Navigating the Loop: HITL → HOTL → HOOTL in Autonomous Networks

As autonomous networks transition from theory into practical operations, a critical question emerges: What role do humans play in this new paradigm?

In traditional views of automation, autonomy is often mistakenly associated with the elimination of human involvement. However, within the context of telecommunications and critical operations, this perspective is not only simplistic but also dangerous.

True autonomy does not mean removing humans but redefining where their contribution adds the greatest value. To achieve this, new interaction models are required between operators, automation systems, and artificial intelligence agents.

Three Models, One Transition

In the book “AI-Driven Autonomous Networks,” Matías Lambert clearly categorizes these models:

• Human-in-the-Loop (HITL): Automated decisions require human validation before execution. This model prevails during initial phases when perceived risk is high or systems are still maturing.
• Human-on-the-Loop (HOTL): Systems operate autonomously but under human supervision. Humans don’t approve of every action but are prepared to intervene when anomalies or unexpected behaviors arise.
• Human-out-of-the-Loop (HOOTL): The system functions completely autonomously within predefined boundaries. Humans define the objectives, constraints, and success criteria but do not intervene in routine operations.

This phased approach ensures a secure transition toward higher autonomy levels while establishing a clear framework for responsibilities, risks, and governance.

Where Are We Today?

In practice, we observe three distinct contexts for Human-in-the-Loop (HITL):

Critical Operations:
HITL is essential in safety- or business-critical scenarios—production changes, sensitive network element adjustments, and cross-domain integrations—where human oversight mitigates high risk.

Transitional Phases:
During early-stage deployments and pilot programs, HITL acts as a stepping-stone toward greater autonomy, allowing teams to validate automation outcomes, refine processes, and build trust in the system.

Change Aversion:
When HITL is adopted solely out of fear or resistance to change, it creates unnecessary bottlenecks and hinders progress. It is crucial to distinguish between valid risk management and defaulting to manual checks that no longer add value.

The key is to align each use case with the appropriate interaction model—HITL, HOTL, or HOOTL—based on domain maturity, criticality, and business exposure.

Human-in-the-Loop: A Lever for Acceleration

It may sound contradictory, but HILOP is a powerful accelerator for AI implementation. In transitional stages, leveraging agents as intermediaries allows organizations to iteratively expand autonomous capabilities with human oversight at critical checkpoints. This measured approach ensures each automation milestone delivers value while maintaining control and confidence.

Autonomy with Governance

The concept of autonomous networks isn’t about removing humans; it’s about freeing their time to focus on tasks of greater strategic importance: defining objectives, assessing impacts, overseeing risks, and adjusting strategic direction.

HITL, HOTL, and HOOTL models transcend mere technical labels. They are organizational design tools that enable us to build smarter networks without sacrificing traceability, ethics, or the human oversight essential to every critical operation.