Kelly Fletcher thinks this may be the best time yet to work in technology.
“It’s the coolest time to be a technologist,” the Department of State chief information officer (CIO) told MeriTalk.
Generative artificial intelligence (AI) is one reason. Fletcher sees the technology carrying enormous implications for cybersecurity and for how quickly government can respond to its customers.
“It is truly generational technology,” she said. “We’re going to see this big of a change once in our lifetimes. This is that moment.”
But Fletcher is just as excited about something that gets far less attention: connectivity.
The State Department supports roughly 100,000 users across 190 countries, including remote locations where bandwidth and latency challenges can make technologies that work seamlessly in Washington much harder to use. Fletcher points to advances in satellite communications as a potentially game-changing way to improve that global communications foundation. Her point is simple: The most advanced application in the world does little good if employees cannot reliably reach it.
“If the Wi-Fi doesn’t work, no one wants to hear about AI,” she said.
That grounding shapes much of Fletcher’s agenda heading into fiscal year 2027 (FY27). State is pushing aggressively into AI agents, frontier models, and the shift toward more automated cyber defense.
At the same time, Fletcher is focused on the less glamorous work required to make those capabilities useful every day – stronger infrastructure, better user experiences, workforce adoption, and tighter accountability with vendors.
Democratizing access to AI agents
State dubbed 2026 “the year of the agent and the app,” Fletcher said.
The department expanded modern applications, particularly in its classified environment, rolled out generative AI at the classified level, and introduced Beacon, an administrative AI agent embedded in Microsoft Teams.
Fletcher is candid about where Beacon stands today.
“Is it good today? I would give us a C,” she said. “But it’s going to be a bloody A-plus. Give me three months.”
The agent is designed to chip away at the small administrative tasks that accumulate across a workforce of 100,000 people. Rather than asking an employee to navigate several systems to find leave information, request IT support, or approve an equipment request, Fletcher wants Beacon to understand who the user is and complete more of the workflow from one place.
Today, for example, a supervisor can receive an employee’s IT request directly through Teams and approve it there rather than follow a link into another system.
One interaction might save less than a minute. Across thousands of employees and repeated transactions, the productivity boost is substantial.
State is initially focusing Beacon on human resources, employee moves and logistics, and IT support. The larger FY27 ambition is to move beyond a centrally built administrative agent.
“I want to democratize agents,” Fletcher said. “I want customers to be able to build their own agent for their own workflow.”
That shift will run alongside continued infrastructure optimization. The department is rolling out a new laptop form factor that moves from perimeter-based security to application-layer security. The goal is to give users in high-latency, low-bandwidth posts a noticeably faster, more reliable experience when accessing core State applications.
Driving adoption through iteration
State’s AI work taught Fletcher a lesson she did not expect.
“I thought building the thing would be the hard part,” she said. “You know what’s so much harder than that? Getting people to use the thing.”
State rolled out generative AI broadly two years ago, but it took “months and months” to get even half of the department to use it. The experience convinced Fletcher that deploying AI requires an intentional adoption campaign alongside the technology itself.
For her, that means repeated communication, top-down support, grassroots engagement, and a feedback loop that shows employees their input is changing the product.
State’s AI Champions program is one piece of that strategy, giving employees local advocates who can show colleagues what the tools can do. Fletcher said her team also built trust by responding quickly when users asked for improvements.
“When they say, ‘I want it to do more of this,’ and then three weeks later it does that, they believe that it matters,” she said.
AI use cases are also moving closer to State’s core diplomatic mission.
Fletcher describes a two-pronged AI strategy: reduce administrative toil and help advance diplomacy.
On the mission side, she wants AI to help diplomats quickly understand patterns such as how other nations have voted alongside the United States in international bodies and where engagement may have the greatest value.
The goal, as she puts it, is “foreign policy dominance” – putting relevant information in diplomats’ hands quickly enough to sharpen decisions and engagement.
Turning detection into action
The push toward operational use is also shaping Fletcher’s cybersecurity agenda.
State is already using frontier models to examine both systems in development and systems in production for vulnerabilities. Fletcher sees an advantage in identifying problems before new applications ever reach production – and increasingly, she sees models helping teams determine how to fix them.
The next step is turning what remains highly specialized work into routine, automated defense.
“We’re seeing automated offense. We need automated defense,” Fletcher said.
Today, some of State’s best cyber specialists are using frontier models to interrogate code, identify vulnerabilities, and determine patches. Fletcher wants that process built into normal operations rather than dependent on one-off efforts.
That includes automating patching wherever possible.
“This has been true my entire career,” Fletcher said. “Finding a vulnerability, what that does for you is exactly nothing. Patching the vulnerability, creating an alert for the exploit of that vulnerability – now you’re doing something.”
State is pairing the push for automated patching with strong alerting, reducing end-of-life technology, greater use of software-as-a-service capabilities that inherit automated patches, and continued segmentation. The department already has macro-level segmentation and is working toward more microsegmentation across individual systems.
Configuration management and application programming interface inventories are also part of the effort.
With more than 500 systems, Fletcher does not pretend State can eliminate every weakness.
“We’re not going to be perfect,” she said.
The job is to concentrate effort where it produces the most risk reduction – and increasingly, automate it at the speed threats demand.
Owning the outcome together
Fletcher sees execution, rather than authority, as the larger challenge facing CIOs. She also thinks agencies need to get better at deciding what not to do.
Asked whether FY27 calls for starting more initiatives or stopping more low-value work, she chose the latter.
“I think that’s a place where we have struggled to be successful,” Fletcher said.
Asked to choose between insufficient funding and too many competing priorities as a top modernization constraint, Fletcher pointed instead to “a failure to prioritize appropriately.”
At State, she said she has leadership support, resources, funding, and a strong team. One thing she wants is more collaborative relationships with vendors – which the department is starting to see, Fletcher noted.
Fletcher said State is receiving significantly more patches from vendors and increasingly working with partners that proactively explain risks, brief the department on their response, and help solve problems rather than simply point them out.
She wants that model to become the norm – particularly as the government prepares for quantum-resistant cryptography and increasingly depends on the cybersecurity of commercial technology providers.
“The idea that a vendor will come in and say, ‘Oh, we see that you’re not doing this right’ – like a passive-aggressive shrug. We’re not doing that anymore,” Fletcher said. “You own that with me.”
“We have a true partnership with a subset of vendors that … stand arm in arm with us and help us do what we need,” she said.
For Fletcher, that is where federal technology leadership appears to be heading more broadly: closer collaboration between builders and users, shared ownership between government and vendors, and more accountability for mission success.
“[We want] true shared ownership of outcome-based delivery,” she said.
More from MeriTalk’s Federal CXO Spotlight series:
- Basia Sall, Chief Data Officer, Nuclear Regulatory Commission
- Adam Starr, Chief Information Officer, Office of Personnel Management
- Mangala Kuppa, Chief Information Officer, Department of Labor
- Dawn Zimmer, Chief Information Officer, Department of Energy
- Kimberly McManus, Acting Chief AI Officer, Department of Veterans Affairs