Why AI needs human judgment to keep networks running right

AI as a collaborative assistant in network operations

AI is often seen as a disruptor, but in network operations, it’s more effective as a collaborator than a replacement. Its role isn’t to overtake human expertise but to enhance it. Networks demand extremely high reliability, known in the industry as “three nines,” or 99.9% uptime. Even an AI system with 80% accuracy doesn’t meet that threshold. That’s why human judgment remains essential. The future of network management depends on building systems where AI amplifies human decision-making, not automates it entirely.

Jason Lovelace, Outbound Product Management Leader at IBM, described this relationship well. He said, “To implement AI in the network, you need to think about maintaining or improving on three nines. The goal here is the illumination of human judgment through AI as a partner.” His message is simple but vital: AI should help engineers make sharper, faster, and more reliable decisions without taking away their control or understanding of the system.

For executives, this calls for a mindset shift. AI integration isn’t just about automation. It’s about designing workflows where AI augments situational awareness and speeds up response times. Businesses that embrace cooperative AI, where systems and people work in tandem, will see more stability, quicker diagnostics, and clearer accountability across their network operations.

Leaders should invest in infrastructure that supports human-AI collaboration rather than full delegation. This approach reduces operational risk and improves decision precision. The result is a network environment that’s smarter, faster, and still grounded in human oversight. The organizations that balance speed and control here are the ones that will set new performance standards for intelligent network systems.

The “See” phase – gathering real-time telemetry and context

The first step in IBM’s framework for AI-driven network operations is “See.” This phase is about visibility, understanding what’s happening across the network in real time. To make reliable decisions, AI systems require two things: telemetry and context. Telemetry gives live reports on what’s happening in the network. Context explains why it’s happening. You need both to understand the full picture.

Telemetry works best when paired with time-series monitoring, which tracks metrics continuously over time. This allows the system to identify issues early, even before alarms trigger. It’s the difference between reacting to a failure and preventing one. As AI processes streams of telemetry data, it builds predictive insight, anticipating potential disruptions and guiding engineers to act before performance degrades.

Jason Lovelace explained that combining agentic AI with time-series analysis gives engineers a “clear view of what’s happening.” Without that visibility, decision-making becomes reactive and less precise. Context, on the other hand, helps teams distinguish between isolated technical noise and systemic problems that could scale into outages.

For executives, the strategic takeaway is clarity. Network resilience comes from data timeliness and context accuracy. Investing in telemetry pipelines and contextual analytics gives teams the operational intelligence they need to keep performance stable, secure, and predictable.

However, data balance matters. Too much context can overwhelm AI models and reduce accuracy. The right approach is controlled feeding, enough data for insight but not so much that it derails precision. Building this capability means setting disciplined data governance standards and ensuring teams understand how their telemetry and context layers interact.

For a business leader, mastering the “See” phase isn’t about adding complexity; it’s about enabling precision. When AI sees clearly and engineers understand what it sees, network reliability strengthens. That’s how you transform AI from a monitoring tool into an operational advantage.

The “Use” phase – prioritizing AI tools over specific models

Once the network environment is fully observed and understood, the focus moves to how AI is applied, this is the “Use” phase. The critical shift here is understanding that success doesn’t depend solely on what model is used but on how effective and secure the AI tools are. Models evolve rapidly, but tools define day-to-day utility. Jason Lovelace, from IBM, explained that the richness of tools determines what network engineers accomplish each day.

AI tools that can generate network scripts or automate configuration tasks expand an engineer’s capacity substantially. They provide instant functionality, reduce repetitive manual work, and lower operational latency. What matters is building systems that let engineers act faster and more confidently while maintaining full control of network operations.

For executives, this phase is about investment priorities. Developing scalable and adaptable AI toolsets must take precedence over constantly chasing newer models. The organization should direct resources toward tool ecosystems that remain secure, integrable, and adaptable across changing AI architectures. This strategic focus creates flexibility, allowing operations to evolve as technology shifts, without incurring the costs of constant model replacement.

The message is clear: choose tools that empower your teams, protect your infrastructure, and grow with your network demands. Over time, this creates operational resilience, not through model sophistication but through functional diversity and robust configuration management. For enterprise leaders, the long-term value lies in owning AI tools that remain relevant as models come and go.

The “Prove” phase – validating AI reliability with guardrails

The “Prove” phase ensures that AI outputs remain accurate, reliable, and free from accumulated errors. In complex network environments, where automation sequences can chain across multiple decision layers, a small inaccuracy can scale into major disruptions. Jason Lovelace pointed out that “if your context window is filling up, the compounding effect of a hallucination at step four or five then makes steps 17, 18, and 19 not reliable.” The message emphasizes continuous validation. Engineers must verify every stage of AI reasoning to avoid cascading faults.

Reliability is built through structured guardrails, defined verification steps, error containment strategies, and audit checkpoints. This framework ensures the AI remains accountable at each point in its decision process. Engineers need accessible, interpretable data from AI tools, enabling them to trace back any recommendation and correct it before execution.

