Energy research is entering a new era.

Policymakers face increasingly complex questions amid a growing volume of information and constantly changing data. Production, prices, demand, technology, infrastructure, and policy are all moving targets. At the same time, the issues energy policy addresses—security, affordability, reliability, and environmental performance—are deeply interconnected and affect every aspect of modern life.

Making affordable, reliable, and cleaner energy widely available therefore requires grappling with questions that are both enormously complex and essential to human flourishing.

Artificial intelligence gives researchers new ways to work in that environment. At the National Center for Energy Analytics, we are using AI to accelerate research, improve access to evidence, and build analytical tools that can be updated as the world changes.

Our purpose is not to replace human judgment with machines. It is to equip researchers, policymakers, and the public with better tools for understanding the energy system.

Our Purpose

NCEA’s work is grounded in a straightforward principle: energy is foundational to human flourishing. As Founder and Executive Director Mark P. Mills and Advisor Scott W. Tinker, PhD, wrote in NCEA’s foundational declaration, The Choices We Face: Energy for the 21st Century:

"Energy will not end poverty, but we cannot end poverty without energy."

Understanding energy requires more than collecting facts. It requires connecting science, engineering, economics, markets, policy, and human behavior—and distinguishing what is measured from what is modeled, what is known from what is assumed, and what is possible from what is merely aspirational.

Our aim is to make that understanding more accessible and useful. For researchers, AI can accelerate the search for evidence, the analysis of data, and the development of new research. For policymakers, it can make timely information and clearly documented analysis easier to access. For the public, it can make complex energy issues easier to explore and understand.

AI is a means of pursuing that aim, not the aim itself. Its value depends on how it is designed and used. NCEA’s AI systems are intended to support source-grounded research, transparent methods, and informed human judgment—not to obscure uncertainty or automate conclusions.

Why We Created a Purpose-Built AI

General-purpose GPTs and large language models can summarize documents, write code, answer questions, and organize complex projects. With guidance, users can also shape a conversation around a particular topic or task. But a general-purpose model is not an energy research system.

It does not automatically distinguish among NCEA’s published research, original sources, and outside articles or commentary. Nor does it inherently understand NCEA’s research framework, standards, source hierarchy, or expectations for citation and verification. It does not approach energy through a structured analytical lens.

For serious energy policy work—and for anyone seeking to understand energy systems—those distinctions matter. A fluent answer is not necessarily a reliable one. A plausible citation may not support the claim attached to it. A number may be accurate for one year but outdated the next. And a model may not recognize when a question requires a primary source, a current data release, or human judgment about methodology.

NCEA built its AI agent to serve two audiences: the public and our research team.

For policymakers, journalists, scholars, and the public, it provides an accessible way to explore NCEA’s growing body of work, surface relevant publications and outside commentary, and understand connections across research on energy systems and policy. For NCEA’s researchers, it can examine that expanding corpus at scale—surfacing related claims, tracing citations to original evidence, comparing sources, and identifying information that might otherwise remain buried across dozens of publications.

The agent combines models from Anthropic and OpenAI with NCEA’s research library, source-aware search and retrieval, citation workflows, web access, and editorial safeguards. Its purpose is to make NCEA’s research easier to find, understand, verify, and extend.

The Public-Facing NCEA Energy Chat

The public-facing agent is integrated into NCEA’s website and available at ai.energyanalytics.org. It allows policymakers, journalists, scholars, and members of the public to explore NCEA’s published work through plain-language questions.

Users can:

• Find relevant NCEA studies, issue briefs, analyses, testimonies, and supporting data.
• Explore what NCEA has published on a particular energy issue.
• Identify related publications and compare their findings.
• Locate relevant passages, supporting evidence, and underlying sources.
• Receive accessible explanations of complex research.
• Follow answers back to the NCEA publications on which they are based.

The public agent makes NCEA’s research more discoverable without requiring users to know a publication’s precise title, terminology, or format. It is a guide to NCEA’s published research—not a substitute for reading the underlying work.

Research Mode

NCEA’s team uses a more advanced version of the agent in Research Mode, which supports the full research process, including source discovery, verification, analysis, and quality control.

Research Mode helps NCEA researchers:

• Define research questions and identify the evidence needed to answer them.
• Search NCEA’s research library and supporting documents.
• Trace claims in NCEA publications to studies, datasets, government releases, and other first-party sources.
• Use citations and footnotes as leads to original evidence.
• Find current external sources when a claim requires updated information.
• Compare definitions, methodologies, assumptions, and findings across sources.
• Extract and organize information from research documents and data files.
• Write, test, and revise analytical code.
• Identify missing citations, inconsistent definitions, unsupported precision, and gaps in an argument.
• Develop findings into public explanations, data products, and interactive tools.

Research Mode is not simply a larger chatbot. It is a source-first workflow that helps researchers move from a question to NCEA research, from NCEA research to original evidence, and from evidence to a documented, reviewable conclusion.

The public agent and Research Mode serve different purposes, but they share the same foundation: AI should make research easier to find, investigate, and understand while preserving source quality, transparency, and human judgment.

From Static Publications to Living Tools

AI is also changing the form research can take.

Traditional reports will remain essential. They provide context, explain methods, document assumptions, and preserve an analysis for a particular point in time. But reports can be complemented by living databases, dynamic indexes, interactive models, and tools that allow users to explore evidence themselves.

