The U.S. Patent and Trademark Office (USPTO) is using AI agents to help patent examiners handle routine work and reduce the time needed to review applications, according to a NextGov report. 

Acting Commissioner for Patents Barry Schindler discussed the effort on October 20 at the Google Public Sector Summit in Washington, D.C.

USPTO faces a backlog of about 700,000 patent applications and employs roughly 9,000 examiners. Schindler said AI could improve consistency across examiners with different levels of experience while helping reduce review times from as long as five years to about one year, according to NextGov.

The agency has turned in part to Compass, an agentic tool that performs formality checks, such as verifying submitted DNA sequences before examiners assess whether they can be patented. 

By automating parts of the review workflow, USPTO aims to free skilled examiners for direct conversations with applicants rather than extended written correspondence.

AI Agents Enter Established Workflows

The USPTO example is significant because the use case goes beyond general-purpose AI assistants. Agents would operate within an established professional workflow where decisions can have legal and commercial consequences.

That raises questions about how agencies should control agent access to sensitive information, determine which tasks can be automated and establish points where human examiners must review or approve an AI-generated result.

Schindler said the goal is for agents to assist rather than take over. He also cautioned that excessive reliance on AI could weaken examiners’ understanding of the fundamentals, making training and human judgment ongoing requirements.

Organizations Must Adapt Data Management Practices As AI Scales

For Chief Data Officers (CDOs), this illustrates why data governance becomes more important as AI moves deeper into operational work. 

Patent examination depends on large volumes of technical, legal and historical information, making data quality, access controls and traceability important considerations.

It is imperative that organizations adapt data-management practices as AI scales across operational workflows.

The USPTO’s approach reflects a broader federal push to use agentic AI for specific operational tasks. The challenge now is determining where agents can safely take on work while preserving the expertise and accountability required for government decision-making.