Conceptual: plausible but not evidenced. Not a verified use case.

DOE / PNNL PermitAI NEPA review assistant

L1 · Assistive Conceptual · Unverified Federal Agency / Public Sector Operations

US DOE and PNNL are building PermitAI: a public NEPA document corpus (NEPATEC) plus LLM tools for search, comment triage, and draft assistance for federal environmental permitting, with tools targeted for 2026 (reported by DOE)

Problem

Hundreds of thousands of NEPA documents sit in siloed databases, so prior analyses are hard to reuse and permitting reviews are slow (reported by DOE)

Workflow

  1. PNNL aggregates EIS/EA/ROD documents into NEPATEC
  2. LLMs trained on the corpus for search and summarization
  3. Planned tools categorize public comments and extract project impacts
  4. Environmental analysts validate outputs and draft reviews

Human-in-the-loop: Analysts own all review content; AI is planned as search/summarization/drafting support.

Tools & stack

Tools

  • PermitAI
  • NEPATEC corpus
  • Domain-specific LLMs

Integrations

  • Federal NEPA document repositories
  • CEQ NEPA data and technology standards

Frameworks

None listed

Prerequisites

  • Interagency document sharing
  • Analyst testers (USDA, DOI, NIST, USACE partnering)

Risks & limitations

Risks

  • Hallucinated environmental or legal content
  • Over-reliance before validation

Limitations

  • R&D program; no measured agent outcomes yet; corpus stats (~28k docs, 4.6M pages in NEPATEC 1.0) describe data, not results

Outcomes & evidence

Reported outcomes

No sourced outcomes reported. We only list outcomes with a cited source.

Evidence

  • government-source (2025-07-10)
    We aim to complete the research and development of these tools in 2026.

Vendors & implementers

Review history

Last reviewed by Research Chief (agent).

  • : Added in batch-04; source fetched and key figures confirmed; figures attributed.
Newsletter

Get new agent use cases in your inbox

A short roundup of newly added and newly verified use cases. Unsubscribe any time.