Over the last decade, while working with Fortune 500 enterprises and some of the most secure government networks on the planet, I’ve consistently heard the same question: “Can your software tell me what will happen if I push this change?”
Every change goes through roughly the same high-level process, whether it’s a vulnerability patch, a new service to launch, or a new office coming online. Someone designs it based on their level of experience and skill with the network. There may be some limited testing, usually in a lab that bears only a passing resemblance to production, followed by a change review board that tries to identify any errors. And finally, the change goes live, and production becomes the test environment. This can take weeks, or even months. No matter the rigor of the process, changes are still the riskiest element of network operations and most common cause of outages and security exposures.
A telling symptom of how regularly changes cause problems, are the “war room” calls that are regularly convened during the change window. These calls require stakeholders from all critical applications to commit hours of availability to determine what (if anything) broke during the window. Known errors are problematic, but the most insidious and dangerous errors are silent and undetected. These issues cause gaps in security, decreases in redundancy, and have no identified impact until a future date when something that should be innocuous causes them to wreak havoc on the network.
Networks are the backbone of business and the government. Every other engineering discipline, (semiconductor design, aerospace, civil engineering, software development), uses mathematical models and exhaustive testing to prove outcomes before anything touches production. Networking has never had that. Not because teams weren’t rigorous enough, but because the technology to do it simply didn’t exist.
That changes today. Announcing Forward Predict.
After each spending over five years at Stanford building cutting-edge Software Defined Networks, my co-founders and I were struck by how limited the software tools that networking professionals depended on to run global enterprises were. The complexity and criticality of these networks had been scaling up exponentially since the dot-com boom, yet investment and innovation in operator tooling had not kept pace.
In 2013, after graduating, Brandon, Nikhil, Peyman, and I started Forward Networks. We set out to apply the best techniques and research available from computer science to modernize how networks are secured and run, not just to bring these tools to parity with the rest of IT, but to lay the foundation for a fully autonomous networking future.
The first step for us was to build Forward Enterprise, the industry’s first mathematically accurate network digital twin. It made the current and historic network knowable: every device, every vendor, every layer, searchable and provable. Most of the pain of diagnosis, drift, and compliance went away. Forward AI added a layer on top of that, agentic operations that let any engineer analyze a complex multi-vendor network through natural conversation.
But all of that work was still focused on how the network is (or was) behaving. The harder question, how will it behave after the next change, was still unsolved.
Forward Predict delivers the world’s first multi-vendor prediction of the full impact of a network change atop the mathematical foundation and accuracy of a true network digital twin. Forward Predict models the full control and data plane of network devices, including distributed protocols such as BGP and OSPF, alongside their interactions and convergence. After modeling the future state of the network based on the proposed change, our entire suite of algorithms and analysis from Forward Enterprise are available for use. The model spans every vendor, every layer from L2 to L7, on-prem and cloud. It is the exact production network mathematically reproduced, not a sample, not a representative subset, not a vendor-specific view, everything in one model.
Forward Predict should be part of every network change. It dramatically reduces the time from design to delivery, while eliminating risks that have always made changes the most feared part of running a network. Here is how it fits into your existing workflow.
Intent. The first step of any change is defining what success looks like. Whether that is creating new connectivity between endpoints, blocking existing connectivity, or improving compliance as part of an audit finding. Forward Predict provides extensive, easy to use capabilities to define that intent, which is then used to grade the change. Think of it like test-driven development, but for your network.
Design. For non-templated changes, the user designing the change can use Forward Predict as a safe space to quickly test potential changes and see exactly what the effect will be across the entire production network, without ever touching it.
Intent testing. After iterating toward a potential change, Forward Predict can execute all user-defined intent tests to verify if the network will behave as intended. The answer is a deterministic pass or a fail, not an opinion.
Regression testing. In addition to intent tests, Forward Predict runs the proposed change against your full suite of regression tests, the same way a software CI pipeline does. Unanticipated side effects such as paths you did not mean to open or policies you did not mean to break, are caught before the change ever reaches production.
Approval. Because Forward Predict has tested the change across the entire production network, and provided evidence of every security, connectivity, and compliance impact, the people in charge of approval can make data driven decisions faster. They no longer have to hope for the best, they’ve got mathematically proven outcomes.
Deployment. The change window becomes a formality. Forward Predict’s exhaustive pre-deployment testing has already done the work, war rooms don’t exist, and everyone stays in their beds.
Advanced API driven workflows can easily integrate with any or all of the above stages, and deliver benefits immediately. And when you bring AI into the picture, the pace of what becomes possible changes entirely.
Forward Predict enables both manual and fully automated networking change workflows, dramatically reducing the time required for design, approval and deployment, while increasing confidence, ensuring security, and verifying compliance.
Great AI requires great software and tools to support it. In software development, when AI writes code, it can't know whether the syntax is correct until a compiler checks it, and it can't know whether the code actually works until it runs and passes unit and regression tests. If problems surface, the AI can iterate at machine speed fixing, retesting, and repeating until the code works.
Forward has built the world’s first AI networking agent for multivendor network change, creating the networking equivalents of the compiler and testing framework. When you ask Forward AI to design a change, it draws on Forward’s network digital twin to understand the current state of your network in full detail. It proposes a change and uses Forward Predict, the compiler equivalent, to validate syntax and build a “runnable” model of the updated network. Next the agent runs Forward’s exhaustive test suites for connectivity, security, and configuration compliance against the future network. Any issues that surface are resolved by the AI at machine speed, iterating until it delivers a tested, working result.
This is a major paradigm shift in network change management, and it isn’t limited to one change at a time. You can feed your entire backlog of changes into Forward AI and have every change designed in parallel. This has the potential to accelerate network operations the same way AI coding agents have accelerated software development.
I never liked “move fast and break things.” Forward Predict enables your engineers to move fast and break nothing. Changes that used to take months of planning and review can be designed and proven in minutes.
We have been building toward this for over a decade. The network digital twin made it possible to understand with precision the behavior of your present, and historical network. And now Forward Predict makes it possible to deterministically know how it will behave next.
We will be demonstrating Forward Predict at Cisco Live. Stop by our booth to see the full running against real configurations, from change proposal, verification, and test execution, through to, post-delivery validation. Or book a meeting with me and our expert team at the show.
If you can’t make it to Cisco Live, learn more at forwardnetworks.com/predict or request a technical session.
The era of running the network on hope is finally over. Knowing the future of your network is the future of networking.
