and values
  • Prioritizes human well-being over automation efficiency
  • Prevents over-reliance on AI in critical decisions

The challenge isn’t just keeping humans in the loop and ensuring that humans have real decision-making power over their AI agents, especially critical decisions.

Sector-Specific Applications of HITL for AI Agents

Healthcare: AI Supporting, Not Replacing Clinicians

  • Medical Diagnostics & GIRL: AI can analyze patient data, but doctors must validate findings before making treatment decisions.
  • AI in Telehealth: AI-powered chatbots can assist with triage, but a healthcare professional must give final medical advice.
  • Patient Agency: Patients should have control over how AI systems interact with their health data, with clear opt-in/opt-out options.

Higher Education: AI as a Learning Partner, Not a Substitute for Educators

  • AI-Assisted Teaching: AI tutors can support students but should not replace human educators who bring context, mentorship, and more profound learning experiences.
  • Ethical AI in Student Evaluations: AI grading must include human review to prevent unfair assessment biases.
  • Student Ownership of AI-Generated Work: Universities must set clear guidelines for evaluating and attributing AI-generated content.

Financial Services: AI Agents in Banking and Investment Advising

  • Robo-Advisors & GIRL: AI can suggest investment strategies, but human advisors must assess risks and ethical considerations before execution.
  • Fraud Detection and Decision Authority: AI can flag suspicious transactions, but human compliance teams must take final action.
  • Customer Consent & Transparency: AI should not execute financial decisions without explicit customer approval and understanding.

Guidelines for Ethical & Human-Centered AI Agents

To create AI agents that empower rather than control, leaders must adopt the following strategies:

  • Define Boundaries of AI Autonomy: Establish clear decision-making limits where human intervention is mandatory.
  • Ensure Explainability: AI agents’ design should provide users with clear reasoning behind their critical decisions and allow users to be in the loop.
  • Prioritize AI-User Collaboration: Users should be able to override AI decisions when necessary.
  • Continuous Learning & Feedback Loops: AI agents should learn from data and human feedback.
  • Mandate Ethical Audits: Periodic AI agents’ assessments and monitoring are key to closing the gap between decision outcomes and human ethical standards aligned with the users’ goals and values.

Leadership Reflection: Key Questions to Consider

  1. How does your organization define when and where humans must intervene in AI decision-making?
  2. Are your AI agents designed to enhance user autonomy or replace human decision-making?
  3. What safeguards are in place to prevent AI agents from making unethical or unintended decisions?
  4. How are you ensuring that AI-driven processes remain transparent and explainable?

Conclusion: The Future of GenAI Agents Depends on Intentional Leaders Who Keep Critical Decisions in the Hands of Humans

As GenAI agents become more advanced and autonomous, leaders must require that humans remain at the center of AI decision-making. GenAI agents will most likely enhance efficiency and automate routine tasks; however, some user-critical decisions—especially in healthcare, education, and financial services— will always require human intervention, ethical frameworks, and context. This framework not only ensures safety but also mitigates organizational exposure.

A Human-in-the-Loop (HITL) approach is not just an ethical necessity—it’s a leadership imperative that ensures AI operates transparently, aligns with human values, and remains accountable. Organizations that embed HITL strategies today will mitigate risks, comply with evolving AI regulations, and build AI systems that enhance human expertise rather than replace it.

Successful leaders in the AI era understand that this technology’s potential can be enormous to the extent that it is optimized and aligned with human and organizational goals and values. The future of AI agents is about enhancing human capacities and delegating routine and low-value tasks while keeping critical decisions and agency in human hands. This human-machines partnership is a human-centered collaboration centered on ethical, responsible, mission-aligned growth.

Are your organization’s AI agents strategy centered on empowering your stakeholders, including your employees (educators, staff, doctors, nurses), or replacing them? Leaders’ well-thought decisions, centered on GenAI Ethical principles, will shape your organization and society.

About the Author

Freddie Seba is a distinguished thought leader and educator specializing in Generative AI ethics. He holds an MBA from Yale and an MA from Stanford and is pursuing a Doctorate in Education at the University of San Francisco (USF), focusing on GenAI Ethics. Since 2017, Freddie has served as a faculty at the Masters of Science in Digital Health Informatics at USF’s School of Nursing and Health Professions (SONHP). He teaches and mentors graduate students in this program and collaborates closely with the healthcare ecosystem. He developed and taught a course on Generative AI Ethics in Education and Healthcare Ecosystems. Freddie is a seasoned Silicon Valley entrepreneur, co-founding and working with innovative startups in financial services,  healthcare, and education. As a speaker, faculty, and writer, Freddie inspires others to navigate GenAI ethics complexities with purpose. You can find more information at www.freddieseba.com

About the Project

This article is part of a continuous exploration – a joint journey to share insights, foster discussions, and empower leaders with the frameworks they need to navigate the complex ethical landscape of Generative AI (GenAI). I want this series to be a space to critically interrogate, question, and leverage GenAI to drive the best possible societal impact together and shape our organizations and ecosystems as a conscious, intentional set of choices – not something we just fall into because we fail to see the new opportunity space. We can all be agents of change in our organizations, communities, homes, and professional networks. Hence, I see this as a joint exploration with fellow travelers. GenAI tools are utilized for this series, including ChatGPT, Grammarly, Speechify,  ZoomAI, and others.

Useful Information & References

  • Altman, S. (2025). Reflections on AI Agents & The Future. Retrieved from blog.samaltman.com
  • Drori, I., & Te’eni, D. (2024). Human-in-the-Loop AI: Opportunities & Risks. Communications of the ACM, 67(3), 72-79. ResearchGate
  • ​​Human Decision-making is Susceptible to AI-driven Manipulation (Sabour, et.al 2025) https://arxiv.org/html/2502.07663v1
  • Tschiatschek, et.al (2024) Challenging the Human-in-the-loop in Algorithmic Decision-making. https://arxiv.org/pdf/2405.10706
  • Stanford AI Research (2025). AI Model Trained for $50: A Case Study. TechCrunch