Generative AI is changing more than the tools we use at work. It is changing the capabilities people need to work effectively. As organizations invest in AI training and adoption, much of the attention has focused on tools, prompting, and technical proficiency. Those skills matter, but they are only part of the equation. The bigger opportunity for Learning & Development may be helping people develop the judgment, critical thinking, and discernment required to use AI well.

So perhaps the question isn’t simply, “How do we teach people to use AI?” It is, “What do people need to become better at because AI exists?”

Organizations are moving quickly to build AI capability. Employees are being introduced to tools like ChatGPT and Copilot, learning how to write better prompts, attending workshops, and experimenting with ways to incorporate AI into their daily work. All of these steps are important. But the more I think about AI adoption through the Learning & Development lens, the more I wonder whether we’re focusing too much on the tools and not enough on the capabilities people need to use them well.

Teaching someone how to write a prompt is relatively straightforward. Teaching someone to recognize when AI should be used, evaluate what it produces, question an answer that sounds convincing, and decide when human expertise needs to take over is much harder. This distinction matters because AI isn’t simply changing how people complete tasks. It’s changing what people need to be good at.

From AI proficiency to AI judgment

The first wave of workplace AI learning understandably focused on proficiency. People needed to understand what generative AI was, what the available tools could do, and how to interact with them effectively. We’re now moving beyond that stage.

As AI becomes embedded in everyday work, employees need more than technical proficiency. They need the judgment to determine where AI adds value and where it introduces risk. They need enough understanding of their work to recognize when an output is incomplete, misleading, inappropriate, or simply wrong. Consider a manager preparing for a difficult employee conversation. AI could help organize thoughts, identify questions to ask, or suggest clearer ways to communicate. But the manager still needs to understand the employee, the context, the situation’s history, and the conversation’s potential impact. Or consider someone using AI to analyze information and recommend a course of action. AI may produce an impressive analysis in seconds, but someone still needs to ask whether it considered the right or even correct information, whether it missed important context, and whether the recommendation actually makes sense.

The technology can accelerate the work. It cannot assume responsibility for the judgment behind it.

AI is revealing an “old” L&D challenge

In many ways, this isn’t a new learning problem. Learning & Development has always faced the challenge of moving beyond knowledge transfer. We can teach a framework, demonstrate a system, explain a process, and provide resources. The harder part is helping someone apply that knowledge when the situation is ambiguous, and there isn’t a perfect answer.

AI makes that challenge much more visible.

When information is instantly available and a reasonably polished answer can be generated in seconds, knowing information becomes less differentiating. Knowing what to do with it matters more. That shifts the learning conversation from “Can you use the tool?” to “Can you use the tool well?” These are very different questions. Using AI well requires critical thinking, context, curiosity, discernment, verification, and an understanding of the work itself. It also requires the confidence to challenge an AI-generated answer rather than assume a polished response is correct. In other words, the next generation of AI learning shouldn’t just be tool training. It should be judgment training.

What would that look like in practice?

It could mean starting AI learning somewhere other than the prompt box. Before teaching someone how to construct a sophisticated prompt, we could first ask them to think about the work they are trying to accomplish. What problem are they actually solving? Which parts of that work could AI accelerate? Which parts depend on experience, relationships, organizational context, creativity, or judgment?

Then we can teach prompting within that context.

Instead of simply showing employees how to get a better response from AI, we can teach them how to evaluate the response they receive. What assumptions did the AI make? What information might be missing? What needs to be verified? What would happen if the answer were wrong? Is there confidential or sensitive information that shouldn’t be entered in the first place? This approach also changes how we measure AI learning. Completing a course or demonstrating the ability to write a prompt tells us very little about whether someone can use AI effectively in their actual job. A better measure may be whether employees can identify meaningful use cases, incorporate AI into appropriate workflows, evaluate outputs critically, and improve the quality or efficiency of their work without surrendering the judgment that made the work valuable in the first place.

A different opportunity for L&D

This is where I think Learning & Development has an important role. The opportunity isn’t simply to become the department that teaches everyone how to use the latest AI tools. Tools will keep changing, and employees will increasingly learn many of their features on their own. The larger opportunity is to help organizations build the human capabilities that make those tools useful.

This means connecting AI learning to actual work rather than treating it as a separate technology initiative. It means working with employees and managers to identify meaningful use cases, creating opportunities to practice with real scenarios, and developing the critical thinking required to evaluate results. It also means recognizing that adoption isn’t the same as capability. An employee can use AI every day and still use it poorly. Another employee may use it selectively but demonstrate excellent judgment about when it adds value.

The goal isn’t to use AI more. It’s to use AI to do better work.

The question I think we should be asking

AI will keep getting easier to use. Interfaces will improve, tools will become more integrated into our workflows, and many of the prompting techniques we’re teaching today may eventually become unnecessary. Human judgment won’t. That’s why I’m increasingly less interested in asking, “How do we teach people to use AI?” I’m much more interested in asking: What do people need to become better at because AI exists?

For Learning & Development, I think the answer to that question may shape some of the most important work we do next.