It has been a week since we announced Forward Predict at our Innovation Day broadcast, and I'm still taking it in.
Since the inception of networking, the industry has been working without a safety net, making changes in the production network without knowing their impact beforehand. The result has been outages and security breaches. This wasn’t a lack of diligence, it was because there was no way to know, with certainty, what a change would do to the production network before it was pushed.
Certainty stayed out of reach. The production network was, ultimately, its own test environment. Everyone in this industry knew it. Even though everyone knew the risk there was no other way. We thought that was unacceptable, and we set out to resolve the most frequently asked yet never deterministically answered question in networking, “What will happen to my network when I push this change live?”
On May 20, we announced Forward Predict. And in the week since launch, the response has confirmed something I suspected but couldn't quite let myself believe until I saw it: the industry was ready. More than ready. Customers, partners, former colleagues, analysts, journalists, people I haven't talked to in years reached out. The reaction from every direction has been the same. This is the capability we have been waiting for.
That kind of alignment doesn't happen by accident. It happens when you're over the target. When a technology that has been long discussed but never quite delivered finally shows up in a form that actually works, people recognize it. Not because we told them it was important, but because they had been waiting for someone to actually deliver it.
If you missed the SDxCentral Innovation Day broadcast, "The self-driving network is no longer a thought experiment," I'd encourage you to watch it on demand. What struck me watching our CTO and co-founder Brandon Heller moderate that panel was the caliber of the people on the panel and where they were coming from. Nick McKeown, Professor Emeritus of Computer Science and Electrical Engineering at Stanford, representing decades of foundational academic work on networking. Josh Matheus, VP and Chief Portfolio Strategist at Cisco, who spent years on the operator side running one of the largest enterprise networks in the world before moving to the vendor that ships to most of the industry. Jon Green, CTO and CSO of HPE Networking, speaking for one of the major networking vendors that enterprises run on.
Foundational academic research, hands-on operations, and two of the largest networking vendors on the planet. Three people who could be expected to see the challenges of autonomous networking differently, but they didn't. They agreed on the diagnosis, they agreed on the stakes, and they agreed that a mathematically accurate digital twin is the way through. That was the moment it really landed for me: this is no longer a debate about whether deterministic, autonomous networking is possible. That debate is over. The question now is how fast the industry can adopt what's already here.
Nick framed the problem the way only Nick can: "When you build a network with hundreds of devices, the complexity mushrooms to a point where no human can truly understand it so you need automation. In industries like aerospace, they simulate the entire system with a mathematical model. Networks are now reaching that same point where you can understand the consequences of a failure or misconfiguration before making a change, with high confidence." That is exactly the standard Forward Predict delivers.
Josh, who has lived this problem at the scale of one of the largest enterprise networks in the world, put it bluntly: "What we've always wanted is the equivalent of software development for networking: the ability to test, validate and know a change will work before deployment. It's taken 25 or 30 years of networking to finally get here." And on what makes this moment different from the broader AI noise: "Anything based purely on probabilistic AI is hard to trust in production infrastructure. But when AI operates against a digital twin, it can perform experiments, validate outcomes, and prove a change will work before deployment. That's a completely different ball game and a genuine game-changer for networking."
Jon brought the operator's reality home: "In some industries, it's even illegal to have downtime so the change management cycle is enormous, because teams don't always have a deterministic understanding of what a change is going to do. Sometimes they really don't know the final impact until they push a change into production."
That is the gap that Forward Predict eliminates.
Every network engineer knows the feeling. The change window opens. The plan looks right. The CAB signed off. And still, there is a knot in your stomach, because nobody actually knows what is about to happen.
That anxiety is not irrational. It is the logical consequence of operating critical infrastructure without proof. We have all been there, staring at a terminal at 2 a.m. during a maintenance window, hoping the change we just pushed does not cascade into something ugly or worse catastrophic. That is not engineering. That is hope. And hope is not an operating model for production networks.
Forward Predict replaces that dread with certainty and confidence. When a team proposes a change, the platform evaluates the end-to-end outcome across the full network – reachability, routing behavior, policy compliance, segmentation integrity – and delivers a verified result. If the change would break something, the platform enables the team to iterate until the intended result is achieved.
This is not an incremental improvement. For the first time, network teams can operate with the same rigor that aerospace and defense engineers have demanded for decades: don't deploy until the model says it's safe.
As autonomous networking systems begin operating on networks, Forward Predict becomes the closed-loop verification layer. An AI agent proposes a change, validates it against the twin, receives specific failure feedback, iterates until the change is mathematically safe, and executes only then.
Josh said it on the broadcast better than I could: “the marriage of an AI workforce with a predictive digital twin is what enables AI at machine scale.” I’m proud that our team has delivered it. This wasn’t easy, it’s a journey we started a decade ago developing the mathematically accurate model of the entire production network. In the last two years the technology to support the computations necessary for predictive reasoning have become available and our engineering team has worked nonstop to make this industry requirement a reality for every network.
The thing I am most looking forward to right now is Cisco Live (Las Vegas the week of June 1 - 4, 2026). That is where the conversation moves from announcement to hands-on. After a week of reading reactions and answering questions, what I want most is to put Forward Predict in front of practitioners and let them see it.
If you are going to be there and want an in-depth session, request an executive meeting. Our best engineers will be on the floor to walk you through the platform and answer questions.
I want to be clear about something. Yes, everyone at Forward is proud of what we have built. This team has worked for years to reach this moment, and the response over the last week has validated that work in ways that matter.
But we are not treating this as an arrival. Forward Predict is not the finish line. It is the starting gun. That means listening to customers to ensure our roadmap matches your requirements, extending integrations, adding features, and proving every day that the certainty we deliver holds up in the most demanding production environments worldwide. Our customers are trusting us with their most critical infrastructure. That trust is not earned by a launch. It is earned by what comes after.The networking industry spent a generation debating whether this kind of pre-change verification was even possible. We ended that debate. As Nick put it on the broadcast: "We've moved into an era where immensely complex systems can now have provable mathematical properties giving you the ability to test against them with a high degree of confidence that they will work." Now the question is how fast we can deliver on the full potential of that capability — and that is entirely on us.
See you at Cisco Live.
