PubMatic launches governance framework for ad agents

PubMatic has launched a governance framework for autonomous advertising on its AgenticOS platform, with Rise, a Quad agency, among the first agency partners to pilot the system.

The framework lets advertisers and agencies set rules for how AI agents run campaigns while creating an audit trail of decisions from planning through measurement.

The architecture is built into campaign creation and activation, rather than added later as an oversight layer. An authorised account administrator can set governance parameters through natural-language prompts, defining rules, thresholds and approval structures that apply across campaigns.

For buyers already using the Ad Context Protocol, the framework serves as a separate system check on transactions handled through PubMatic, regardless of which AI buying agent initiates them. The controls are designed to sit alongside existing workflows rather than depend on a single tool or agent.

How it works

The system follows five steps. PubMatic sets platform-wide constraints for all agents, while buyers can add their own policies covering budget thresholds, creative approvals, inventory allowlists and audience restrictions.

Agents start with a preset state based on a buyer's business rules for targeting, pacing and creative selection. They are also limited to pre-approved libraries of inventory, creative and audience segments that have already gone through the buyer's review process.

If an agent reaches a decision outside its delegated authority, it must escalate the matter for human approval. Approvers receive real-time notifications and verify their decision through a one-time passcode, creating a record of each intervention.

The final layer focuses on audit trails and drift detection. Every agent action is logged with details of what happened, when it happened and whether a human approved it, while the system monitors for behaviour outside expected parameters and flags it for review.

The launch addresses a central issue in the advertising market as companies test more autonomous media-buying tools. Agencies and brands have shown growing interest in using AI to automate campaign execution, but governance, accountability and visibility into decision-making remain major barriers to wider adoption.

According to PubMatic, Rise has already run its first agentic campaign on AgenticOS. The agency's involvement offers an early indication of how ad buyers want safeguards to work as AI systems move from recommendation to execution.

"Transparency and accountability have always been non-negotiable for us. They're core to how we work and how we earn our clients' trust," said Klaudia Smykowska, Group Director, Media Investment at Rise, a Quad agency.

"As we bring agentic advertising into our practice, those standards don't change. The governance architecture PubMatic has built reflects exactly the kind of structured, auditable approach we need to deploy autonomous buying with confidence," Smykowska said.

Broader push

The new framework extends PubMatic's broader effort to build more visibility tools into AgenticOS. The platform already includes a Fee Transparency Agent and a Detailed Reasoning Agent, both designed to show buyers where money is spent in the supply chain and why the system made particular decisions.

AgenticOS debuted earlier this year and has supported more than 80 autonomous campaigns globally across more than 100,000 websites, apps and streaming services, according to PubMatic. More than 4,000 direct deals have also been executed on the platform.

PubMatic also cited internal and partner data showing that advertisers using AgenticOS have seen up to 50% better total fee efficiency, up to 87% faster campaign setup times and up to 70% faster issue resolution. The figures were presented alongside claims of improved performance across more qualified audiences and premium inventory.

The platform draws on bidstream signals from the sell side, first-party auction data and signals spanning identity, commerce, behavioural and contextual data from more than 300 partners. PubMatic said each campaign adds new signals back into the system.

Nishant Khatri, EVP, Product Management at PubMatic, said the framework was shaped by discussions with buyers about what they would need before expanding their use of autonomous advertising.

"We did deep discovery work with a broad range of buyers to understand what governance needed to look like before they could scale agentic campaigns with confidence," said Nishant Khatri, EVP, Product Management at PubMatic.

"What we've heard consistently is that buyers know what they want this technology to deliver, what controls they need to see, and where the system has to hold. This five-step architecture is the answer to what they told us. We believe that if we get this foundation right, there's no ceiling on what this technology can do," Khatri said.