Do more with less has been the federal mantra for so long that for many it stopped meaning anything. This year it grew teeth. The administration’s push on efficiency and effectiveness is showing up in concrete form, from workforce reductions to GSA’s recently published Eliminate, Optimize, and Automate playbook, IT leaders are being asked to prove impact, rather than defend headcount.

That was the terrain for the latest GIST 360 webinar, From the Edge to the Enterprise, hosted by Swish and HPE. Sean Applegate, CTO of Swish, moderated a candid conversation with Preston Werntz, acting CIO at CISA, and Ken Rich, Federal CTO for HPE.

Mr. Wertz opened with a reframe that framed the conversation. You do not actually do more with less. You do less with less, and that is not a bad thing, because it forces the prioritization that abundance always let agencies avoid.

The Next Five Years Are Not a Mystery

Federal IT rarely gets to plan against a clear horizon, but this is one of those moments. Nearly all of the work ahead falls into two buckets, the transition to stronger security and the move into AI. That clarity should discipline every dollar. Investments made now need to serve those two ends, and legacy systems should be retired with the same logic. The refresh cycle becomes the lever. Keep what is stable and reliable, replace what is going end of life before it becomes a vulnerability, and standardize the operational model so the network is simpler to automate.

Optimizing a footprint bloated by years of additions is itself a form of doing more with less. This is also where federal mandates converge. Zero trust targets and the government’s AI agenda are not competing priorities. They are the same modernization, and agencies that treat them as one program will move faster than those funding them apart.

Telemetry Is the Whole Game

The conversation kept circling back to a single unglamorous concept, it’s all about context. AIOps and self-healing networks make big promises, and skepticism is warranted, because autonomy over a production network should be earned rather than assumed. But the real constraint is not the model. Models are already capable, and each new one is an incremental gain rather than a leap. What separates a useful AI operations capability from a reckless one is the telemetry feeding it.

A generic model dropped onto a network knows nothing about that network. The systems that work are trained on the environment they run in, fed continuous observability, and pointed at the tasks AI genuinely does well, such as root cause analysis, baseline drift, change validation, and compliance checks. The architecture decisions and the final risk calls stay with people. That division of labor, low risk and reversible actions to the machine and high risk and irreversible ones to the human, is what earns the autonomy back over time.

The 10X Engineer, Not the Smaller Team

Participants throughout the webinar opined that the workforce fear around AI is misplaced, and the lesson from software already landed. AI did not shrink engineering teams, but rather it has given a senior engineer the output of an entire team, and the same shift is arriving for network operations.

Picture walking in each morning to a network that has already flagged the port flapping, the weak wireless coverage, and the circuits buckling under low-priority traffic, waiting to be told what to fix, with changes that roll back cleanly.

That is leverage, not replacement.

It matters most in a workforce stretched thin and spread far beyond headquarters, with staff in the field, in partner facilities, and in tactical settings where compute now runs at the edge because a decision cannot wait for a round trip to a distant data center. The payoff is human. When routine toil is handled, people stop reacting to individual tickets and start seeing the systemic problem underneath them, which is exactly the judgment the mission most needs.

How IT Leaders Manage Risk to Ensure Successful Adoption

None of this works without guardrails, and recent headlines about AI agents escaping their sandboxes made that vivid. In most cases the agent did exactly what it was told with too little context, which is the whole lesson. The answer is not to avoid agents but to prepare the ground before turning them loose. That means a real data inventory, labeled data, identity built on roles and attributes, and cost controls that can throttle an overeager agent before it burns through bandwidth and budget.

IT leaders can manage risk by managing what the AI is allowed to do. Move quickly where the risk is low and reversible and stay deliberate everywhere else. Budget discipline follows the same logic. With the fiscal year closing, the strongest investments lower the cost of the network over time, through telemetry, standardization, and reduced complexity, rather than simply buying another tool bolted on to paper over the last one. Expect early inefficiency with AI, start with one or two use cases, and bring the workforce along as fast as the capabilities arrive.

Watch It. Share It. Then Act on It.

The efficiency mandate is not going away, and the agencies that thrive will treat security, AI, and modernization as one connected build rather than separate programs. Watch the webinar on-demand on the GIST 360 platform, and bring your network, security, and AI leads into the conversation. Explore the podcast, whitepapers, and other recordings at gist360.com, and tell us what you want covered next. When you are ready to map where your infrastructure stands against where the mission needs it, reach out to the team at Swish.