- Keysight is bringing agentic AI to its software, enabling engineers to direct AI agents to automate complex design tasks
- RF engineering has been slow to adopt agentic AI because it relies on specialist expertise and schematics and layouts that LLMs cannot read
- Separately, Keysight launched its Keysight Velocity 4900 Series, which is Layer 1 automation switches designed to transform and automate physical connectivity in network test labs
Keysight Technologies is bringing agentic AI to its software, enabling engineers to direct AI agents to automate complex design tasks. Customers can use their large language models (LLMs) to generate and optimize designs, with Keysight simulation validating the agents’ progress. This shifts how engineering is done, as teams can evaluate significantly more options in the same development time.
More than 60% of organizations expect to deploy AI agents by 2028, according to 2026 research from Gartner (see chart). However, radio frequency (RF) engineering has been slow to the agentic AI plate because it relies on specialist expertise and schematics and layouts that LLMs cannot read in a consistent, deterministic way, according to Keysight.
As a result, the process has remained manual. Seeing a need, Keysight developed its Advanced Design System (ADS) 2027, which allows RF engineers to record workflows as macros so that their agents can learn from and convert graphical designs into code they can read.
Where does Model Context Protocol fit in?
Model Context Protocol (MCP) servers connect agents to the software, guiding LLMs and agents in their interaction with ADS, noted Keysight in an announcement about ADS.
“Foundation models are removing the usability barriers that have long slowed engineering in these fields,” Daren McClearnon, senior product manager at Keysight EDA, told Fierce Network. “Engineers ask for what they need, and agents, through their skills and context, read the documentation for them by working through product MCP servers and APIs.”
When an engineer makes a request in natural language, their agent completes the task in ADS, noted the announcement. MCP servers provide the agent with documented skills and tools built by Keysight that execute specific RF work consistently, reducing variability caused by AI inference models generating answers statistically. The agent then runs Keysight simulation to validate its output, according to the company.
ADS also creates an open ecosystem for agentic workflows because MCP servers work with AI assistants and LLMs that organizations are already using, and with tools from multiple vendors in the same workflow.
In addition, Keysight said ADS speeds and expands design cycles because agents can handle repetitive setup and simulation, so engineers can cover more scenarios and find more optimal designs while reducing time to market.
“Much of a company's intellectual property sits in human experience, organizational processes, proprietary methods, and years of product history and data, and restrictions from foundries and partners further limit what AI can be trained on,” McClearnon noted. “We help customers capture that knowledge and make it usable within their design flows, so AI amplifies what makes each company distinct in its markets.”
Plus, macros recorded in ADS can capture an experienced engineer’s methods for colleagues and agents to reuse across teams.
“We are investing in speed to result in several ways, while making sure engineers can still trust what they get back,” he said.
Keysight rolls out new automation switches, too
Separately, Keysight launched its Keysight Velocity 4900 Series, which is Layer 1 automation switches designed to transform and automate physical connectivity in network test labs. These switches allow engineering teams to connect and automate the management of test equipment via a GUI so that they can build out test workflows for complex networks and infrastructure for AI, 6G and next-gen service.
“Teams supporting AI infrastructure and future 6G development need to test more combinations of devices, interfaces, and network topologies. They simply cannot keep pace if setting up and configuring those environments depends on manual work,” said Stephen Brown, director of product management, Automated Test & Assurance, for Keysight.
The new Velocity 4900 series addresses the “challenge of the exponential growth in the number of tests required and the increasing complexity of recreating these environments in the lab,” said Brown.
“The Velocity 4900 series provides a dynamically configurable foundation for integrating physical test environments into automated CI/CD workflows. By making physical connections programmable, teams can automatically configure and recreate test environments, expand test coverage, and reduce setup time and connection errors,” he said. “Through the broader Velocity portfolio, teams can access and share test equipment globally across lab locations and fully automate test environments and processes, from setup and test execution to reporting and insights from the results.”
Read more about agentic AI on Fierce Network
Vecima: AI traffic, reliability concerns are driving a new wave of cable network upgrades