When Bridget Carper Arnone joined the Department of Energy (DOE) in 2023, she remembers wondering why some processes still moved between paper and legacy systems when better technology was already available.
Then, technology could often feel like an add-on.
Today, “technology is part of the mission infrastructure,” said Arnone, DOE’s deputy chief information officer for Architecture, Engineering, Technology, and Innovation, chief data officer, and senior agency official for Geospatial Information.
She said the department’s scientists, engineers, and other mission employees are often eager for the newest capabilities – and expect the enterprise to keep up.
“If our staff hears about a new, better technology, they’re like, ‘All right, we need this yesterday,’” Arnone told MeriTalk.
That demand is shaping how Arnone sees the next phase of federal technology leadership. The question is increasingly how to give employees powerful new tools quickly while ensuring the cybersecurity, data, workforce, and budget foundations underneath them can support that pace.
Trust comes before the AI multiplier
Asked to rank DOE’s fiscal year 2027 (FY27) technology priorities, Arnone put cybersecurity and risk management first, followed by data management and analytics, and then artificial intelligence (AI) and machine learning.
“Trust and resilience underlines everything else,” Arnone said. “If people don’t trust the tools enough to put in their data and to leverage it, then we’re not going to have adoption.”
Trusted data comes next. DOE needs to know that information housed across different systems means the same thing, is accessible to the right users, and can support reliable decisions.
Only then, Arnone said, can AI do what DOE wants it to do.
“We look at AI as more of a multiplier,” she explained.
Leadership flipped the AI conversation
DOE has also seen what happens when the technology is ready, but the workforce is not yet convinced.
The department began rolling out Joulix, its generative AI tool suite, at the end of 2023. Arnone said the platform initially had just 1,000 users, even though it had security authorization and operated within DOE’s environment.
The turning point in adoption came from leadership. DOE undersecretaries, office leaders, and component CIOs began telling employees that they supported the technology, were using it themselves, and expected teams to explore it.
“The dialog flipped by leadership saying, ‘I’m using it,’” Arnone said.
Instead of employees asking whether they could or should use AI, conversations moved toward how to use it and which data sources could help.
Joulix has now grown to roughly 22,000 users, Arnone said, earning the team a Secretary of Energy Achievement Award.
That experience reinforced an important lesson for Arnone: Leaders have to actively create permission for people to change how they work.
Agents are shifting the work
With adoption growing, DOE is now looking at the next step: agents.
Arnone said her team is examining repeatable workflows that agents could handle, allowing employees to shift from manually performing every step to reviewing the result for quality.
The department’s technology onboarding process offers an example. Rather than having employees repeatedly carry out the same onboarding steps, Arnone said the team is exploring how an agent could observe and replicate the process.
Then, employees could focus on quality control and the decisions that genuinely require judgment.
DOE is also working with its cybersecurity leadership on how to secure agents and manage multiple agents operating within an environment.
Cybersecurity moves upstream
Arnone sees cybersecurity as both DOE’s top technology priority and one of the biggest potential sources of friction if agencies approach it incorrectly.
Cybersecurity and innovation cannot operate as competing forces, she warns.
DOE now brings chief information security officers and information system security officers into customer and technology discussions early so they understand how a capability will be used, what data it will touch, who needs access, and what protections will be required. Security documentation can develop alongside the technology rather than appearing as a final hurdle after user acceptance testing.
“It’s not, ‘We built it, and then now security needs to look at it,’” Arnone said. “They’re involved now the whole way through.”
The approach also gives DOE a chance to scrutinize whether every review is actually reducing risk. Arnone questioned the value of repeatedly reviewing a capability already operating in a controlled environment when those resources might deliver more risk reduction by red-teaming a genuinely new model or threat.
Arnone would like to see DOE move further toward enterprise-level risk decisions so different organizations do not repeatedly solve and authorize the same problem.
Success hinges on mission impact
Arnone ranked insufficient funding ahead of competing priorities as the bigger modernization constraint and said FY27 requires agencies to stop more low-value work rather than launch new initiatives.
“Our budgets aren’t going to increase,” she said.
That means leaders need to ask harder questions about systems and processes that have accumulated over time: Who is still using them? What value are they producing? Could they be consolidated or decommissioned so the resources can go somewhere more useful?
“We definitely, wholeheartedly agree that we have to change how we’re looking at things and potentially decommission systems that aren’t being used,” Arnone said.
And for Arnone, the same evaluation standards should apply to the new tools.
“We don’t look at success like, ‘Hey, we’ve put out these tools and we have these available,’” she said. “It’s actually the adoption. Have we reduced administration burden? Have we moved, increased the speed? Have we strengthened collaboration? Is there more scientific discovery?”
For Arnone, the measure of success is whether the technology is advancing the mission.