Cisco and Omdia published research on agentic AI in enterprise networks, saying 51 percent of large organizations now run agentic systems that make autonomous production changes. The same research says more than 84 percent of technology executives expect to run an agent-directed operational framework within the next 12 months.
Alert loads are already pushing operations teams toward automation
The research says corporate environments now register an average of 4,100 monitoring alerts per day, with 51 percent tied directly to network infrastructure. Operations personnel manually review and resolve about 21 incidents each workday, and the analysis estimates that handling that volume without automation would require roughly 100 dedicated technical specialists per enterprise.
Operations departments discard 46 percent of incident alerts without preliminary investigation, and engineering personnel spend 45 percent of active troubleshooting hours on false positives. Production routing and policies are revised daily at 59 percent of businesses, while 29 percent push those changes multiple times each day.
What Cisco and Omdia say autonomous systems are being asked to do
About 82 percent of engineering leadership teams allow autonomous systems to make production adjustments without prior human sign-off for defined tasks. Permitted automated actions include dynamic traffic rerouting, wireless parameter tuning, and isolating compromised hardware endpoints.
Joe Vaccaro, SVP and GM of Network Platform and Assurance at Cisco, said agentic AI is a fundamental paradigm shift from running operations to orchestrating intent. He said agentic AI for NetOps requires trust built on visibility into every decision, explainable context behind every recommendation, and guardrails that ensure deterministic outcomes.
What the numbers say about the operating model
Ninety-five percent of engineering leaders say existing AIOps packages fall short. Cross-domain faults account for 92 percent of major performance failures, and support staff correlate evidence across an average of ten separate management interfaces to find root causes.
The mean resolution time is 88 hours, while the median is 12.5 hours. Traffic to machine learning services doubles every six months across campus and branch installations, per-client consumption is growing by 54 percent, and autonomous software agents generate up to 450 percent more network traffic per completed workload than human-prompted queries.
Read for enterprise NetOps teams
The research gives enterprise network teams a benchmark for how far agentic control has already moved into production tasks, and it ties that shift to specific actions such as rerouting traffic and tuning wireless parameters. It also shows that governance requirements are not optional for buyers: 69 percent want detailed reasoning trails, and 36 percent require end-to-end tracing plus immutable post-execution audit logs.
That is the part procurement teams and platform owners will compare against existing AIOps tooling, especially where monitoring is still split across compute, storage, security, and networking.
What mature deployments are reporting
About 86 percent of enterprise operators want monitoring consolidated across compute, storage, security, and networking into a single platform. Mature enterprise deployments record a 30 percent uplift in individual staff productivity and maintain 3.2 times greater visibility across infrastructure perimeters.
Mark Rodrigue, Senior Network Engineer at Room & Board, said questions that once required manual searching across dashboards and data sources now return in minutes with the executive summary first and supporting evidence underneath. He said the system shows its work, allowing him to follow the logic and evidence behind every recommendation.