Energy demand is expected to increase significantly due to the increasing use of artificial intelligence (AI) technologies, which are very energy intensive. As we noted in a recent Thought leader contributionthere is a close intersection between nuclear power and AI. Not only is nuclear power generation well positioned to meet the growing demand for AI power, but the nuclear power industry and its main regulator, the U.S. Nuclear Regulatory Commission (NRC), are looking to leverage AI to increase efficiency and strategic decision-making.

NRC staff recently published a report Identifying how AI can be used to improve work within the NRC. The report is the result of an October 2023 memorandum from NRC Chairman Christopher Hanson in which he “directs NRC staff to examine how AI can streamline operations, optimize processes, and make informed decisions.”

The report follows the publication of NUREG-2261 in 2023, also known as the AI ​​Strategic Plan (Strategic Plan), in which the NRC acknowledged that while it does not currently use AI technologies, it “expects to increase the use of AI in NRC-regulated activities.” The NRC explained that the purpose of the strategic plan is to ensure the readiness of NRC staff to review and evaluate the use of AI in NRC-regulated activities. Now, in the report, the NRC provides recommendations on how it can leverage AI technology within the agency. With the report, the NRC continues its efforts to recognize the role of AI technology in the nuclear industry.

The process

To conduct the review described in the report, the Executive Director for Operations assembled a team from the Office of the Chief Information Officer and the Office of Nuclear Regulatory Research (AI Team). The AI ​​Team asked staff to evaluate and propose ways the NRC could use AI to improve its work, called “AI use cases.” The AI ​​Team then worked with NRC data scientists and outside AI experts to evaluate whether the proposed AI use cases were feasible.

The results

The AI ​​team identified 36 potential AI use cases that would increase employee productivity and support the workforce. Of the 36 cases, 16 would “leverage AI to automate routine tasks and implement workflow and process improvements, making many of the daily tasks employees perform more efficiently.” An example of these AI use cases is summarizing meeting minutes, documents, and web pages. The remaining 20 AI use cases leverage other forms of “AI to perform predictive analytics, automate the review of public comments on proposed regulations, and assist inspectors in planning and scheduling their availability in the regions.”

Next Steps

Staff identified two next steps for the NRC to successfully implement AI. First, they recommend that the NRC “develop an enterprise-wide AI strategy to advance the use of AI within the agency.” This enterprise-wide AI strategy would ensure that “the NRC is consistent with federal policy that federal agencies use AI in a safe and responsible manner” and would include preparing AI governance to ensure responsible and trustworthy AI implementation preceding the full rollout of AI.

In addition, the agency management program would evolve by implementing and supporting data architecture and data management while ensuring that the entire NRC workforce has the data skills necessary to fully understand how data supports and improves the agency’s work. This would also include strategic hiring and upskilling of the existing workforce to ensure workplace adaptability. In addition, it would include allocating resources to support the integration of AI tools as part of the IT infrastructure.

Second, the staff recommends that the NRC “invest in foundational tools to advance the use of AI.” Specifically, in fiscal year 2025, the staff plans to invest in two foundational AI tools that will cover multiple AI use cases and facilitate staff learning and understanding of various AI tools needed to develop other AI cases. The staff recommends focusing on governance, data readiness, and training prior to implementing these foundational AI tools.