The Department of Energy continues expanding its enterprise artificial intelligence capabilities by combining user feedback, strong governance, and secure data management, according to Bridget Carper Arnone, Deputy Chief Information Officer for Architecture, Engineering, Technology, and Innovation and Acting Chief Data Officer.

Arnone says the department's AI journey has evolved significantly since the launch of Energy GPT in 2024.

What began with approximately 100 users has grown into Julix, an enterprise AI suite now supporting 20,000 users across 88 Department of Energy organizations.

The platform's name itself reflects the department's mission.

"We actually wanted to do an energy spin," Arnone says, explaining that the team looked at measurements of energy before selecting the name Julix.

Rather than developing capabilities based on assumptions, DOE expanded the platform in response to requests from employees across the department.

Human capital teams wanted assistance creating position descriptions. Other users requested help interpreting executive orders. Additional teams wanted collaborative AI tools for working on documents.

As those needs emerged, the platform continued to grow.

"The suite grew based on use cases that came in from the users that were leveraging Julix," Arnone says.

Not every idea moves immediately into production.

DOE evaluates proposed capabilities through a governance board that includes super users representing headquarters, laboratories, and field sites.

The department also establishes an enterprise threshold before investing resources.

Arnone says DOE generally looks for capabilities that can benefit roughly 20 to 25 percent of the organization before making them broadly available.

That process helps ensure enterprise tools provide value across multiple organizations rather than serving only a single office.

The department also collaborates closely with its national laboratories throughout the development process.

Although many laboratories already had experience using AI tools, DOE compared lessons learned across different models before expanding Julix within the department's secure environment.

One milestone Arnone highlights is web grounding.

Previously, AI models relied on information that might be several months old.

Julix now connects to current web information with an approximately 20-hour delay, allowing users to reference much more recent information while remaining within the department's environment.

"It helped adoption because now you don't need to use the public model when the DOE version is just as fast," she says.

Alongside Julix, DOE has introduced Quanta, a secure enterprise data and AI platform designed to strengthen how employees work with organizational information.

Arnone describes Quanta as the platform that brings together multiple AI models alongside DOE's public and non-public data sources.

Because "AI models are only as good as the data," she says, the department continues investing heavily in enterprise data strategy.

Quanta supports several commercial and open-source models while giving users secure access to organizational information needed for presentations, speeches, and other mission work.

The platform has experienced rapid adoption.

After beginning with only three offices, Quanta now supports 44 of DOE's 88 organizational elements, with plans to expand department-wide by the end of the fiscal year.

According to Arnone, leadership interest continues driving that growth.

Managers recognize that automating administrative work allows employees to devote more time to mission responsibilities.

Building these platforms also requires significant attention to cybersecurity.

While implementing technology has been relatively straightforward, Arnone says that secure integrations require careful planning to ensure organizational data is used appropriately.

"Technology is very simple. Ensuring that it's safe and secure is the little harder part," she says.

Looking ahead, DOE continues emphasizing data governance and standardization.

As additional systems are modernized, the department wants enterprise data to remain consistent and easier to integrate across platforms.

Quanta also helps address one longstanding challenge associated with unstructured information.

Documents that once required months to organize can now be processed in minutes, allowing employees to convert paper records and other unstructured formats into information that supports decision-making more quickly.

Throughout the discussion, Arnone emphasizes that enterprise AI success depends on more than technology alone.

By combining user-driven development, governance, secure architecture, and robust data management, DOE continues to expand AI capabilities that support employees across the department while maintaining trust in the information those systems produce.