United States – PubMatic has announced the launch of AgenticOS, a new operating system designed to support agent-to-agent advertising execution across premium digital environments, as the company expands its focus on agentic AI within programmatic advertising.
The AI-powered ad tech firm said AgenticOS is intended to serve as a system-level layer that enables autonomous agents to plan, transact and optimise advertising campaigns across PubMatic’s platform. The launch is being carried out in collaboration with several advertisers, agencies and publishers, including WPP, Butler/Till and MiQ, which are participating in early testing and deployment of agent-led workflows.
According to PubMatic, the increasing complexity of digital advertising — driven by the proliferation of devices, formats, advertisers and audiences — has placed strain on traditional programmatic systems. AgenticOS is positioned as a response to this environment, with a focus on enabling continuous decision-making while maintaining operational control and reducing fragmentation.
Under the new system, advertisers are able to define campaign objectives, guardrails, brand-safety requirements and creative parameters through their preferred large language model interfaces. PubMatic’s platform then translates these inputs into coordinated actions carried out by multiple intelligent agents, which are responsible for planning, execution and ongoing optimisation within predefined limits.
AgenticOS brings together PubMatic’s previous investments in agentic AI, including work on agent-to-agent communication, yield optimisation, audience and inventory discovery, planning and troubleshooting. These capabilities are now integrated into a single operating environment designed to manage agent-led campaigns end to end.
The operating system runs on NVIDIA-accelerated computing and is embedded within PubMatic’s global infrastructure. The company said this enables low-latency inferencing and coordination across millions of advertising transactions per second, supporting campaign execution in live market conditions.
PubMatic also highlighted an early agentic campaign conducted in December 2025 with independent agency Butler/Till and Geloso Beverage Group’s Clubtails brand. In that campaign, AgenticOS used natural-language input via Claude to recommend tactics, execute media buying and optimise performance in real time, within parameters set by the agency.
AgenticOS is built on what PubMatic calls its Architecture of Advertising Intelligence, a three-layer framework comprising infrastructure, application and transaction layers. The infrastructure layer uses NVIDIA-accelerated computing to enable real-time data integration and fast decision-making across tens of millions of auctions per second. The application layer embeds agentic capabilities across PubMatic’s platform, allowing agents to automate functions such as planning, forecasting, pacing, yield management, troubleshooting and measurement. The transaction layer connects agentic decisioning directly to PubMatic’s Activate buying platform, supporting execution across Programmatic Guaranteed and Private Marketplace transactions.
PubMatic said early tests of agentic campaigns have shown reductions in campaign setup time and issue resolution, as well as improvements in operational efficiency.
To support broader adoption, the company is also launching an Agentic AI Acceleration Program aimed at helping advertisers, agencies, publishers and partners move from testing to live deployments within weeks. The program is expected to support in-market implementations through the first quarter of 2026, with a focus on execution, governance and measurable outcomes.
