Agentic AI holds the promise of making networks smarter and staying steps ahead of impairments, all while taking stress off the people tasked with running them. But as companies begin handing over control of network operations to AI, the amount of responsibility to cede to digital teammates remains a question of trust.
Having trust in AI is critical, but it could become a moving target as agentic AI is more widely adopted in network operations, according to a new Omdia study backed by Cisco data and analysis. Omdia is a sister company to Light Reading.
The question about whether agentic AI is being deployed is already “outdated,” based on the level of adoption that’s underway, according to the study, which was based in part on a survey of about 1,000 IT and network operations professionals working in companies with more than 500 employees. It found that nearly three-quarters of organizations have deployed AI for network operations, with 51% saying that agentic AI systems already in use “act” rather than just advise.
And even with today’s guardrails, four in five survey respondents said they are comfortable granting AI a “high or fully autonomous role” in network operations, including 24% who said they are comfortable with fully autonomous roles that don’t require human oversight.
Additionally, 84% of the people surveyed expect their companies to implement an “AI-led” operating model within the next 12 months.
IT can’t hire its way out of the problem
The study found that the number of alerts related to the network is becoming too overwhelming for humans to monitor, process, diagnose and act on without some help from agentic AI.
The average organization generates about 4,100 monitoring alerts and events per day, with more than half (51%) being network related, said the study. Omdia and Cisco reckon that a typical network practitioner is capable of reviewing, investigating and resolving about 21 network alerts per day, meaning that clearing the daily backlog manually would require a team of about 100 specialists – a lot more than most companies employ in network operations teams.
“IT is managing more with fewer people than ever before,” Joe Vaccaro, SVP of network platform and assurance at Cisco, told Light Reading. “And network connections have become more complicated than ever before. IT just can’t hire their way out of this anymore.”
AI must show its work
The growing adoption of agentic AI comes amid headlines and concerns about AI systems going rogue before humans become aware. As a result, establishing trust by implementing security and governance frameworks has never been more important.
“You need a level of trust as you bring [agents] into the fold as you would with a new junior network operations engineer,” said Vaccaro.
And humans still need to stay in the loop to supervise the work of the agents and audit their output. A full 99% of survey respondents said they require some type of guardrail before trusting AI to act, and 69% said they demand tracking of the agentic actions that are taken.
“The core thing we see through the survey … is the importance of [AI agents] showing their work,” Vaccaro explained.
While trust is a huge factor, the study also found that embedding AI into the operations of the company is critical to overall productivity. For example, companies with the highest levels of AI adoption in network operations are close to 30% more productive per employee than their less mature peers, according to the study. This suggests that AI maturity is becoming a competitive advantage rather than just a path to operational improvement.
“[The] biggest barrier might not be technology in the first place,” Vaccaro said. “It might be organization readiness.”
