NRC to reformulate its AI strategy amid OCIO reorg, other changes

The Nuclear Regulatory Commission is preparing to formally alter its AI plans following a reorganization of its office of the CIO and other changes, per top technology officials speaking Tuesday during the agency’s seventh AI Workshop. 

The regulator is planning to focus on repeatable, high-value use cases where success is measured by reduced administrative work and faster access to relevant information — not by the number or novelty of tools deployed. 

“A year from now, we don’t want our success story to be: we deployed five more AI tools,” Caroline Carusone, NRC chief AI officer, said in her opening remarks. “What we really want is to hear that an inspector spent less time on administrative work … we want to hear that a technical reviewer found relevant information faster, or that a newer employee could easily tap into decades of institutional knowledge.”

“Those are the outcomes that matter,” added the official, who also serves as an executive director for one of NRC’s programs. “And we need to achieve them without compromising the quality or independence of our regulatory decisions or the public’s trust in NRC.” 

The comments come just months before NRC debuts its “reformulated” and “refreshed” AI strategy, which is expected to reflect 2027 goals and beyond. 

“Over the last year, NRC has taken significant steps to move beyond AI exploration and directly into mission impact,” Carusone said. “Our objective is not just to chase the technology; it’s to apply it deliberately to the mission, and AI doesn’t have to be revolutionary every time we use it to be valuable.”

Carusone took over the chief AI officer position from CIO Scott Flanders in June, the same month that the NRC office of the CIO underwent a reorganization. A spokesperson said the reorg was required by executive order 14300 and was consistent with provisions outlined by the ADVANCE Act. 

“It’s exactly three months since the reorg, and one of the exciting things that happened in OCIO and directly under my division is we not only have AI and data but now we include application development,” Basia Sall, chief data officer at NRC, said during the event. “As a CDO, that gave me full access to the entire data lifecycle.”

NRC wants the additional visibility to inspire improved data quality and governance — key ingredients to AI success. 

“AI is not going to work if we don’t have the data,” Sall said. 

As part of the reorg, per the official, there is also a new branch dedicated to data analytics. 

“We used to have them placed across the agency,” Sall said. “Now, we’re one, and we’re able to support [the organization] holistically.”

The NRC spokesperson said the reorg included consolidating IT functions, establishing new leadership roles for data management and AI, and strengthening support for emerging technologies.

“The new model brings technical staff together within DATA and embeds them back into program offices to identify AI opportunities grounded in mission needs,” the spokesperson said via email. “This approach strengthens partnerships and governance, reduces silos, and accelerates secure, practical solutions that improve decision-making and advance the NRC’s mission.”

Many federal agencies have taken a look at their OCIO structures and made alterations in the past year or so, including the departments of Homeland Security, Transportation and Interior. 

The NRC is working to combine digital work, too, bringing together its AI governance, data governance and investment review boards. 

“Our CIO Scott Flanders has a vision that we’re excited to execute on, starting tomorrow,” Sall said. “So, more to come [and] that then builds onto our steering committees that really get to our other structures in governance.”

While the NRC teams have undergone major change in recent months, more is on the horizon for the regulator and its stakeholders.

The NRC has been working to redesign its website since January. The agency debuted a prototype for the public in April and underlined its desire to improve user experience through the modernization effort. The new website is set to soft launch later this week. 

“Once that’s settled, we’re going to go ahead and update all the AI and data pages,” Sall said. “You’ll be able to see in ‘27 some fresh content on there and more in real-time information.” 

AI approach

Amid the changes, NRC has relatively stayed the course on its metered approach to AI adoption, a path that stands in stark comparison to the rapid pace of expansion and hundreds of use cases tallied at other agencies.

When giving examples of NRC’s current AI portfolio, Carusone pointed to well-documented use cases: Microsoft 365 Copilot and the agency’s in-house tool SimplifAI. Those are two of the four line items in the regulator’s most recent AI use case inventory. 

The agency is also evaluating shared services, like the General Services Administration’s USAi. 

“Beyond these tools, NRC is exploring AI-driven efficiencies in licensing, rulemaking and inspections,” Carusone said. “AI doesn’t replace the expertise, but it can help our experts spend more of their time actually using it.”

The regulator’s top AI leader said generative AI training and related policies for employees and other stakeholders are expanding in line with the NRC’s shift in AI maturity. 

“Adoption goes well beyond learning how to write a good prompt,” Carusone said. “One of our focuses is really having our employees know when AI is appropriate, how to validate the output, how to challenge the result and when human judgment must prevail.”

As part of its regulatory function, the agency is looking into its response to applications or licensees that use AI. 

“Our job isn’t to decide whether AI is inherently good or bad, but what we do need to know is we need to understand what it does, what evidence it supports and how it could fail,” Carusone said. “That means that we need to be building expertise now, evaluating where regulatory approaches may need to evolve and engaging industry as these technologies develop.”

Part of that work is being done through NRC’s collaboration with the Department of Energy and its Genesis Mission. 

The initiative aims to capitalize on AI, quantum and high-performance computing gains to improve scientific workflows. One of the core focuses has been the intersection of AI and the nuclear energy industry.

DOE ran an experiment recently to streamline the nuclear regulatory process by condensing license application timelines, an effort it’s referred to as a “milestone test case.”

There are other projects, too.

An initiative called Prometheus, for example, is housed under DOE’s Genesis Mission and is an effort to accelerate nuclear energy deployment. Industry partners have committed more than $200 million in cost share and $30 million in capital, proprietary design data and commercial deployment pathways, according to the Idaho National Laboratory, one of the project’s leads.

Federal investment in the project has totalled $60 million. 

The project is a three-year effort to build a platform that can speed up multiple levels of the nuclear energy lifecycle, from manufacturing and construction to licensing and operations.  

“It’s going to be really important to have good relationships with the NRC to make sure we’re proceeding in the right direction,” Peter Suyderhoud, INL’s project lead for Prometheus, said during the event Tuesday.