As networks grow increasingly complex and distributed, the benefits of deploying artificial intelligence (AI) technology are becoming increasingly evident. In short, AI is poised to fundamentally change the way networks are monitored and managed.
A key AI benefit is rescuing skilled network teams from routine and mundane tasks. "AI can help monitor the health and configuration of the network, identifying anomalies and potentially taking corrective actions automatically," says Marc Herren, network advisory director with the technology research and advisory firm ISG, in an email interview.
More importantly, the emergence of software-defined WANs (SD-WANs) is opening the way for network managers to integrate AI technology into network operations and management. "For the industry to deliver on the promise of a self-healing or self-correcting WAN, AI tools can help automate routine network operation tasks, set policies, measure network performance against set targets, and respond to and rectify the networks as needed," Herren explains.
AI can also make snap decisions to remediate a variety of serious network issues. "Although human operators can more effectively triage complex and multi-step problems, AI is a powerful tool that can supplement the work of network engineers to add robust controls and automation to mature networks," says David Brauchler, a principal security consultant with cybersecurity and software assurance services firm NCC Group via email. "AI should be considered an addition to a company’s network team rather than a replacement, accelerating the work of engineers and creating new efficiency improvements to developed workflows."
When laying the groundwork for AI network management, it's necessary to understand the network's infrastructure, devices, and connections and to evaluate data sources and data flows, says Portia Crowe, chief data strategist with Accenture Federal Services' defense and applied intelligence unit in an email interview. It's also important to understand your team's talent level and AI expertise. "Exploration and implementation on a small scale will get you started and provide an ability to iterate and learn, as well as to capture metrics to help with scalability," she advises.
Before deploying any AI tool, it's important to identify existing network needs as well as the tools currently being used to address those requirements. "In many cases, organizations have lower-hanging fruit that should be addressed before considering the move to AI-powered toolsets," Brauchler observes. Meanwhile, network engineers and operators should be trained to react and respond to network outages, threats, and other anomalies without complete dependence on AI technology.
While AI isn't a "'silver bullet,'" it does provide powerful tools that can help organizations make data-driven decisions about their network, Brauchler says. "Autonomous AI behavior should be validated, and the final say should be given to humans behind the screen who can pivot and course-correct when the tools encounter a situation they weren’t prepared to handle." Meanwhile, it's always important to remember that any AI network tool is only as powerful as the quality of the data fed into it, and a poorly managed network now will typically lead to a network that's poorly managed by AI in terms of both stability and security.
Herren warns that current network engineers may lack the necessary skills to successfully implement AI capabilities into network management. "Upskilling existing staff or hiring new staff with the required skillset may be lengthy and expensive," he notes. Organizations can mitigate those challenges by looking outside of their organization to vendor-provided solutions and managed service providers that already have the staffing, skillset, and experience to successfully implement AI-driven network management.
Considering its rapidly continuing advancement, Herren predicts AI is destined to become a critical network management technology. Meanwhile, network engineers must become accustomed to training AI software on existing network data. Understand, however, that even the best-trained AI software can sometimes spew out incorrect results, so training and retraining will be a continuing process for developer teams.
Many more network areas will benefit as AI technology evolves and improves. "This includes edge network management, where reduced processing and latent-deficient networks can be optimized based on AI findings," Crowe says. For sensitive and critical tasks, a need may eventually emerge for human-machine AI collaboration. Also, as networks become more complex and distributed, AI will likely begin managing multiple data sources and flows. "Humans will still play a critical role in data quality, data labeling, and data trustworthiness," she predicts.
It's important to be realistic about AI's potential value as a network tool. "Understand that AI doesn’t think on its own—it's software that's trained to look at existing datasets and to come up with an actionable output for specific events," Herren says. If the data fed into it is wrong or coded incorrectly, it could spit out wrong actions. This could degrade network performance or security and potentially bring down the entire network, he notes. "Applying the right controls and validating the actions at regular intervals, just like you would if the operations were run by humans, is critical for successful inclusion of AI in network automation."
Related articles:
Modernization of the data domain to include new practices, tools, and technologies is the third phase of digital transformation.
Think of Hypershield as an AI-enabled security fabric where every network port becomes a high-performance enforcement point.
With the rise of AI and its massive impact on networking, Cisco has pivoted swiftly to enable customers to utilize its converged networks to run taxing AI apps.
