AI Is a Lever, Not the Machine: Advocating for the Profession While the Tools Get the Credit

I want to tell you how this post came to exist, because the journey is the point.

Like most people in this field, my first encounter with generative AI was not frustration. It was awe, threaded through with more than a little fear. I watched AI generate a complete, coherent course outline in seconds and felt two things at once: this is remarkable, and what does this mean for me?

Then I started using the tools for actual work. Not demos. Real projects. And what I found was more clarifying than I expected. AI is very good at producing things that look like training. Structured content. Plausible learning objectives. Cleanly formatted assessments. What it was not producing was effective training. Training built on a real understanding of the performance gap, its causes, and what the learner's actual world looks like rather than the ideal world leadership imagines exists.

I started calling these outputs "training-shaped things." Recognizable from the outside. Hollow at the core.

That realization shifted my concern. The question I had been carrying (will AI replace people like me?) transformed into something more troubling: will people assume it can? Those are not the same concern. The first is about capability. The second is about perception, and perception is harder to correct.

Here is what makes that concern real and not abstract. Most of us work for organizations that exist for reasons that have nothing to do with learning and development. They make loans, build aircraft, run hospitals, manufacture goods, provide software. Their leaders are experts in those domains, not ours. Many of them have a partial or incomplete understanding of what qualified L&D professionals actually do and why it matters. Into that gap, AI has arrived looking capable, fast, and cheap. Some organizations have concluded (incorrectly but understandably) that it can do what we do. Some practitioners have already lost their jobs to that conclusion. That is the reality this post is written into, and it is why advocacy is not a professional nicety but a professional necessity.

The Tool Is Not the Field

Content creation is not what this field is for.

It has always been executable by people without L&D credentials. Technical writers, subject matter experts, communications teams, and interns have created training content throughout the history of this profession. What they produce is training-shaped things: content that covers the material, satisfies a compliance checkbox, and produces no meaningful change in performance.

What qualified L&D professionals bring is not faster or better content creation. It is the entire system of thinking and practice that determines whether content is the right intervention at all, what it should accomplish, how it should be designed to actually transfer to performance, and how to know whether it worked.

Human Performance Technology has been making this argument since the late 1960s, when the field emerged specifically because training was consistently failing to produce the performance improvements organizations needed. The foundational insight then is the foundational insight now: underperformance is rarely a training problem. It is a systems problem. And systems problems require diagnosis before any intervention can be designed.

AI can generate a training program in response to almost any prompt. It has no mechanism for asking whether training is warranted. That question requires expertise, organizational access, and the willingness to tell a stakeholder that the solution they want is not the solution they need.

Where AI Actually Belongs

This is not an anti-AI argument. Used well, AI is a genuine accelerant for production work: drafting content, formatting materials, generating assessment variations, building first outlines. These are real contributions, and practitioners who use AI for this work are not compromising their professional standards. They are working more efficiently.

But production is one phase of a much larger body of work. Here is where human expertise is required throughout the rest of it:

Before any solution is designed, someone has to understand what is actually happening in the organization. That means conversations with people who know what the work actually looks like versus what the process document says. AI can help draft the interview guide. It cannot conduct the interview, read between the lines, or recognize when a performance gap is actually a management problem.

Someone has to determine whether learning is even the right solution. Misaligned incentives, unclear expectations, missing tools, and inadequate feedback are all performance problems that a well-designed course will not fix. AI does not ask whether training is warranted. That judgment belongs to the practitioner.

Learning design requires knowing the difference between how a process is supposed to work and how it actually works. That knowledge lives in the people doing the job. AI can synthesize what is documented. It cannot surface what has never been written down.

Evaluation is not a survey. It is an interpretive act. Someone has to understand why learners responded the way they did, identify the gap between what participants say they learned and what they can actually do, and translate findings into specific redesign decisions. AI can help analyze data once it exists. It cannot conduct the conversation that reveals the training missed something critical.

And perhaps most invisibly: L&D professionals advocate within organizations. They build credibility that earns them access before a solution is commissioned. They push back on misframed requests. They protect the integrity of the discipline inside organizations that would prefer to skip the diagnosis and go straight to the content. None of that appears in a course file. All of it requires a human with standing.

The Advocacy Imperative

The field is defined by its questions, not its tools.

What is the performance gap? What is causing it? What intervention, at what level of the system, will close it? How will we know? These questions preexist any technology and will outlast any technology. They cannot be answered by a prompt. They require a practitioner who knows how to ask them, and who has the organizational access, diagnostic skill, and professional judgment to act on the answers.

AI is a lever. A real one, at a genuine leverage point in the information flow of learning work. Used well, it makes this field's work faster, broader in reach, and more consistent in execution.

But the machine that produces genuine performance improvement is the practitioner. The one who asks the diagnostic questions before reaching for any solution. Who reads the organizational system before designing the intervention. Who knows that a polished, AI-assisted training program built on the wrong diagnosis is still the wrong answer, and who has the expertise and the standing to say so.

The tools are getting the credit. It is time to claim what is ours.