Brazil INPI cuts patent examination times 50% and clears 80% of backlog with AI search

Brazil's national patent office faced a multi-year examination backlog, with pending applications totaling 15,134. Examiner search workloads were high, and many applications required substantial additional prior art searching beyond initial results.

INPI partnered with CAS to deploy AI-assisted prior art search in production patent examination. The system supported examiners by reducing the search burden and, in a substantial share of cases, delivering results that required little or no additional manual search.

Examination times were reduced by 50%. The multi-year backlog was cut by 80%, from 15,134 to 1,052 pending applications over 2.5 years. Search times were reduced for 77% of applications, and a meaningful share required minimal additional search—indicating the AI search was sufficient on its own in many cases.

• Production-scale AI prior art search can materially cut examiner workload and clear large backlogs.
• A meaningful share of applications needing little or no extra search shows AI can be sufficient as a primary search aid in many cases.
• National patent offices are a proven setting for AI search ROI via backlog and cycle-time reduction.