Aravindhan G

Position: View B — Keep Humans in the Loop (Evolved to “Human-on-the-Loop”)
I strongly support View B — organisations should retain humans in the loop, even when AI can automate processes end-to-end. However, this human role must evolve from execution to strategic oversight — what I call “Human-on-the-Loop (HOTL).”
Why Humans Must Remain — Even in Fully Automated Systems
1. Managing Algorithmic Drift and Hidden Risks
AI systems degrade over time as real-world conditions change. Without human oversight, errors can silently scale. For example, in automated supply chain planning, a small forecasting drift can cascade into large inventory imbalances before detection. Human monitoring ensures early intervention.
2. Handling Black Swan Events and Contextual Nuance
AI is trained on historical data and struggles with unprecedented events. During COVID-19, many automated demand forecasting systems failed dramatically. Human judgment was essential to reinterpret signals and redesign responses.
3. Preserving Institutional Knowledge and Control
Fully removing humans leads to “deskilling.” If the system fails due to cyberattack or infrastructure issues, organisations may lose the ability to operate manually. Keeping humans engaged ensures continuity and resilience.
4. Ethical Accountability and Governance
In processes like loan approvals or healthcare decisions, accountability cannot be assigned to an algorithm. Human oversight is essential to ensure fairness, compliance, and ethical alignment.
Where AI Should Lead — and Humans Should Step Back
That said, AI should absolutely handle execution-level activities:
Real-time decision-making in supply chains
High-frequency trading adjustments
Automated quality inspection in manufacturing
Humans cannot match AI’s speed and scale here.
The Right Model: Human-on-the-Loop (HOTL)
The winning organisations will not choose between automation and human involvement. They will:
Let AI run the process at machine speed
Keep humans at the system level for supervision, exception handling, and governance
Example: Autonomous Warehouse Operations
In modern e-commerce warehouses:
AI systems manage picking routes, inventory allocation, and robotic movement in real time
This improves speed and reduces errors significantly
However:
Humans monitor system performance dashboards
Intervene during exceptions (system anomalies, demand spikes, equipment failure)
Continuously improve system rules and logic
Outcome:
30–50% improvement in throughput
Reduced operational errors
Maintained control and adaptability
Conclusion
Fully autonomous systems maximise efficiency but introduce systemic risk.
Human-dependent systems ensure control but limit scalability.
The optimal approach is not choosing one over the other, but redefining the human role.
Organisations should keep humans in the loop — not in execution, but in control.
That is where true operational excellence lies.