Cisco deploys agentic AI in network operations with an emphasis on deterministic decisions
Cisco describes deploying AI agents in network management using a four-phase framework (sense, diagnose, plan, act). Determinism in the action phase is ensured by predefined workflows so that AI agents do not improvise network changes.
According to Cisco, enterprises are increasingly accepting situations where AI agents make changes to the network without human approval. According to Joe Vaccaro, SVP and GM of Cisco's Network Platform and ThousandEyes division, the reason is the volume of alerts in network operations that people can no longer process manually – according to him, companies could not hire enough people to handle it.
Cisco uses a four-phase framework when deploying AI in network operations: sense (detecting the problem), diagnose (diagnosis using telemetry across domains), plan (remediation plan), and act (carrying out the intervention).
A challenge remains, however: networks need deterministic, not probabilistic, outcomes. According to Vaccaro, Cisco addresses determinism specifically in the act phase by using predefined workflows instead of allowing AI agents to improvise changes to the network. The goal is for the actual execution of a network change to proceed with a predictable, deterministic outcome.
The source does not provide further details about the specific deployment, timeline, or other companies involved. See the source article for details.
Why it matters
For companies running larger networks, this approach illustrates a concrete way to combine AI agents with risk control: AI is allowed to diagnose and propose, but the actual execution of a change follows predefined workflows rather than free-form improvisation. This reduces the risk of unexpected outages caused by unpredictable AI decisions.
Two audiences, two different impacts
What this means
For individuals
For network and operations specialists, the role is shifting from manually resolving individual alerts to overseeing AI agents and defining the workflows within which agents are allowed to act.
For a business
Companies can reduce the burden on network operations by having AI agents handle the volume of alerts that would otherwise require an unsustainable increase in headcount – provided that the actual execution of network changes is bound by predefined workflows rather than free-form AI decision-making.
ProcessesCheck the original
Event sources
only one source so far · 1 publisher, 1 independent. We count feeds from the same owner only once.