News

Jan 30, 2026

The new Forward AI capability builds on the vendor's digital twin and is designed to allow network teams to ask complex questions, understand network behavior, validate outcomes and safely automate workflows.

Credit: PopTika/Shutterstock

Network operations teams face a flood of AI-powered tools promising simplified management through natural language interfaces. Many amount to conversational wrappers around existing capabilities, offering convenience but not fundamental changes to how networks are managed and verified. That’s what Forward Networks’ first AI tool, called AI Assist, provided when it launched in 2024.

Now in 2026, Forward Networks is taking a different approach with an agentic AI system that goes beyond a conversational interface. Rather than simply answering questions, it dynamically plans and executes multi-step workflows across hybrid and multi-cloud environments while maintaining mathematical verification of its recommendations.

Forward Networks was founded in 2013 by four Stanford PhDs, and the company spent more than a decade building mathematically accurate network digital twin technology before introducing AI capabilities. The new agentic capabilities in Forward AI are firmly grounded in the founding vision of the company, which is all about delivering an accurate representation of a network.

“When we founded Forward Networks, we started with a simple but strategic question: Does the network actually behave as intended, by design, in production, and across changes,” David Erickson, CEO and co-founder of Forward Networks, told Network World. “Answering that question requires more than visibility. It requires a mathematically accurate model of the network itself.”

Agentic AI vs. AI assist

Forward AI differs fundamentally from the company’s 2024 AI Assist feature, which translates English questions into queries against a normalized data model. The new system is an agent that dynamically constructs multi-step execution plans that reach both the internal digital twin and external systems like ServiceNow.

“It’s a conversational agentic system that is designed to simplify operations overall,” Nikhil Handigol, co-founder and chief AI officer at Forward Networks, told Network World.

The practical difference becomes clear in troubleshooting workflows. When asked to triage a ServiceNow ticket, the agent reads the ticket content, gathers context about entities mentioned from the digital twin, automatically performs path traces for connectivity issues, and returns a diagnosis. The complete workflow remains visible to operators throughout the process.

This differs from simple natural language queries where the system translates a question and returns an answer. The agent is building and executing a plan that may involve multiple data sources and analysis steps.

Custom framework for context control

Forward Networks built its own agentic framework rather than adopting existing tools like LangChain or CrewAI. The decision centered on maintaining precise control over context engineering, which Handigol described as the core engineering challenge for agentic AI.

“We built our own because we wanted complete control over how the agent executes,” Handigol said. “The engineering problem mostly comes down to context engineering. How do you define and maintain the context that is necessary for the agent, that you send to the LLM, to get the right answers?”

Network World Smart Answers

Learn more

Explore related questions

• How are agentic AI solutions architecturally structured?
• How will agentic AI impact enterprise decision-making by 2028?
• What are key challenges in scaling agentic AI?
• What are the security implications of agentic AI systems?
• What are Forward Networks' 2024 AI Assist feature capabilities?

Ask

The team defines context engineering as providing all relevant information without excess noise. Too little information produces wrong answers. Too much information distracts the model from the correct task.

Context is drawn from Forward Network’s hierarchical data stack. At the base layer sits raw configuration, state, and statistics collected directly from devices. The next tier normalizes this raw data into a queryable model showing everything present in the network and its configuration.

The top tier performs behavioral analysis. This layer can answer questions about arbitrary packet paths through the network, asset exposure to external networks, and lateral movement blast radius if assets become compromised. The system also correlates network data with CMDB systems, IPAM, and ticketing systems.

Addressing the hallucination problem

Large language models are fundamentally probabilistic systems that predict the next token. This creates inherent hallucination risk that is dangerous in network operations where a single misconfiguration can cause widespread outages.

Forward AI takes a two-pronged approach to mitigate this risk. First, it bases recommendations on a deterministic digital twin rather than relying solely on probabilistic predictions. Second, it makes all reasoning transparent.

All evidence used to support recommendations is directly accessible during conversations. This allows operators to validate the agent’s reasoning before taking action.

The company also employs an evaluation framework during development. As the agent code evolves, the framework is designed to ensure no regression in existing capabilities while it validates improvements in new areas.

Multi-vendor foundation across hybrid environments

The digital twin underlying Forward AI supports dozens of network vendors and spans Layer 2 through Layer 7 protocols. This includes on-premises infrastructure, major cloud providers (AWS, Azure and Google Cloud), and Kubernetes environments.

“The major advantage that Forward Networks has is that we are multi-vendor,” Erickson said. “We support all network vendors that are out there, and we make the lives better for the customers of all of these vendors.”

This multi-vendor support eliminates a significant source of operational toil. Network operators no longer need to log into multiple vendor CLIs, learn vendor-specific command syntax, and manually normalize data from disparate sources.

The breadth of protocol and technology support is critical for the agentic system to function reliably. Without comprehensive data collection and analysis, recommendations would have gaps that could lead to errors.

“Everyone is building some sort of AI-powered capability for their own systems,” Handigol said. “What differentiates one AI-powered capability from the other is the foundation on top of which it is built.”

NetworkingArtificial IntelligenceNetwork Management Software

Don’t miss a thing

Join the Network World First Look mailing list for the latest news, analysis, and insights. Sign up now!

By submitting your information, you agree to our PRIVACY POLICY.

SUBSCRIBE

Related content

News

Cisco: Latest news and insights

By Network World staff

Jan 30, 202636 mins

Network Management SoftwareNetwork SecurityNetworkingNews

Top 11 network outages and application failures of 2025

By Denise Dubie

Jan 30, 20268 mins

InternetNetwork Management SoftwareNetwork MonitoringNews

Cisco adds intelligent policy enforcement to mesh firewall family

By Michael Cooney

Jan 29, 20264 mins

FirewallsNetwork SecurityNetworking Devices

Other Sections

• PODCASTS
• VIDEOS
• RESOURCES
• EVENTS
• Spotlight: IT Leadership

SUBSCRIBE TO OUR NEWSLETTER

From our editors straight to your inbox

Get started by entering your email address below.

Subscribe

by Sean Michael Kerner

Contributing Writer

1. Follow Sean Michael Kerner on X

Sean Michael Kerner is an IT consultant, technology enthusiast and tinkerer, and has been known to spend his spare time immersed in the study of the Klingon language and satellite pictures of Area 51. He has pulled Token Ring, configured NetWare and has been known to compile his own Linux kernel. He consults to industry and media organizations on technology issues.

Sean's writing has appeared in VentureBeat, InternetNews, TechTarget, ITPro Today, Data Center Knowledge, and TechCrunch, among other outlets.

© 2026 FoundryCo, Inc. All Rights Reserved.