The federal government just unveiled its grand vision for AI adoption—without a plan for the people who'll make it happen.
Last month, the White House's Office of Management and Budget released two memoranda that shift how our government will approach artificial intelligence. These documents—M-25-21 and M-25-22—lean toward faster implementation and reduced regulation, establishing new mandates for AI governance, procurement, and innovation across all federal agencies.
Imagine giving a complex aircraft to pilots who've never received training—that's essentially what we're doing with AI in government right now.
There's just one problem: the people tasked with implementing these ambitious directives haven't been prepared for this moment.
Imagine giving a complex aircraft to pilots who've never received training—that's essentially what we're doing with AI in government right now. While these memos lay out what agencies must do, they offer remarkably little guidance on how the government workforce will develop the necessary skills to execute this vision responsibly.
Policies Need People
The success of this new vision for AI depends entirely on whether public servants are equipped with the skills and understanding needed to translate guidance into real-world impact. Training is not a nice-to-have. It is the foundational infrastructure that will determine whether this transformation is thoughtful, equitable, and sustainable. And right now, this element is missing in the current White House AI policy landscape.
And yet, there is a glaring absence of human infrastructure to support this directive. There are no centralized training repositories, no clear onboarding requirements, and no defined skills benchmarks.
Both memos acknowledge the importance of staffing, especially with directives to establish Chief AI Officers and governance boards. However, AI is not something that only technical experts will touch on. It is already impacting nearly every part of government operations, from legal and procurement offices to frontline service design and delivery. This is underscored in section ii.F of Memorandum M-25-21, which calls on agencies to “recruit, hire, train, retain, and empower an AI-ready workforce” and achieve “AI literacy for non-practitioners involved in AI.” In other words, AI fluency is not just for computer scientists. It is for every decision-maker, implementer, and oversight actor involved in bringing AI into the public sector responsibly.
And yet, there is a glaring absence of human infrastructure to support this directive. There are no centralized training repositories, no clear onboarding requirements, and no defined skills benchmarks. The memos do not reference any strategy for leveraging existing federal programs, partnering with academic or nonprofit providers, or deploying open-access training already available. The “what” and “how” of training remain undefined. This is not just happening with these memos, but it’s the same problem with the White House’s new executive order on AI in education which aims to prepare students for an AI-driven future, yet it coincides with significant cuts to education funding and research, raising concerns about its effective implementation.
What Effective AI Training Could Look Like
So what could effective AI training for the government look like? Ideally, it is tailored to specific roles, grounded in real-world context, and aligned with the day-to-day realities of public service. Procurement officers need the ability to assess vendor claims and prevent vendor lock-in. Legal and policy teams must understand how to oversee algorithmic systems and ensure accountability. Program managers need tools to evaluate where AI can improve outcomes in meaningful ways. And everyone, from agency executives to frontline staff, should have a shared vocabulary that supports collaboration and informed decision-making. Building this kind of capacity will require a combination of formats, including online courses, interactive workshops, peer learning sessions, and active communities of practice within and across agencies.
At InnovateUS, where I lead partnerships, we deliver free, self-paced online courses tailored to public sector professionals, covering topics like integrating generative AI tools into government work and developing AI policies for jurisdictions.
Some of the most promising training efforts today are co-created with public servants to ensure relevance and usability. These programs often include asynchronous modules that build foundational understanding, along with live sessions that focus on practical applications and discussion of real use cases. When aligned with agency goals and workforce needs, these blended formats allow for more effective uptake.
The General Services Administration’s AI Community of Practice used to offer a training series in partnership with institutions like Stanford HAI, George Washington University Law School, and Princeton’s Center for Information Technology Policy, providing specialized tracks tailored to various government roles, but is no longer running.
At InnovateUS, where I lead partnerships, we deliver free, self-paced online courses tailored to public sector professionals, covering topics like integrating generative AI tools into government work and developing AI policies for jurisdictions. These collaborative efforts underscore the importance of diverse, role-specific training programs and the value of cross-sector partnerships in building a competent and confident AI-ready public workforce. What is still lacking is greater coordination, larger-scale investment, and a stronger recognition that workforce development must be treated as a core element of any national AI strategy.
The Risk of Moving Fast Without Skills
If legal teams, mission owners, IT leads, and contract officers do not speak the same AI language, collaboration slows, risk increases, and opportunities are missed. Without a common baseline of understanding, AI procurement becomes fragmented at best and dangerously uninformed at worst. Training is not just a tool. It is a prerequisite. And we need to be precise about what roles need what knowledge, and how to teach it effectively.
The call to prioritize high-impact AI use cases, and to decrease governance only for “high-risk” use cases, will only deepen the urgency. As agencies move toward deploying AI assistants and automated agents that operate in daily service delivery, the complexity of those systems and their real-world consequences will grow. This is not about future potential. It is about preparing public servants today to navigate systems that will shape eligibility determinations, public communications, and even policy enforcement. Without broad, scaled training, these tools risk becoming black boxes — used but not understood, deployed but not trusted.
Deploying new AI systems can take years. But training can start immediately, and the return on investment is fast and measurable: better project selection, faster iteration, sharper oversight, and stronger governance. In an environment where public servants are already stretched thin, we owe them every tool we can offer. Not just to meet mandates, but to meet the moment. If AI is poised to transform the government, people must come first in that transformation.
This is a rare window of opportunity. We have the eyes of the executive, growing public interest, and unprecedented momentum. Agencies that act now to provide relevant, contextual, role-specific training will not just keep up with compliance. They will lead with confidence. For public policy leaders, this is your chance to shape implementation from the inside. Advocate for AI training to be embedded in every agency strategy, funded as a priority, and tracked with the same rigor as deployment milestones.
The infrastructure for trustworthy AI does not begin with code. It begins with people. Let us build it accordingly.