NCEA is developing this next generation of research infrastructure through projects including the Index of Leading Energy Indicators and NCEA-EMIT.

The Index of Leading Energy Indicators

The Index of Leading Energy Indicators is being designed to help policymakers and the public track how policy and market conditions affect the energy system through a structured set of indicators. It will also include a forward-looking projection and modeling component to help users examine how energy-system conditions could evolve under different assumptions.

The index will examine four dimensions:

• Security
• Affordability
• Reliability
• Environmental footprint, focusing on impacts to land, water, and local air quality

The long-term goal is an interactive dashboard offering a comprehensive view of the U.S. energy system and demonstrating the connections among these dimensions. Users will be able to examine historical and current conditions and explore how different policy or market assumptions could affect future outcomes.

The project is intended to be open source. Researchers will be able to download the underlying data and models, reproduce the analysis, and contribute to the project.

Rather than treating an index as a static chart published once and then left unchanged, NCEA is building it as an ongoing research process. Data can be updated as new releases become available, while the definitions, sources, methods, assumptions, and interpretation of each indicator remain documented.

The projection and modeling component will likewise make its inputs and assumptions visible, allowing users to distinguish observed results from forward-looking estimates.

AI-assisted workflows can support tasks such as:

• Monitoring designated source publications.
• Identifying new or revised data.
• Converting data into standardized formats.
• Checking for missing values and unusual changes.
• Supporting calculations, projections, and visualizations.
• Flagging updates that may require further review.

NCEA’s researchers and subject-matter experts remain responsible for defining the indicators, selecting appropriate models, evaluating source quality, validating the data, and designing visualizations that communicate the results clearly.

NCEA-EMIT

NCEA-EMIT is being developed as a data-driven modeling tool for exploring replacement scenarios in U.S. power generation.

It addresses questions such as: If a specified generation asset were retired, how much of another type of generation asset would need to be built to replace its energy output while achieving a specified reduction in CO₂ emissions?

To examine these scenarios, EMIT uses data on:

• Installed generation capacity
• Actual electricity generation
• CO₂ emissions over the most recent year of available data

Considering both capacity and actual generation allows the model to distinguish between the amount of electricity an asset could theoretically produce and the amount it actually produced. Users can then examine how different replacement choices affect the amount of new capacity required and the resulting emissions.

The analytical workflow begins with data downloaded from the U.S. Energy Information Administration through its API. NCEA’s researchers analyze the data and run the model in R, then export the completed tables to Supabase. The website can access those tables and present them through a public-facing interface.

The system is designed to connect: EIA API → data ingestion → analysis and modeling in R → tables in Supabase → public interface

EMIT is not intended to predict the future power system or determine a single preferred policy outcome. It is a structured tool for examining the implications of specified assumptions and replacement scenarios. Its usefulness depends on the quality of the underlying data, the transparency of the calculations, and the clarity with which its assumptions and limitations are presented.

Human researchers will determine which variables to include, how scenarios are defined, how the data are validated, and what caveats users need to see. The tool is designed to help policymakers and the public understand the trade-offs involved in changing the generation mix—not simply to display more numbers.

AI Requires Accountability

AI can accelerate research, but speed alone is not the standard.

NCEA’s AI-assisted workflows are built around several principles:

• Source grounding: Important claims should be connected to credible, identifiable evidence.
• Primary-source verification: Researchers should examine the original study, dataset, statistical release, filing, or official publication.
• Human accountability: Researchers—not the model—remain responsible for published claims, calculations, and conclusions.
• Transparent methods: Definitions, formulas, assumptions, and model changes should be documented.
• Data traceability: Users should be able to understand where data came from and when they were updated.
• Appropriate uncertainty: Forecasts and model outputs should not be presented with more confidence or precision than the evidence warrants.
• Editorial review: Humans write, edit, and fact-check research, content, and code even when assisted by AI.

These principles are consistent with NCEA’s wider approach to energy analysis. The Choices We Face grounds energy policy in “the laws of nature, fundamentals of economics, and standards of civil governance,” as well as what is “possible, practical, and reasonable.”

That standard applies to the tools we build as well as the research we publish.

Our Aspirations

NCEA’s ambition is to create a research environment in which high-quality energy analysis is easier to produce, inspect, update, and use.

We want AI to help NCEA:

• Investigate more questions with the same commitment to rigor.
• Update research products more efficiently.
• Make complex energy relationships easier to understand.
• Give users more direct access to data and methods.
• Connect published analysis to dynamic, public-facing tools.
• Improve the consistency and transparency of research workflows.
• Expand the reach of fact-based energy information.

The future of research will not be defined by AI-generated output alone. It will be defined by the combination of expert judgment, reliable data, well-designed models, and tools that help people understand the difference between evidence and assertion.

NCEA built a custom AI agent because energy research demands that combination. Through projects such as the Index of Leading Energy Indicators and NCEA-EMIT, we are working toward a model of research that is not only published, but continuously useful.

Have a question about energy? Ask NCEA Energy Chat →

NCEA’s AI projects are a continuous work in progress. Index of Leading Energy Indicators and NCEA-EMIT are under construction. Project architecture, data sources, models, interfaces, and release details remain subject to ongoing development and review.