Washington is focused on the wrong aspects of artificial intelligence.
Discussions in federal government circles center almost entirely on safety measures, purchasing regulations, and oversight structures. While these issues are legitimate—and I certainly don’t dismiss them—they overlook a more subtle danger that merits serious consideration: What occurs when AI gradually teaches government workers to arrive at the same conclusions?
This phenomenon is already underway. Evidence of it is visible in the way work products are beginning to look and sound alike across federal departments.
As generative AI becomes deeply woven into everyday agency workflows—composing memos, summarizing policy alternatives, and assembling briefing materials—it inherently gravitates its outputs toward a statistical average. This isn’t a defect; it’s how the systems were engineered. AI processes enormous datasets and produces outputs that mirror established patterns. The outcome is intellectually homogenizing: proposals that are logically organized, well-justified, and nearly identical to each other. Streamlined, certainly. But for a government that routinely confronts unprecedented, intricate, and high-pressure decisions, efficiency rooted on uniformity is a concealed liability masquerading as progress.
Consider what the federal government is tasked with. It frequently addresses situations that lack clear precedents—pandemics, cybersecurity breaches, breakdowns in supply chains, and sudden geopolitical upheavals. In those moments, the significance of human reasoning extends far beyond following established procedures. It lies in the ability to spot what the data misses, to pose questions that no dataset ever anticipated, and to balance conflicting concepts long enough to discover a creative alternative. When teams across multiple agencies all rely on identical AI-produced analyses and interpretations, this intellectual variety slowly deteriorates.
Having spent years navigating the overlap between corporate reinvention, government, and academia, I’ve witnessed countless organizations confuse operational improvements with genuine strategic progress. AI integration in the public sector is dangerously prone to repeating this mistake. Leaders are evaluating success through adoption metrics—such as how many tools have been deployed or how many processes have been automated. But the true measure of success is far harder to gauge: Have people’s thinking capabilities improved as a result of AI, or has their thinking diminished?
That question carries enormous weight in government, where policy choices have impacts that accumulate and intensify over time. A recommendation may appear sound solely because it aligns with what the AI recommended, yet it could be precisely the type of recommendation that collapses catastrophically under the very conditions that determine an organization’s reliability and credibility with the public.
The staffing reality intensifies this urgency rather than diminishing it. Agencies have undergone substantial workforce reductions. Those who remain are expected to accomplish more, at greater speed, with AI serving as the bridge to productivity. Under those pressures, the natural tendency is to defer to AI-generated outputs uncritically. You adopt the AI’s perspective, incorporate its terminology, and proceed to the next assignment. That response is completely understandable. However, it’s also how organizations silently erode the independent reasoning that genuinely makes policy work. It’s how they become trapped by the AI era’s greatest misconception: that productivity and speed are all that matter.
So what steps should leadership take?
To begin with, acknowledge that leveraging AI is not the same as comprehending it. Make certain that every government employee recognizes these tools are merely the beginning. In truth, when generative AI is employed thoughtfully, it can strengthen and enhance human insight and creative reasoning. People are naturally inclined to seek originality and value fresh thinking. Given the freedom to choose, we apply any resource—including AI—to advance that drive. But AI is a tool unlike any other: its continuous interaction with the user demands more than a basic instruction manual. It calls for training that establishes a higher bar—a clear understanding of how AI enhances a particular task, and an appreciation for the individual’s distinct skills in applying it within their specific position.
Next, design AI integration strategies that are tailored to specific roles, rather than implementing a one-size-fits-all approach. Not every government function values originality equally. Automating routine administrative work is simple and delivers real value. But positions requiring analysis, judgment, discretion, or responsibility to the public operate on a completely different level. Handling them identically is how departments unintentionally automate away the very qualities that make governance trustworthy.
Finally, monitor cognitive skills over time rather than focusing only on technology adoption. The question agencies need to pose isn’t “What percentage of our staff uses AI?” It’s “Are our employees sharpening their judgment or growing increasingly reliant on the tool?” These are fundamentally different paths, and they produce radically different outcomes for the workforce. Leaders who lack a clear, honest answer to that question are operating without insight.
The federal government is placing enormous bets on AI at this critical moment. Some of those bets will yield positive returns. But defining success solely through the lens of speed and productivity is insufficient. A government whose workforce converges on uniform thinking, amplified across the entire system, is a government less capable of addressing the challenges that fall outside the algorithms.
The AI era doesn’t reduce the demand for original human thought in public service. On the contrary, it makes such thinking more precious and more vulnerable than ever before. Washington needs to begin adapting its policies to reflect that reality.
Jonathan Aberman is co-founder and CEO of Hupside, a partner at Ruxton Ventures, and the founding dean of Marymount University’s School of Business, Innovation, Leadership and Technology.
Copyright
© 2026 Federal News Network. All rights reserved. This website is not intended for users located within the European Economic Area.