For business leaders, reliability validation isn’t optional, it’s essential for trust. Without dependable AI performance, automation creates unpredictability, not efficiency. Implementing layered guardrails protects operational stability and assures both technical teams and stakeholders that the system performs as intended under pressure.

In the context of governance, this step also reinforces compliance and operational integrity. Regular auditing of AI decision paths aligns network operations with internal policy and external regulation. Strong validation systems ultimately define whether AI can scale into broader use across critical infrastructure.

Executives should view the “Prove” phase as a continuous process, not a one-time verification. As AI tools gain capability, oversight measures must evolve to match. Reliability and governance strengthen reputation, maintain client trust, and sustain the long-term viability of AI-driven operations.

The “Act” phase – tailoring AI application based on experience levels

In the final phase of IBM’s framework, “Act,” the emphasis is on execution and how engineers apply AI insights to real network decisions. This stage is not uniform, its effectiveness depends heavily on the experience level of the engineer using it. Junior engineers benefit most from AI-guided assistance, which provides structured recommendations and detailed explanations of each action. This helps them understand network logic while executing reliable decisions safely.

For senior engineers, AI should serve as an informed assistant rather than a decision-maker. They should be able to evaluate, challenge, or refine the system’s recommendations. Jason Lovelace from IBM highlighted this by stating that engineers must “push back on the model and have the model reconsider its actions.” His point underscores the idea of retaining human oversight where performance and risk intersect.

For executives, implementing this phase effectively means designing AI access and control layers suited to each role. This ensures that less experienced engineers have guidance while advanced users maintain decision authority. It also encourages skill growth at every level, AI supports learning and competency while protecting system integrity.

Organizations that adopt this adaptive approach enhance both efficiency and capability. Junior staff build confidence and accuracy; senior staff maintain strategic control. For business leaders, this balance translates into operational scalability, the system evolves as the workforce advances. The “Act” phase, done correctly, integrates AI as a practical and educational force, not just an automation mechanism. Over time, that cultivates a stronger bench of engineers and a smarter, more reliable network operation.

Establishing a structured AI framework to build trust and ensure governance

A structured AI framework, such as IBM’s four-step model, provides the governance backbone that network operations need to manage complexity and maintain trust. It defines how data is gathered (“See”), tools are used (“Use”), results are validated (“Prove”), and actions are taken (“Act”). This structure keeps network teams aligned across workflows while ensuring leadership can track AI influence and operational impact.

Jason Lovelace, Outbound Product Management Leader at IBM, explained that a solid framework allows organizations to “reap the benefits of agentic NetOps” while preserving governance and safety standards, regardless of how the underlying AI technology evolves. This approach prevents disjointed AI adoption and ensures consistency, reliability, and accountability across all network functions.

For executives, the value of this framework lies in control and scalability. It gives visibility into how AI is used, guarantees compliance with corporate and regulatory standards, and builds confidence that automation is being handled responsibly. It also helps define measurable performance standards across teams, a key factor when integrating AI into mission-critical environments.

From a leadership perspective, structured frameworks also strengthen workforce trust. Engineers know that consistent procedures protect both their work and the system they manage. Leaders, in turn, can confirm that governance remains intact as AI capabilities expand. The framework becomes a blueprint for disciplined innovation, it allows adaptation to new AI capabilities while maintaining the reliability and governance expected in enterprise-grade operations.

For long-term success, executives should enforce frameworks that evolve alongside technology but remain grounded in human oversight and clear accountability. This dual focus, advanced AI use with stable governance, positions the organization to scale safely, innovate responsibly, and sustain operational excellence in an increasingly automated world.

Main highlights

• AI as a partner, not a replacement: Executives should view AI as a collaborator that enhances human capability rather than replaces it. Pairing AI precision with human oversight ensures the reliability and operational continuity demanded in network environments.
• Seeing through data for Real-Time clarity: Leaders should invest in systems that merge real-time telemetry with contextual data. This combination enables predictive insight, minimizes downtime, and guides more informed infrastructure decisions.
• Tools over models for Long-Term flexibility: Focus on developing adaptable AI toolsets instead of chasing evolving models. Versatile tools empower network teams to automate safely while staying responsive to new business and technology demands.
• Proving reliability through guardrails: Governance and validation frameworks should be non-negotiable. By enforcing structured guardrails and ongoing AI audits, executives can safeguard performance, maintain compliance, and sustain stakeholder trust.
• Acting through Experience-Based control: Leaders should structure AI access and responsibilities by skill level. Junior engineers can rely on AI guidance to learn, while senior engineers retain decision-making authority to ensure operational accuracy.
• Building structured frameworks for trust and governance: Executives must standardize AI use across the organization through structured frameworks. This ensures consistent governance, transparent accountability, and scalability as AI capabilities evolve.