In the past year, AI agents have moved from research prototypes into production EDA environments. At the forefront of this transition is ChipAgents, an agentic AI system purpose-built for RTL design and verification. Since its introduction, ChipAgents has been deployed across leading semiconductor companies, engaging with design engineers, verification teams, and toolchain owners on real projects. This workshop shares what we have learned from those deployments: the successes, surprises, and lessons that can guide the broader adoption of agentic AI in EDA.

We will present case studies where ChipAgents accelerated UVM testbench development, improved coverage closure, and reduced debug turnaround time, while also discussing scenarios where agents encountered limitations and required human guidance. Particular emphasis will be placed on integration patterns: how ChipAgents accelerate a variety of workflows, how teams structured evaluations, and which workflows benefited most from agentic AI.

Just as importantly, we will cover the challenges discovered in practice: from handling complex corner cases, to measuring ROI, to work with production environments. Audience members will gain an honest perspective on both the potential and the practical hurdles of introducing agentic AI into high-stakes semiconductor design and verification flows.

By distilling insights from one year of real-world usage, this session aims to equip engineers, managers, and researchers with actionable knowledge on how to evaluate, adopt, and scale AI agents responsibly in their own verification environments.