The Autonomous NOC Still Runs on People

An article by Vivek Bhamre, Director of Technical Sales & Delivery, VIAVI Solutions.
Ask any NOC leader what keeps them up at night and you’ll usually hear two answers. The first is restoring service quickly when something breaks. The second is finding the time to stop it breaking again. In a largely manual, reactive NOC, those two goals compete for the same people, and the first one almost always wins.
Automation and AI promise to change that. But after years of working with operators on autonomous network journeys, I’ve come to believe the technology roadmap is only half the plan. The other half, the one that determines whether the investment actually pays off, is what happens to the people.
The trap of the reactive NOC
Consider a typical Resource Trouble Management process: detect and correlate alarms, register and prioritize the incident, diagnose, authorize and execute a fix, verify recovery, and capture lessons learned. In an illustrative manual baseline, Tier 1 and Tier 2 together absorb around 75% of the person-hours spent running that cycle, with Tier 1 handling the sheer volume of alarms and tickets, and Tier 2 carrying the heaviest load of investigation and repair.
What gets squeezed out is the preventive work: permanently eliminating recurring faults, preparing changes carefully enough to avoid self-inflicted outages, catching performance degradation before customers notice, reviewing capacity and failover headroom, and testing resilience before a real outage does it for you. Defer that work long enough and recurring problems generate more incidents, which leaves even less time for prevention. It’s a cycle most operations teams will recognize.
Released capacity doesn’t redeploy itself

Here’s the uncomfortable part. When automation takes over alarm triage, evidence gathering or routine fixes, the hours it releases don’t automatically flow into prevention. Without a deliberate plan, they quietly disappear back into the daily incident queue.
That’s why we approach NOC transformation in three connected phases, each with an explicit people dimension.
1. Prepare the foundations. Automation is only as reliable as the data and procedures beneath it. Before anything is automated, operational teams validate inventory and topology, clean up operational data, standardize processes, and turn engineering experience into tested runbooks with clear policies and recovery checks. Take a loss of connectivity at a mobile site: if the inventory links that site to the wrong router, automation will confidently investigate the wrong equipment. The people who understand the network’s dependencies and failure patterns are the ones who prevent that. This phase even includes rehearsing a failed recovery action, so the team can prove it can stop execution, roll back and escalate correctly.
2. Redeploy released capacity. Start by measuring where time goes today, then validate the net capacity automation releases after accounting for exceptions, oversight and workflow maintenance. Rank the deferred preventive backlog by service impact and risk, match people to the work, and protect that time in the operating model. The test is simple: released hours need named owners, clear priorities, protected time and measurable outcomes. Released hours represent capacity for reassignment; they don’t automatically translate into fewer positions.
3. Build evolving expertise. Autonomous operations create jobs that didn’t exist in the tiered NOC: runbook owners and automation engineers, operational data stewards, AIOps model operations specialists, autonomous operations supervisors, intent and policy owners, and assurance leads who independently verify that autonomous actions respect their controls. Existing staff are well placed to grow into these roles, given practical training, mentoring and protected learning time.
From tiers to functions
As autonomy matures, the familiar Tier 1/2/3 structure can evolve into three complementary functions: Service Reliability and Prevention, Automation and AIOps Engineering, and Incident, Change and Autonomy Governance, with NOC leadership accountable for service outcomes and workforce development across all three.
The experience carries over. Tier 1 operators bring the instincts needed to supervise autonomous operations and review trends. Tier 2 domain engineers become reliability engineers, automation engineers and data stewards. Tier 3 specialists contribute across all three functions, from problem prevention to AI assurance. Incident and change managers expand into change-risk and policy governance.
What autonomy does, and doesn’t, mean
Mapping this to the TM Forum autonomy levels keeps everyone honest. At Level 1, automation helps people see better, removing screen-switching and ticket administration. At Level 2, closed loops handle known, bounded scenarios, each with explicit eligibility rules, approved actions and rollback conditions. Levels 3 and 4 bring adaptive domain loops and then predictive, cross-domain, service-driven operation.
Two principles hold throughout. Using AI is not, by itself, evidence of autonomy; what matters is how reliably insight connects to action within governed boundaries. And even at Level 5, “full autonomy” does not mean zero people. Someone still sets service objectives and risk limits, engineers resilience against new failure modes, maintains the AI and data platforms, and leads when the autonomous system itself is in trouble.
The real opportunity
The goal isn’t just a NOC that resolves incidents faster. It’s a network that experiences fewer avoidable incidents in the first place, run by people whose expertise has grown alongside the technology. That outcome is achievable, but only if operators plan for their people as carefully as they plan for their platforms.
I will explore the role transitions and maturity gates in more detail in our session, Autonomous Networks + Human Intelligence: Rethinking the NOC, at FutureNet North America in New York, November 4-5, 2026. The VIAVI team will be at Stand 8 for anyone who wants to continue the conversation.
