New EMA Research Discovers High Hopes and
Hard Lessons with AI-Driven NetOps
by Shamus McGillicuddy, VP of Research, EMA
The hype around artificial intelligence (AI) is out of control. It seems like a million influencers jumped off the cryptocurrency bandwagon this year and landed squarely on AI. Not long after OpenAI released ChatGPT-4, its buzzy, all-purpose chatbot, the company raked in a reported $10 billion investment commitment from Microsoft. Meanwhile, pundits started pondering whether AI would steal the jobs of millions of information workers since it automates everything from essay writing to software development.
As industry analysts at Enterprise Management Associates (EMA), our job is to look past the hype and discover how our clients can benefit from emerging technology. I am focused on technology and trends that impact network infrastructure and operations professionals.
My new research report, “AI-Driven Networks: Leveling Up Network Management” does just that by examining how AI, machine learning (ML), and so-called AIOps solutions can help IT organizations optimize and automate their networks. EMA found that there remains a gap between hype and reality.
Based on a survey of 250 IT professionals who have experience with AI-driven networking solutions, our research found that 92% of IT stakeholders believe that applying AI technology to network infrastructure and operations can potentially lead to better overall business outcomes for a company. However, only 36% of research participants believe they have been completely successful with AI-driven network management so far. IT pros clearly see the potential value of AI-driven networks, but they are also learning some hard truths as they experiment with it.
Finding AI Solutions for Network Management
EMA asked respondents to reveal how they consume or plan to consume AI for network management. A new market of domain-agnostic AIOps vendors emerged in recent years that deliver capabilities across the entire IT stack. However, most organizations are looking to networking vendors for AI-driven networking capabilities. Only 47% of respondents told EMA that they are relying on domain-agnostic AIOps solutions for network management. Instead, 77% are looking for AI/ML features from their network management vendors and 79% are looking for such features from their network infrastructure vendors. In fact, most of them now believe that AI/ML features are not product differentiators for network management and network infrastructure products.
The Benefits of AI-Driven Networking
Research participants identified the primary benefits to applying AI/ML and AIOps to network management. First, it helps them optimize the overall network for better performance. Second, it makes the network more agile, allowing IT pros to respond more quickly to the needs of the business. Third, it reduces overall security and compliance risk. Finally, it makes the network more resilient, reducing overall downtime.
While many respondents saw room for success with how they achieve these benefits, EMA saw signs that AI-driven network management is delivering some return on investment already. For instance, 69% of research participants reported that overall end-user experience has improved since they started applying AI to their networks.
Roadblocks to Success
The first stumbling block with AI and networking is product evaluation. Only 40% of respondents felt that their organizations were fully effective at evaluating the AI/ML and AIOps solutions that they apply to network management. They also told EMA that explainable AI features that can illustrate how an AI system derives insights from a network through natural language, graphics, and other methods can help them close the evaluation gap.
Respondents also told us they are struggling with security and compliance risk. They are worried about how safe AI is for network management. Will these systems make mistakes that open up vulnerabilities? Will sharing network data with a cloud-based AI system pose a risk? These are questions that network teams are asking right now.
Network teams also said that the complexity of their existing networks is posing a challenge to AI solutions. In the past, EMA found this complexity to be a common challenge for network monitoring, too. When a network is extremely complex, IT organizations struggle to identify how to extract sufficient telemetry to provide good insight into the state of the network. Organizations are struggling to understand whether they are providing the right data to AI systems to ensure that the system can deliver value.
Regardless of the challenges, hype continues to win out. Most IT pros anticipate a bright future for AI-driven network management. In fact, 59% told us that it is inevitable that AI solutions will lead to fully autonomous networks, where human intervention is rarely needed for managing and troubleshooting a network.
To learn more about this research, attend our upcoming free webinar, AI-Driven Networks: Leveling Up Network Management
.