A lot of federal leaders hoped artificial intelligence would be a productivity tool.

In many cases, it is. Tasks that once required hours now take minutes. Workflows that once required people can increasingly be automated. The catch is that every productivity gain seems to arrive attached to a new oversight problem.

Every AI deployment arrives with companions: new identities, new interactions, new attack surfaces and another layer of governance, said Felipe Fernandez, chief technology officer at Fortinet Federal. Leaders have to determine what information an AI can access, what systems it can interact with and how its activity will be monitored.

Fernandez compares the experience less to deploying software than to hiring an employee. That helps explain why AI conversations inside government increasingly revolve around guardrails rather than capability.

“It’s not just a set-it-and-forget-it, fire-and-forget type tool that we used to have,” Fernandez said during Federal News Network’s Cyber Leaders Exchange 2026. “It is certainly like adding a very talented human to the environment.”

Agencies have found work for AI agents. They have also inherited the obligation to supervise them. As those agents gain access to data, tools and mission workflows, oversight becomes part of the job.

Keeping that work inside clearly defined boundaries is becoming a significant management task. The more autonomy agencies give these systems, the less realistic it becomes to assume every unexpected action can be prevented in advance.

Plus, it’s essential that agencies focus on creating resilient operations. “We have to think about operating through incidents, not just recovering from them,” Fernandez said. “We have to understand we’re operating in an environment where we assume something bad has already happened, something bad is happening, something bad will happen.”

Instead of focusing exclusively on keeping adversaries out, agencies are increasingly asking whether critical services can remain available when a cloud service fails, a system is compromised or a key resource becomes unavailable. Simply preventing disruption is no longer sufficient, he said. The objective has shifted to sustaining operations during disruption.

That philosophy quickly runs into the physical reality that many federal networks were designed for a world where users interacted with applications. Today, applications interact with other applications. Automated workflows interact with other automated workflows.

AI agents increasingly communicate with systems, data repositories and other agents. That creates demands that older infrastructures were never designed to handle.

“The infrastructure wasn’t going to be laid out forever, like roads and highways,” Fernandez said. “It needs to be upgraded to support the additional traffic that’s being put on it.”

As agencies deploy more AI-enabled capabilities, their networks are carrying traffic they were never designed for. AI certainly consumes more computing power, but it also changes the shape of the traffic itself, Fernandez said. Greater bandwidth, more flexible architectures and higher-capacity security platforms are necessary to keep pace with the volume of machine-to-machine activity.

And AI isn’t the only modernization project competing for attention. At the same time, many agencies are preparing for post-quantum cryptography, placing an even greater demand on their networks. Before they can replace vulnerable encryption, agencies first need to understand where that cryptography exists, which applications depend on it and whether aging hardware can support newer approaches.

Fernandez said many organizations are still in the inventory phase, identifying dependencies before they begin large-scale migration efforts.

AI is generating productivity gains. It’s also creating new responsibilities around governance, resilience and modernization. That’s great news, Fernandez pointed out. But, he added, agencies that hoped automation would reduce work must be ready for the ensuing governance, resilience planning and modernization efforts that cannot be postponed.

Discover more articles and videos now on our Cyber Leaders Exchange 2026 event page.

Copyright © 2026 Federal News Network. All rights reserved. This website is not intended for users located within the European Economic Area.