Aviatrix Blog 
AI is the gold rush of modern cloud networking: everyone’s excited and anxious to cash in, but there’s enough misinformation, confusion, and deception about AI to create some chaos. The real, actionable intelligence you can get from AI is pure gold, but there is a lot of gold spray-painting going on with solutions that don’t really use AI to analyze patterns or make predictions. The practice of sticking an AI label on a feature or tool that doesn’t really merit the term has become known as “AI-washing.”
In the journey to AI Ops, how do you spot the real gold and move away from AI-washing towards actionable artificial intelligence? I’ll go over some basic AI terminology and then outline four key milestones in the journey towards an advanced, predictive use of AI.
What is AI? Standard Terms and What they Mean
Sometimes, you can spot AI-washing just by knowing fundamental AI terminology. Here’s some vocabulary to start with:
- Heuristics — Generally, heuristics are mental shortcuts that help people make decisions quickly. Heuristic AI creates shortcuts that guide the search for solutions more efficiently than exhaustive search methods. In a complex network, heuristics can help teams identify anomalies quickly.
- Machine Learning — This type of AI is about training machines to make decisions, set up automation, and generally think more like a human.
- Automation — Automation is technology used to perform tasks with little to no human intervention. In networking, you can use automation to replicate environments or auto-scale gateways and storage containers based on traffic and usage.
- Predictive AI — With predictive AI, you use machine learning to analyze patterns and predict future behaviors. In networking, you can use predictive AI to forecast patterns such as traffic flows.
- Generative AI — With generative AI, you use machine learning to generate new text, images, or other content based on multiple examples of the same type of content.
Milestones on the Journey to Actionable AI
To differentiate organizations or solutions that are still limited to these areas from those who are actually journeying towards predictive AI, check if they’ve reached one or more of these four milestones:
Milestone 1: Reactive Automation
Reactive automation is the kind of logic, rule-based automation we see in networking, if-this-then-that: for example, scaling up or down with scale groups, SQL queries for metrics, automated alerts, and scripts querying systems to understand what’s going on. It’s very reactive and less predictive.
An example of this would be the implementation of reactive automation in a cloud infrastructure to manage sudden traffic surges. When a website experiences a sudden increase in traffic while measuring by the concurrent network sessions, the system automatically scales up the resources to handle the load. This reactive automation ensures a seamless user experience and prevents potential downtime.
Milestone 2: Enhanced Visibility Through Natural Language Processing (NLP)
Natural Language Processing (NLP) is like putting a bot interface or a chat interface over the front of existing data. An example of NLP-enhanced reactive automation would be to give a command: “show me yesterday’s peak latency.”
Most companies have achieved this level of automation. They may send out a press release about AI Ops, but all they’ve really done is put a natural language engine over the top of the metrics, which is likely a time series database running in the background. It’s not doing much that’s predictive. It’s just allowing you to type in human words or natural language to talk to the data that was already there.
This type of automation is sometimes useful for picking out trends instead of having someone learn your filtering system on your product. You can ask questions like, “who are the heavy hitters of egress traffic?” and it will give you an answer. Many people have reached this level of AI use, but are stuck there.
Milestone 3: Machine Learning for Predictive Analytics
This type of AI provides forecasting instead of automatic responses. For example, it can show you that traffic goes up on Fridays. It can watch patterns and know what will happen based on past data. It’s useful for predicting congestion at certain times or equipment failures, making it possible to do things like proactive maintenance.
An example of this use of AI would be using machine learning capabilities to predict network congestion during major events, such as live sports broadcasts. By analyzing historical data, the system can forecast peak traffic times and allow for proactive resource allocation, ensuring a smooth streaming experience for viewers.
We’re not seeing a lot of this type of automation yet, but some organizations are getting there.
Milestone 4: Augmented Decision-Making in Operations
With this type of AI use, AI can make suggestions for optimization and configuration. For example, AI can propose routing based on current and forecasted traffic, class multiple known vulnerabilities, and find anomalies in traffic patterns.
AI is much better at looking at large data sets than humans are, so this is where AI becomes very useful at spotting potential issues. For example, healthcare organizations are beginning to use this type of AI to automate repetitive tasks and perform precision diagnostics such as radiotherapy planning. In cloud networking, it can recommend that you tweak a security policy or monitor a traffic flow until you can make more informed decisions.
An example of this milestone would be leveraging augmented decision-making capabilities to optimize network security. The AI system can identify unusual traffic patterns that indicate a potential security threat. Based on the AI’s recommendations, the IT team can adjust security policies and monitor the traffic flow, preventing possible data breaches.
Aviatrix: Turning AI Insights into Real-Time Predictions
While Aviatrix offers features and solutions for the other three types of automation, this fourth type of automation, Augmented Decision-making in Operations, is where we’re really beginning to shine. Our Cloud Perimeter Security solution uses AI recommendations to offer health scores for your network, monitor traffic flows, and provide intelligent insights.
- Natural Language Processing — We offer a simplified, intuitive user interface where you can query data in human languages.
- Security Posture Health Scoring — Our solution evaluates your network security and gives you a score to help you improve it.
- Intelligent Insights — We help you monitor traffic for suspicious activity and unnecessary expenditures, helping you keep your network safe and manage costs.
These features are available through our new Aviatrix Platform-as-a-Service (PaaS) offering, which provides all the value of AI-enhanced insights as we manage backend maintenance and upgrades for you.
Want to explore how Aviatrix’s AI-driven insights can help your network?
- Schedule a demo to learn more about the AI enhancements in our solution.
- Learn more about Aviatrix’s integration with Microsoft Copilot for Security, which uses AI insights to help you protect your network.