DOGE should use AI to fix environmental review

The recently conceived Department of Government Efficiency (DOGE), headed by Elon Musk, is the big, new Trump administration idea on the block for cutting costs and making government work better. It should tackle a problem of government inefficiency that is holding up investment and job creation associated with development projects of many kinds, including siting clean energy and connecting it to a grid.

DOGE should focus its tech talent on making the National Environmental Policy Act (NEPA) work the way it was intended: to make federal decision-making sensitive to environmental impacts but not create the byzantine paperwork exercise that haunts many projects. To do that, DOGE should leverage artificial intelligence (AI) technologies to streamline bureaucratic processes.

Sign up for PowerPlay, the ’s bimonthly newsletter keeping you up to date on all facets of the energy transition.

NEPA doesn’t need to be so cumbersome

On January 1, 1970, then-President Richard Nixon signed NEPA, and it quickly became a cornerstone for environmental protection in the United States. NEPA doesn’t establish limits for harm—it is a “process” statute requiring federal agencies to identify planned actions that may significantly affect the environment and to describe those impacts in detail, for both the project as proposed and for a range of alternatives. Federal agencies must then state which action they will take, and which measures they’ll implement to mitigate the impacts.

But NEPA has long been a cumbersome process. The law and its amendments call for brevity in words and time, but the collective parts of an environmental impact statement (EIS) can run hundreds or even thousands of pages long and take more than two years to prepare—often by outside firms. Neither the environment nor the participants in the process benefit from that excess—decision-makers rarely even read the EIS.

It’s time for a dramatic change in the way that federal environmental review is carried out. The emergence of AI creates a tool to make that change a reality.

AI can streamline government processes

The Bureau of Ocean Energy Management (BOEM), where I have served, launched an effort in this direction in February 2020 during the last year of the first Trump administration.

BOEM’s initial idea was simple: EISs and other environmental documents were being created anew by the agency for each proposed action. Some parts of those documents were unique to the action involved, but much of the information, such as a required description of the affected environment, was largely identical for activities in the same geographical area. BOEM realized that an information base kept updated by agency scientists would save staff from unnecessary, repetitive review and speed things up.

BOEM named its initiative Status of the Outer Continental Shelf (SOCS) reflecting the agency’s jurisdiction. It began by compiling environmental documents prepared and vetted by the agency over the years and initiating a study to develop a model for decision-making using that information base. The model would not take humans out of decisions but instead provide them with objective indices of impacts on the environment based on defined categories of concern, such as the presence of endangered species and importance to tribal culture.

SOCS is underway now in BOEM, and its potential is made dramatically more significant with the emergence of generative AI.

Here is the concept: couple the SOCS information and model with generative AI and then fine-tune a custom AI tool for BOEM that can prepare EISs and other environmental documents. On top of that, use AI to facilitate public engagement faster and better than is currently done by providing a way for anyone to ask questions directly to the AI tool about projects and NEPA documents.

This concept can work for any federal agency making decisions with environmental impacts, not just BOEM.

How AI can fix NEPA

That said, one approach for developing a new AI-based tool could follow these steps:

• Upload contextual documents, including NEPA, other environmental laws and regulations, plus guidance documents and judicial decisions—the more, the better. Include EISs that are exemplary documents so the AI tool can learn what an EIS should look like—that is, it should communicate key issues concisely, clearly, with supporting graphics, focus candidly on important issues, and specify clear and enforceable mitigation measures (as conditions of approval).
• Have the AI produce an EIS template drawing from these uploads and integrating a decision-making model if an effective one becomes available—something DOGE should include in its NEPA-related efforts. A good model should transparently address the full range of impacts of greatest concern. It also needs to be user friendly for agency staff who are not modelers themselves.
• Task the AI tool to prompt the human team with requests for information specific to the EIS-proposed action.
• Fine-tune the AI tool through iterative refinement. This would include human experts systematically reviewing, correcting, and updating AI-generated output, since generative AI models can “hallucinate” facts that require fixing. The review should also look hard for and correct model bias—such as the Google Gemini AI model which, when asked for images of the Founding Fathers, only came up with people of color.
• Have the human experts closely review completed draft EISs for accuracy and quality. This task should become easier over time as reviewers gain experience.
• AI tools can also enormously improve public engagement with EISs. Google’s NotebookLM is one option currently available for free. Users can upload an EIS (or any other document) and ask questions about it. The answers are reliable and the tool can even generate an engaging podcast.
• Eventually, it may become possible simply to task an AI agent to produce a draft EIS, making sure it can access information specific to the project concerned.

NEPA is fixable

So, why aren’t EISs being prepared this way now? It’s partly because generative AI is still novel and government is slow to change. NEPA itself is not an obstacle. The statute and its regulations provide flexibility for how an EIS should be drafted.

To be sure, agency lawyers will wring their hands about what courts may do with AI, but that’s not a good reason to hold back. With rescission of the Chevron doctrine by the Supreme Court, which eliminated deference to agencies by judges, predicting judicial outcomes is impossible, and NEPA can be amended if warranted.

Government information technology (IT) policies are perhaps an even greater inhibition for AI innovation than nervous lawyers. IT requirements, some of which are legislated, are necessary for system security. But the process of change allowed under them can be suffocating and lead agency program staff to avoid innovation.

These organizational inhibitions make improving environmental review under NEPA a strong candidate for prioritization for the Department of Government Efficiency envisioned under the second Trump administration.

DOGE, which aims to bring in technology-focused staff from outside of government, working with the White House Council on Environmental Quality on the inside, could deliver a needed shake-up. It could bring the NEPA process into the 21st century. That would mean a more efficient path to renewable energy growth and the quest for net-zero carbon emissions, while creating a better understanding of the adverse environmental impacts of projects.

Go for this one, DOGE; it’s waiting for you in plain sight.

William Yancey Brown is a nonresident senior fellow at the Global Energy Center. From 2013–2024, Brown was the chief environmental officer of the Bureau of Ocean Energy Management in the US Department of the Interior, where he oversaw the implementation of NEPA.

The Global Energy Center develops and promotes pragmatic and nonpartisan policy solutions designed to advance global energy security, enhance economic opportunity, and accelerate pathways to net-zero emissions.

Image: Power lines near Baylands Nature Preserve in California. (King of Hearts, Wikimedia Commons) https://commons.wikimedia.org/wiki/File:Palo_Alto_Baylands_January_2013_001.jpg