As federal agencies accelerate artificial intelligence adoption in line with the administration’s AI Action Plan, the conversation has quickly shifted from whether AI belongs in government IT to how agencies can deploy it safely, responsibly and cost-effectively at scale.
In mission-critical federal environments, “mostly right” is dangerously insufficient. Federal IT systems support everything from identity and access management to emergency response operations, endpoint security, citizen services and national security infrastructure, and a single flawed automated action can disrupt operations, introduce vulnerabilities and erode public trust.
On the flip side, AI has already demonstrated its potential to improve how agencies navigate complex IT environments. Reducing alert fatigue, accelerating ticket resolution and streamlining patch management are all ways in which AI has improved workflows for IT teams. Though when it comes to weighing where and how to most effectively incorporate AI in federal IT environments, it’s important to remember that automation alone shouldn’t be treated as a substitute for human capability. Like anything else in business or operations, technology is only one piece of a larger puzzle. When it comes to architecting or recalibrating for AI optimization, your people and your processes play equally critical (and irreplaceable) roles.
Workforce tension
The efficiency argument for AI in federal IT has always carried an uncomfortable subtext, and recent events have made it impossible to ignore.
As agencies face significant workforce reductions, AI is increasingly being positioned not just as a tool to help existing staff work better, but as a justification for having fewer staff in the first place. That is a meaningful distinction. When AI augments a full team, there are humans available to catch the 5% of cases where automation gets it wrong, to apply contextual judgment, and to course-correct when something unexpected happens. When AI is deployed into an environment where that institutional knowledge and human capacity have already been cut, the margin for error shrinks to nearly nothing.
The Massachusetts Institute of Technology studied whether it made economic sense to swap humans for AI across roughly a thousand computer vision tasks and found that in many instances, it didn’t. When you add up implementation, maintenance, hardware, and the separate layer of staff you still need to check the AI’s work, the total exceeds what the human cost would have been.
For federal agencies, the result is a kind of compounded risk that neither the efficiency advocates nor the AI vendors tend to advertise: agencies that are simultaneously less staffed and more dependent on automated systems they may not fully understand, govern, or have the capacity to audit or financially maintain long term. Even as AI pricing models shift, agencies today are signing multi-year contracts and rebuilding workflows around today’s rates, making long-term bets on prices that the vendors themselves aren’t covering.
None of this is modernization. It is fragility and hype cycle-induced adoption cycles dressed up as progress, with the consequences still yet to be fully seen.
The case for a “trust but verify” AI strategy
Knowing where and how to use AI and ML to augment human expertise, and not replace it, is one thing. But when it comes to the technology itself, federal agencies need to adopt a “trust but verify” approach. One that allows them to adopt and scale automation efforts while maintaining operational oversight, accountability and compliance.
The federal government is already moving in this direction. The Office of Management and Budget’s Memorandum M-25-21 requires agencies using high-impact AI systems to implement minimum risk management practices, including ongoing performance monitoring, with an obvious impetus to discontinue AI use if systems are not performing appropriately.
In practice, that means keeping humans involved in high-impact operational decisions, particularly actions that affect endpoint configurations, identity and access management, security policies and mission-critical infrastructure. AI can summarize complex data, identify anomalies, recommend remediations and automate repetitive workflows. But human operators should remain responsible for validating recommendations, approving sensitive actions, and overseeing how automated decisions are executed.
A successful “trust but verify” strategy also requires agencies to prioritize AI systems that are transparent, auditable and governed by clear operational guardrails. IT teams need visibility into how AI systems are making recommendations, what data those recommendations are based on, and how automated actions are being carried out.
Not every use case carries the same level of risk. Lower-risk tasks like ticket summarization and alert prioritization may be appropriate for greater automation. Higher-risk actions like access changes, script execution and system configurations, however, should require stricter permissions, review processes and more rigorous levels of human oversight.
With the right safeguards in place, agencies can modernize confidently while preserving resilience, trust and operational integrity.
Responsible AI starts with solving real problems
As agencies evaluate AI investments, another reality is becoming increasingly clear: Not every problem requires an AI solution. The rush to add AI into every workflow risks creating additional (and often unnecessary) friction, not to mention greater costs, governance gaps, and security exposure instead of reducing administrative burden. Successful AI strategies begin by identifying operational pain points first, then determining where targeted AI use cases can help safely and meaningfully improve outcomes.
In federal environments, intentional innovation will always outperform innovation driven by hype. This is especially important as agencies confront already-murky IT environments growing further obscured with shadow AI.
Without clear guardrails, oversight mechanisms, and communication between IT teams and end users, agencies risk exposing sensitive data, creating compliance gaps, or introducing further operational risks by overindulging in AI.
Responsible AI adoption requires more than deploying new tools. It requires governance models, internal policies, employee education, and operational frameworks that evolve with the technology itself.
Federal IT must balance automation with accountability
As budgets shrink, AI poses a real opportunity for federal agencies to improve efficiency, strengthen resilience and modernize operations.
But only if adopted thoughtfully and meaningfully, and not in a futile attempt to curb expenditure — particularly on human talent. It’s important to keep in mind that the current AI pricing model reflects competitive pressure and investor-backed growth spending, not sustainable unit economics, and price increases are already arriving through quieter channels (model gating, usage caps, feature paywalling).
With the AI IPO queue forming, SpaceX, OpenAI, Anthropic and others will soon face public markets that need actual profitability timelines. When that pressure arrives, prices will go up, and agencies that have by then restructured around AI will be left holding the bag.
Intentionality is the name of the game when it comes to AI. Thoughtful adoption. Specific use cases. And keeping a human in the loop to trust but verify AI-driven outcomes along the way. The efficiency gains posed by AI are there, but only if organizations and agencies stay cautious and pragmatic in their approach, keeping trust and security at the center of their adoption strategies.
Egon Rinderer is senior vice president of global enterprise and public sector at NinjaOne.
Copyright
© 2026 Federal News Network. All rights reserved. This website is not intended for users located within the European Economic Area.
