We’ve all spent plenty of time looking at network diagrams that were out of date before the ink was even dry. You’re trying to figure out if a specific security policy is actually doing what it’s supposed to, or if a new application path is going to break something else. Usually, that means logging into half a dozen different CLIs, cross-referencing spreadsheets, and maybe even doing a bit of manual packet tracing if you’re really lucky. It’s tedious. It’s what we call “toil,” and it’s where most of the risk in our modern infrastructure lives. The gap between what we think the network is doing and what it’s actually doing is where the problems hide.
During the recent AI Infrastructure Field Day, Forward Networks showed off how they’re using generative AI to basically delete that toil. They call it Forward AI. It’s a conversational interface that sits on top of their network digital twin. Instead of writing complex queries or hunting through dashboards, you just ask a question in plain English.
Can this host talk to that database?
It sounds like magic, but it’s math.
The Ground Truth is Mathematical
To understand why a chatbot for your network actually works here, you have to look at what’s under the hood. Forward Networks doesn’t just “discover” your network; they build a mathematically accurate digital twin of it. They’ve been doing this for over a decade, grounded in rigorous research and proven through years of product shipping. They collect every bit of configuration and state data from every device that moves a packet. Switches, routers, firewalls, load balancers, it doesn’t matter if it’s on-premises iron or running in AWS, Azure, or Google Cloud. They take all that data and break it down into mathematical primitives. This is the “ground truth.” It’s a provable model of every possible path a packet could take through the network.
When you ask Forward AI a question, it isn’t just guessing based on a large language model’s training data. We’ve all seen LLMs hallucinate; you can’t have a network tool hallucinate that a firewall rule exists when it doesn’t. Forward AI uses the LLM to understand your intent, but then it queries that mathematical model to get the answer. It’s combining natural language with absolute certainty.
From Symptoms to Diagnostics
I like the analogy they used during the presentation. Most of our current monitoring tools are like measuring symptoms. We look at packet drops or latency, which is like checking a patient’s temperature. It tells you something is wrong, but it doesn’t tell you why. The digital twin is more like a full-body MRI. Because the model is mathematical, you can ask it “what if” questions. What if I change this ACL? What if this link goes down? The system can tell you exactly what will happen because it knows the physics of your network.
Adding the conversational interface makes this diagnostic power available to everyone, not just the two people in the company who know the query language. You can ask, “Show me all the paths from the DMZ to the production database that bypass the web application firewall.” In the old world, that’s a multi-hour project, probably involving several teams. With Forward AI, it’s a three-second conversation.
Agentic Operations and the End of Toil
Where this gets really interesting is when we move into “agentic” operations. This isn’t about letting an AI run wild on your core switches while you’re at lunch. It’s about human-supervised automation. Forward AI can help identify those manual bottlenecks, the stuff that takes up 80% of an engineer’s time but provides 0% of the creative value. By using the conversational interface, you’re basically delegating the “finding” part of the job to the machine.
“Find all the devices running this specific version of OS that are vulnerable to this CVE.”
“Which of these devices are actually exposed to the internet?”
The AI pulls that info from the digital twin instantly. You get a verified list of actions to take. No more grepping through config files or manually checking vulnerability databases against your inventory spreadsheet. It’s about making complex operations radically simpler.
The New Source of Truth
We’ve talked for years about having a single source of truth for the network. Usually, that results in a CMDB that no one trusts because it’s always three weeks behind reality. Forward Networks is making a strong case that the mathematical digital twin is the only source of truth that actually matters. If the math says a packet can’t get from A to B, it can’t. Adding Forward AI on top of that ground truth changes how we interact with our infrastructure. It moves us away from being CLI junkies and toward being architects who manage intent.
It’s not just about the AI, though. The AI is the wrapper, a natural human interface. The real value lies in the underlying, provable model. Without that, you’re just chatting with a bot that might be lying to you. With it, you’ve got a window into exactly how your network is behaving, right now, in plain English.
The Forward Networks presentation at AI Infrastructure Field Day is here on the Tech Field Day website, and all of their appearances at Tech Field Day are on the Forward Networks page.
