By Ernest Lefner, Chief Product Officer, Gluware
The data is clear: autonomous network configuration is technically possible, but organizationally, culturally, and philosophically, most teams aren’t there.Â
There’s a version of the AI-powered network that vendors love to describe. Autonomous agents detect anomalies, diagnose root causes, push configuration changes, and self-heal, all before a human engineer even gets a Slack notification. It’s a compelling vision. It’s also almost entirely fiction for the vast majority of network teams operating today.
Here’s what the data shows.
According to the Network Automation Forum’s (NAF) 2025 State of Network Automation Survey, one of the most comprehensive studies of its kind with 681 respondents across 58 countries and 34 industries, only 1.52% of network organizations have deployed fully autonomous remediation. Another 4.56% are in semi-autonomous territory. That means roughly 94% of network teams have not crossed the threshold into autonomous execution.
It’s not because the automation stack is absent. Configuration deployment automation is fairly common with the automation-forward organizations that NAF tracks, with 30.61% of respondents using it. Deployment-with-validation pipelines, drift detection, monitoring automation, these are real and growing. The tools exist. What doesn’t exist, for most organizations, is the organizational and cultural trust required to let them run unsupervised.
On AI specifically, the picture is starker. Only 3.04% of respondents have AI or LLMs deployed with access to production network data and systems. Another 13.5% are in development or testing. Meanwhile, 44.68% report no current plans or initiatives around AI at all.
The survey’s author, Chris Grundemann, summarized it with a line that deserves to be quoted directly: “Culture eats Automation for Breakfast.” The number one barrier cited was not budget, tooling, or vendor gaps. It was people, specifically a combination of skills gaps (27%), organizational resistance (20%), and cultural friction (14%) that together account for the majority of what’s holding teams back. Over half of organizations are also misaligned with their own leadership on the state of their automation capabilities.
And keep in mind that NAF survey respondents are probably the most automation-forward organizations.
The Trust Deficit Runs Deep
Zoom out to the broader IT landscape and the picture looks similar. In a September 2025 survey of 360 IT application leaders across organizations with 250 or more employees in North America, Europe, and Asia/Pacific, Gartner found that only 15% are currently considering, piloting, or deploying fully autonomous AI agents, defined as goal-driven tools that operate without human oversight. While 75% of organizations are experimenting with some form of AI agent, the leap to full autonomy remains rare.
The barriers Gartner identified are specific and telling:
- Only 19% had high or complete trust in their vendor’s ability to provide adequate hallucination protection
- 74% believe AI agents represent a new attack vector into their organization
- Only 13% strongly agreed they have the right governance structures in place to manage AI agents
- Only 14% strongly agreed that IT, business users, and leadership are aligned on what problems AI will actually solve
These aren’t abstract concerns. Network infrastructure is mission-critical. In network operations, a misconfigured BGP route, a firewall rule applied at the wrong scope, or an ACL pushed to the wrong interface can cascade across an organization in minutes. The blast radius of an autonomous mistake is asymmetric and often requires significant remediation effort to undo.
The Governance Gap Will Claim Victims
Gartner’s May 2026 analysis goes further. It predicts that by 2027, 40% of enterprises will demote or decommission autonomous AI agents due to governance gaps identified only after production incidents occur. The root cause, according to Gartner: organizations are treating AI agent governance as binary, either completely locked down or fully trusted. Neither extreme works.
Gartner’s proposed framework is instructive. It identifies four meaningful autonomy levels: agents that observe (read-only), agents that advise (recommendations for humans to act on), agents that act with approval (execution gated by explicit human sign-off), and agents that act autonomously (within defined guardrails, with humans reviewing exceptions). Most network teams, based on the survey data, are somewhere between Level 1 and Level 2. Almost none are operating at Level 4.
Critically, Gartner flags a risk even at Level 3, act with approval, that practitioners know well: approval fatigue. When human review becomes a rubber stamp under time pressure, “a false sense of safety” is created while the actual risk surface expands. Meaningful human oversight requires meaningful human engagement, not just a checkbox in a workflow.
Vibing is Not Yet Enough On Its Own for NetOps
The convergence across these sources points to the same practical conclusion. The path toward greater autonomy requires three things to mature simultaneously:Â
- AI outputs that are grounded in accurate, real-time understanding of network state rather than stale or hallucinated context
- Explainable actions that show reasoning network engineers can actually evaluate
- Governance frameworks that are proportional to the autonomy level being granted, not uniform across all agent types.
The vendors pushing “autonomous networks” as a near-term reality are getting ahead of reality. The network professionals actually running production infrastructure are telling a different story: they want AI that earns trust incrementally, proves its reasoning, and does not bypass the human approval step just because it is technically capable of doing so.
Vibing your way through a network change is still risky today. The future belongs to AI that is grounded, validated, and held to account. Before VibeOps can be trusted, it needs a foundation in GroundedOps.
If you want a grounded path towards agentic NetOps, check out how Gluware’s DIAL technology and Titan AI offer deeply validated, safe, and predictable agentic network automation. Learn more or request a demo today.
