Articul8, the generative AI Intel spinout, has unveiled a Network Topology Agent tool designed to provide network operators with increased visibility of their systems.

The agentic AI solution, known internally as Weave, launched alongside a model benchmarking tool called LLM-IQ. It was built from the ground up using data sourced from carriers and operators to create what CEO Arun Subramaniyan described as a living, queryable graph of an entire network.

“The holy grail in networking is autonomous optimization,” Subramaniyan said. “But you can’t optimize what you can’t see.”

The Weave agent ingests massive volumes of structured and unstructured network data to provide a holistic view, with Articul8 revealing exclusively to SDxCentral that carriers in the US, Europe, and Asia tested it to handle a variety of topologies, from power-heavy grids to high-throughput backbones.

“We’re looking at roughly a billion records every 10 days, and networks with tens of millions of nodes,” he said. “This isn’t a pretty demo. It’s a living system.”

Articul8 was spun out of Intel last January, partnering with several investment firms, including private equity giants DigitalBridge, to support its goal of developing domain-specific AI models and systems.

Its network-focused AI agent can help network operators prepare for unexpected downtimes, understand traffic patterns, and identify unauthorized connections and potential attack paths.

Compared to traditional network virtualization and emulation offerings, Articul8 claims its agent-based system provides not just observability, but topology intelligence as a service.

“This agent delivers programmatic, composable access to the network's truth – data you can automate, store, compare, and secure,” a company blog post on the product launch reads. “Teams can now approach the network as a structured graph, building consistent workflows for visibility, change detection, scenario planning, and policy enforcement.”

The startup’s latest offering aims to tackle real enterprise problems, in a reverse of the hype cycle which has seen the concept of agentic AI bandied about and shoved into applications, in what a recent Gartner report dubbed “agent washing.”

“Everybody builds an agent and then figures out where it will plug in. For us, the agent is the last step,” Subramaniyan told SDxCentral. “We look to solve actual use cases for a customer, and from there, distilled the smallest possible agent that we think is actually useful for a variety of different players.

“This is not something that looks for a problem to solve. The problem has already been solved, and we are basically distilling the core unit and then presenting that as an agent for other people to use.”

‘ChatGPT moment for the enterprise’

The Network Topology Agent tool is being made available on AWS’s new marketplace for AI agents, a platform where customers can tap into a variety of different agentic AI solutions.

Subramaniyan said that getting Articul8’s network agent tool listed on the AWS platform wasn’t just about expanding reach, but trust: “To be accepted, our agents had to pass brutal security and performance testing by multiple AWS teams,” he said.

In addition, agents run inside the customer’s secure environment, so there’s no need to route data externally.

“For enterprises, it’s just a line item on their AWS bill. No legal renegotiation. No supply chain disruption. Total auditability,” he said. That includes support for sovereign cloud zones, where data locality and compliance are non-negotiable.

The name of it was also deliberate, with Subramaniyan suggesting the offering can be broadly applied: “We're calling it network topology. We didn't call it telco network topology. So anything that's a network would be able to take advantage of it.

“There's a lot of talk and hype about generative AI in general, but from the time a model becomes useful to an enterprise, it's a long road. Agents should be nearly instantly useful, especially from a marketplace and that’s really our hope.”

Referencing the wave of interest in AI following ChatGPT’s unveiling back in late 2022, Subramaniyan claimed the launch of the Articul8 agentic offering was the “ChatGPT moment for the enterprise.”

“What ChatGPT did for consumers, unlocking the potential of everybody, whether you're a techie or a non-techie, that's really what we feel is going to happen in the enterprise world with these agents.”

Articul8’s other new offering, LLM-IQ, is a no-code agent that evaluates language models across 25+ enterprise domains.

Subramaniyan told SDxCentral that it’s not solely a benchmarking tool, but a routing engine for real-time orchestration, capable of evaluating even closed-model systems like GPT-4 or Claude 4, and then plugs the right model for the user’s needs directly into live inference pipelines.