PNNL and OpenAI Sign Partnership to Accelerate Federal Permitting Procedures

The federal government's core infrastructure permitting process has become a bottleneck slowing innovation across various sectors including energy, manufacturing, and transportation. Environmental and technical reviews take years, increasing costs and delaying projects.

In response, OpenAI collaborated with the PermitAI™ team at PNNL (Pacific Northwest National Laboratory), under the U.S. Department of Energy, to design DraftNEPABench, a benchmark for evaluating how effectively coding agents can accelerate federal permitting work.

In tests involving 19 experts, general-purpose coding agents such as Codex CLI showed the potential to reduce NEPA (National Environmental Policy Act) document drafting time by 1 to 5 hours per section, or up to 15%.

These agents perform the following key tasks:

• Accurately summarizing and synthesizing technical and regulatory documents spanning hundreds of pages
• Verifying factual relationships across diverse environmental, engineering, and regulatory sources
• Drafting structured reports that meet highly specific legal and technical standards

This collaboration aims to go beyond simple document drafting, moving away from static PDF formats to generate web-based reports and interactive visualizations, making the expert review process more efficient.