In 2024, almost every service provider featured in Gartner’s Magic Quadrant for global WAN services and Magic Quadrant for managed network services reports confirmed they had begun using artificial intelligence (AI) in multiple areas of enterprise network operations. Key applications include AI for IT operations (AIOps), generative AI (GenAI) as a network assistant, improved service delivery, and AI within secure access service edge (SASE) and network security.
AIOps has become a core capability in managed networking. Major providers like HCLTech, Microland, and NTT Data are now combining AIOps with network automation to streamline service onboarding and enhance the customer experience. Additionally, these providers are applying AI and machine learning (ML) to monitor network health, identify anomalies, and automate routine tasks in network operations centres (NOCs).
The aim is to move from fixing problems after they happen to preventing them before they occur. For instance, if latency on a wide-area network (WAN) link begins to spike irregularly, a machine learning model could detect this pattern as an early warning of a potential link failure and either notify engineers or initiate a failover before a serious outage takes place.
Tata Communications is one such provider that has invested in AI-based fault diagnosis, achieving 85% accuracy through AI/ML, while its AI-driven telemetry helps predict and resolve issues for proactive network monitoring.
Furthermore, many network equipment vendors are now building AI features directly into their products to help service providers with network monitoring.
GenAI as a network assistant
Over the past year, Gartner has observed strong interest from managed network service (MNS) providers in using GenAI for IT operations, including network management. The goal is to create a network AI assistant that can communicate with operations teams through a natural language chat interface, assist with troubleshooting, document network setups, and even make changes by generating configurations from high-level intent.
HCLTech is one example, focusing on integrating GenAI with software-defined wide-area networking (SD-WAN) to achieve full automation across lifecycle operations. The company is developing a supplier-focused GenAI large language model (LLM) as part of its service delivery platform (SDP).
Enhanced service delivery
AI is also being used to improve customer-facing aspects of MNS. Providers are increasingly adopting AI to enhance support and transparency for their clients. This includes AI-powered customer service bots, service portals, and AI-generated reports or insights.
For instance, many MNS providers included in the Gartner Magic Quadrant for managed network services report use bots—now frequently enhanced with AI—to handle repetitive tasks. Some providers deploy thousands of these bots as part of their network automation systems.
AI in SASE and network security
AI and ML are proving just as vital for security in MNS as they are for performance management. Many providers, such as XTIUM and Microland, promote AI-enhanced network security solutions that use advanced analytics, AI, and GenAI to strengthen and simplify security across local area networks (LANs), WANs, and cloud environments.
In SASE and network security, AI enables automated anomaly detection—for example, isolating a suspicious device or triggering multifactor authentication for a user showing unusual behavior.
For policy optimization, AI can recommend tightening or adjusting security policies based on observed usage patterns. It might suggest zero-trust rules for a specific application, taking into account context like location, time, or department.
Advanced providers like HCLTech are exploring the use of LLMs to support security analysts—for instance, summarizing complex, multi-step attacks or even generating firewall rules from high-level threat descriptions.
Many SASE platform vendors also highlight their AI/ML capabilities. Versa Networks, for example, promotes its AI/ML-powered unified SASE solution that combines SD-WAN and cloud security, using ML to continuously adapt to changing network conditions and threats. Similarly, Cato Networks emphasizes its use of AI/ML across its cloud-native SASE service to deliver “reliable, accurate network security,” applying advanced data science to threat prevention and intelligent traffic management.
AI in MNS in 2028 and beyond
The integration of AI into MNS will continue to boost operational efficiency and support smarter decision-making, ensuring networks remain robust and flexible enough to handle evolving demands and traffic patterns. Over the next three to five years, significant changes in MNS are expected as a result of widespread adoption of AI—including traditional, generative, and agentic forms—alongside automation.
Widespread NOC assistants
Given the current pace of development, GenAI is expected to become a mature and trusted assistant in network operations by 2028. The experimental and early-stage deployments seen between 2023 and 2024 will evolve into robust network AI assistants fully integrated into MNS workflows.
These assistants will communicate via natural language (text or voice) and connect with monitoring and ticketing systems. They’ll be able to answer complex network questions, draft change plans, and summarize incidents and problems.
In essence, if 2023 marked the introduction of network AI assistants (see What is a network AI assistant?), by 2028 they will be a standard tool in NOCs for boosting productivity.
The models powering these assistants are expected to become more specialized in network engineering and fine-tuned with each provider’s historical data, making them more accurate and context-aware than today’s tools.
Top providers will leverage proprietary models—or at least proprietary fine-tuning—as part of their intellectual property. For example, a provider might use a model trained on years of network event data that excels at diagnosing telecom network issues or evaluating network security design effectiveness. This will set them apart from competitors relying on off-the-shelf AI assistants.
By 2028, agentic AI will likely appear as automated “Tier 0” responders in NOCs. These AI agents will be capable of detecting network incidents, understanding intent, making independent decisions, and carrying out actions to resolve specific tasks and incident types end-to-end without human involvement.
By 2028, many providers are expected to have fully automated remediation for known issues. For example, if a branch SD-WAN router goes offline, the AI agent could detect the incident, decide on a sequence of fixes—such as restarting a virtual instance or failing over to a backup—and execute them. It would only alert a human if those steps fail.
Another scenario might involve detecting a known bug, like a memory leak in a firewall causing performance issues. After identifying the problem, the AI agent could apply a temporary configuration workaround or initiate a software patch.
This goes beyond today’s static scripts by adding autonomous decision-making and action. The agent can use machine learning to verify whether the issue matches a known pattern and check—via policy—whether it’s safe to proceed (for example, rebooting only after business hours if the issue is critical).
Fully autonomous networks will likely remain out of reach until well after 2028. However, by 2028, self-healing actions are expected to be accepted for narrow use cases, as providers will have built trust in AI for these repetitive tasks through extensive training and proven results.
Still, the complexity of coordinating across domains means humans will continue to handle high-level decisions. But for routine faults and performance adjustments, automated agents could become standard practice, improving overall service reliability.
This article is based on an excerpt from Gartner’s report AI will transform managed network services in the next three years, authored by Gartner senior director analyst Gaspar Valdivia.
