The natural American reaction to something novel, especially something man-made and powerful, is for earnest (usually government) technocrats to Gerberize it into something mortals can digest and overseers can feel they’ve controlled. It happened with penicillin, atomic energy, and personal computers.
Artificial intelligence is getting the same treatment, as it should, just faster. Yet the frameworks are coming as fast as the model upgrades as we move past the 18-month mark for mainstream generative AI. The U.S. Department of Labor’s AI Literacy Framework, released last week, looks and reads much like the others from their sister agencies and nonprofits. It features government-grade prose, block diagrams, and familiar chevrons and icons. But there are nuggets worth extracting.
February has already been a month of AI mania. Substantive model upgrades. Communities of rebel agents. Essays with millions of views warning that everyday people have no idea what’s about to hit them. The tone is breathless, urgent, slightly apocalyptic. And now here comes the ink-stained Labor Department, quietly releasing a framework that includes “understanding file systems” as a prerequisite and “verify the output” as a core competency. The gap between the public discourse and the guidance is vast. But the truth is, the people most affected by AI in the workplace aren’t paying attention to either. They’re not on X. They’re not reading the essays. And they’re definitely not reading federal guidance letters.
Maybe we can take this framework as a small sign the conversation is maturing. One small harbinger of a move from existential handwringing to practical guidance. Less “AI will take your job” and more “here’s how to use it at yours.” We won’t really know the workplace impact of AI until it becomes a little boring. Workaday. Like the PC. Like the Internet.
The DOL guidance checks the expected boxes: understand how AI works, use it to build competency, align with guardrails. The framework also includes seven delivery principles: guidance for how to teach this, not just what to teach. The guidance is aimed at state workforce agencies, community colleges, American Job Centers, apprenticeship programs, and their education counterparts. It’s intended for the people who design and deliver training. DOL isn’t mandating anything—they’re asking these organizations to expand AI literacy efforts and use the framework as a reference. A nudge, not a rule.
But the framework features four areas that get outside the usual box in useful ways:
1) They named “enabling roles.”
Most frameworks focus on the learner. DOL explicitly calls out the people who support learners: managers, trainers, career counselors, mentors. This matters because a worker can complete an AI literacy course and walk back into a workplace where their supervisor is skeptical and their HR team has no guidance. The chain breaks. AI adoption isn’t just an individual skill—it’s an environment. Someone has to prepare the soil, not just plant the seed.
2) They acknowledged basic digital literacy as a prerequisite.
The framework acknowledges something most AI guidance skips: not everyone is ready for AI literacy because not everyone has basic digital literacy. File systems. Browsers versus apps. Uploading and downloading. You aren’t gaining much workplace value from GenAI without them. DOL suggests programs should assess baseline readiness before diving into AI. It’s an admission that the on-ramp isn’t the same for everyone.
3) They name “Direct AI” as a core competency.
Directing means learning to prompt well—providing context, iterating, being specific. But it also means understanding that the best results come when you bring your own data to the conversation. Whether that’s pasting a report into a chat, building a NotebookLM notebook, or using retrieval-augmented generation in an enterprise setting, the principle is the same: the AI gets smarter when you feed it what you know.
4) They go beyond “human in the loop.”
When DOL says “evaluate AI outputs,” that means not taking what the AI gives you at face value. The agency pairs this with a delivery principle called “build complementary human skills.” They call out critical thinking, creativity, communication, and domain expertise. These aren’t soft skills for a commencement speech. They’re the reason the output is any good. The AI drafts; you decide. That division of labor is the whole game.
What’s Missing and What Comes Next: No tools named, no training providers listed, no credentials referenced. Someone reads this framework, gets motivated, and has nowhere to go on Monday morning. That’s not a criticism; it’s the nature of federal guidance. The DOL can’t pick winners.
So what are they actually doing here? The best you can hope for is a common vocabulary. A shared set of concepts that training providers, workforce boards, employers, and educators can point to when building programs. That’s not nothing. It gives the people who actually have to choose tools, design training, and set guardrails something to anchor to.
And yes: we’ve intentionally not addressed agents and vibecoding here. Not because they’re unimportant. Because you can’t vibecode if you can’t harness the chatbot. That’s where most people are. That’s where the DOL is.
I work across education, nonprofit, and corporate settings. These discussions are becoming more urgent. For example, I chair the education committee for a regional AI task force convened by the Mayor of Huntsville, Alabama. While there’s plenty of excitement about the emergence of powerful agentic AI, our mandate is practical and focused on real-world impacts. Our community will welcome the tone of this first step by the Labor Department.
The new federal framework won’t tell you which AI to use or how to teach it. But it does give everyone in the room the same words and a realistic place to start. Sometimes that’s all you need to get going.
Randy Sparkman is an AI strategy advisor and enterprise technology expert, who for 25 years was associate CIO at the NASA Marshall Space Flight Center.