DOL AI Literacy Framework: What It Means for Power Users and Trainers

In February 2026, the Department of Labor published Training and Employment Notice 07-25 , a framework for AI literacy across the American workforce. It’s voluntary, not regulatory. No mandates, no compliance requirements. But it’s the first time a federal agency has laid out a structured definition of what “AI literate” actually means.

If you build AI training programs, teach AI skills, or help others adopt AI tools, this framework matters. Not because you have to follow it, but because it gives you a credible skeleton to build on. Here’s the full breakdown.

The Five Content Areas

The framework defines five areas that an AI-literate person should understand. These aren’t skill levels; they’re knowledge domains. Think of them as the columns in a curriculum matrix.

1. Understanding How AI Works

The fundamentals. How models are trained, what tokens are, why outputs vary, what a context window does. Not deep ML theory, but enough to understand why AI behaves the way it does.

2. Exploring AI Applications

Where AI fits across industries and job roles. Not just “ChatGPT can write emails” but understanding the landscape: coding assistants, image generation, data analysis, customer service automation, document processing. The framework wants people to see the breadth of what’s possible.

For power users, this is the easiest area to teach because you live it. The challenge is scoping it for your audience. A marketing team doesn’t need to know about code review agents. A dev team doesn’t need to know about social media content generation.

3. Effective Prompting Techniques

The framework explicitly calls out prompt engineering as a core literacy area. This validates what the AI power user community has known for years: the quality of your input determines the quality of your output. Clear instructions, structured context, iterative refinement.

I'm training a group of [role/department] on AI prompting basics. Create a 5-minute exercise that teaches three core prompting principles:

1. Be specific (vague vs. specific prompt comparison)
2. Give context (same prompt with and without context)
3. Iterate (show how refining a prompt improves the output)

For each principle, include a before/after prompt pair they can try themselves. Use examples relevant to [their industry/role].

4. Evaluating AI Outputs

Critical thinking applied to AI responses. Is the output accurate? Is it complete? Does it contain fabricated information? The framework positions this as a distinct skill, not just “check the answer.” It includes understanding model limitations, recognizing when confidence doesn’t equal accuracy, and knowing when to verify against external sources.

This is where most beginner AI users fail. They either trust everything the model says or distrust everything and don’t use it at all. A common example: a model confidently cites a research paper that doesn’t exist, and the user includes it in a report. Teaching people to verify citations, cross-check numbers, and recognize when a model is filling gaps with plausible fiction is the core of this area. The framework pushes toward a middle ground: informed skepticism.

5. Managing AI Responsibly

Security, privacy, ethical use, and organizational policies. What data should and shouldn’t go into AI tools. How to handle sensitive information. Understanding bias in model outputs. Compliance with organizational and regulatory requirements.

The Seven Delivery Principles

The content areas define WHAT to teach. The delivery principles define HOW. This is where the framework gets interesting for curriculum designers.

1. Enable Experiential Learning

Don’t lecture about AI. Let people use it. The framework explicitly calls for hands-on, interactive training where learners work with actual AI tools rather than watching slides about them.

This validates what works: workshops where participants bring their own tasks, run prompts live, and see results in real time. If your training involves more than 20% lecture, you’re doing it wrong according to this principle.

2. Build Complementary Human Skills

AI literacy isn’t just technical skills. The framework calls out critical thinking, communication, problem decomposition, and judgment as essential complements. The human skills that make AI useful rather than just fast.

3. Create Pathways for Continued Learning

AI tools change monthly. Training can’t be a one-time event. The framework calls for structured progression: foundational modules that lead to intermediate and advanced tracks, with ongoing learning baked into the system. In practice, this looks like a “Level 1: Use AI for your daily tasks” module followed by “Level 2: Build custom workflows” three months later, with a monthly 30-minute “what’s new” session to cover tool updates and emerging techniques.

4. Design for Agility

Don’t build a curriculum around a specific tool version. Build it around principles and patterns that transfer across tools. This is the framework’s way of saying “don’t make your entire program dependent on ChatGPT 4o’s interface, because it’ll change in three months.”

I'm building an AI literacy program for [audience]. Help me design the curriculum around transferable principles rather than specific tools.

For each module, provide:
1. The core concept (tool-agnostic)
2. A hands-on exercise using [Tool A]
3. The same exercise adapted for [Tool B]
4. The transferable skill the learner should take away

This way, when tools change, I only update the exercises, not the entire curriculum.

5. Embed Learning in Context

Train people using their actual work, not generic examples. An accountant should learn AI with accounting tasks. A project manager should learn with project management scenarios. Context-embedded training has higher retention and immediate applicability.

6. Address Prerequisites to AI Literacy

Not everyone starts at the same baseline. Some learners need basic computer skills before they can engage with AI tools. The framework acknowledges this gap and calls for prerequisite assessment and support. Don’t assume digital fluency.

7. Prepare Enabling Roles

This principle targets the people who support AI adoption: IT staff, managers, trainers, HR. They need their own training track focused on infrastructure, policy, change management, and supporting learners rather than being learners themselves.

Gap Analysis: What the Framework Gets Right and What It Misses

Gets right:

• Experiential learning over lecture (Principle 1)
• Human skills as a prerequisite, not an afterthought (Principle 2)
• Tool-agnostic design (Principle 4)
• Responsible use as a core area, not an appendix (Content Area 5)

Misses or underserves:

• Agent workflows and automation. The framework is oriented toward individual tool use (prompting, evaluating outputs). It doesn’t address multi-agent systems, workflow automation, or the orchestration layer that power users operate in.
• Cost literacy. No mention of understanding API pricing, token economics, or the cost/quality tradeoffs that matter in production use.
• Evaluation beyond accuracy. Content Area 4 focuses on output evaluation but doesn’t cover evaluating models themselves (benchmarks, comparative testing, provider selection).
• Building vs. using. The framework treats AI as a tool to use. It doesn’t address building custom assistants, creating instruction sets, or designing AI-powered systems. For power users, the building side is where the real leverage lives.

Building a Curriculum From This

If you’re actually designing an AI literacy program, here’s a practical mapping:

Using the DOL AI Literacy Framework (TEN 07-25) as the foundation, create a curriculum outline for [audience type] with [X] sessions.

Content areas to cover:
1. Understanding how AI works
2. AI applications in [their industry]
3. Prompting techniques
4. Evaluating AI outputs
5. Responsible AI use

For each session, include:
- Learning objective (one sentence)
- Hands-on exercise (using [preferred tool])
- Assessment method (how to verify the learner got it)

Apply the experiential learning principle: at least 70% of each session should be hands-on, not lecture.

The full framework document is available at DOL TEN 07-25 . It’s worth reading the source if you’re building anything formal around AI training.

Why This Matters for the AF1 Audience

You probably don’t need an AI literacy program for yourself. But you’re increasingly the person others come to for AI guidance. Your team, your company, your clients, your community. This framework gives you a structured way to share what you know instead of ad-hoc “let me show you something cool” sessions.

Use it as the skeleton. Fill it with your experience. Teach what you actually do, not what a framework says in the abstract.