A Gartner analyst recently suggested that GenAI tools could create 25% time savings for network operational teams.
Where might these time savings come from?
How are GenAI tools helping NetOps teams today, and what other tasks might they take on in the future as models continue improving?
In general, these savings come from automating or streamlining manual NetOps tasks. 65% of enterprise network activities were still done manually in 2023, but as networks get more complex, it gets harder and harder for engineers to keep up. Specifically, GenAI can make it easier and faster to interact with the network, shift some NetOps tasks to more junior IT people, and create "first drafts" of networking materials.
Generate "First Draft" Static Network Designs
GenAI could generate basic network designs (via a natural language interface) if given info like SLAs, number and types of applications. This initial AI-generated design will likely need considerable work by a human network architect. Enterprise networks are complex and confusing and often have many years of built-up technical debt or unique requirements that must be met. This makes each one unique. Human engineers are required to customize each network to fit. But starting with a draft instead of from scratch will save time, although it will be out of date fairly quickly without a live network model.
The benefits here will likely increase as AI models improve. Current AI models have limited networking knowledge. But as the cost of training models comes down, perhaps in the future a "networking expert" AI model could be developed that can create better designs and further increase time savings.
Analyze and Interpret Raw Data
GenAI excels in reading, interpreting, and generating human-readable content based on raw data. Once a network change has been made through an automated script (using Ansible for instance), GenAI can take the raw output (CLI results, logs, or automation results) and interpret them. For instance, if the CLI output contains a list of errors, GenAI could analyze the results, correlate them with known issues or network states, and summarize the findings in a more human-understandable format. It could also answer questions about the results, or provide observability by explaining why a particular alert was generated.
Make it Easier to Interact with the Network
Many networking IT vendors are adding chatbots or AI assistants to their products to make them easier to use. This lets IT use the tool through a conversational interface with text, audio, video, or graphics. Rather than using the command line interface (CLI) or normal product dashboard, users can interact with the network with natural language. For example, they could type "Check the overall health of the devices on this network map and summarize the results in a table." This saves significant time for NetOps the same way a tool like ChatGPT can save people time in other fields.
This makes it easier to get network data and to make network changes. Users don’t need to know the CLI commands to do things like find a specific security camera or find an A-B path. They can just ask for it. This makes troubleshooting faster across the board and allows IT staff outside of the network team to get network data themselves rather than emailing NetOps and asking for it. This enables self-service options — in fact, some organizations do this to let Help Desk employees troubleshoot network issues without escalating to the network team, or to allow the security team to look up the location and IP addresses of devices involved in a security incident. As more tasks get handled by more junior employees, (shifting them to the left in a diagram of the usual troubleshooting process) the entire organization becomes more efficient.
Pair GenAI with Other Technologies
GenAI can create even more time savings when paired with other technologies. Here are two that synergize well with it. First is Network Digital Twins. It’s difficult for an AI to interact with the network directly because the commands for different network devices are not standardized (and every enterprise network uses different devices). The AI likely won’t know the differences between Palo Alto and Cisco’s firewall APIs, for example. A digital twin of the network allows the AI to get accurate network information in a standardized way. This means it can generate accurate device lists, network maps and A-B paths — making the AI better across the board.
Second is a strong library of network automations. Automation is a powerful tool for assessing the network and checking if rules, configurations or security policies have drifted away from their intended state and pushing out changes and fixes. GenAI tools or chatbots can orchestrate these automations to gather data and perform tasks more easily. They can be the interface between humans and automation.
Finally, Agentic AI will offer new opportunities as it becomes more widespread. Agentic AI is good at executing tasks autonomously based on predefined rules, triggers or specific requests. It can retrieve basic device properties (IP lookup, L2 & L3 neighbor), run CLI commands, interpret the results, and then take further actions. For example, it could run commands to check the status of routers, switches, or firewalls and then adjust configurations if it found problems.
Implementing GenAI tools in NetOps should begin with calculating the business value the tool will bring. Get specific about how it will reduce MTTR, reduce downtime, or increase the rate at which tickets are closed. Then do a PoC to get familiar with the process, prove success to others, and judge how accurate the AI is (and how much human verification is needed).
Anecdotally, I saw many IT and NetOps teams testing out GenAI use cases throughout 2024. I expect many of these use cases will be implemented in earnest throughout 2025. We’ll start to see what the real time savings and benefits of GenAI are for IT teams.