AI tools can generate a five-module curriculum outline in the time it takes to refill your coffee. Scenario dialogue, knowledge checks, voiceover scripts, the works—clean, polished, and ready to drop into a course shell.
But speed isn’t the same thing as learning.
The production side of instructional design has changed faster in the last two years than in the previous twenty, but the decisions that determine whether training truly changes behavior haven’t accelerated. They still depend on someone who understands the learner, the performance gap, and the difference between content that looks complete and content that works.
The Reality of Output and the Illusion of Outcome
Here’s a familiar scenario:
A stakeholder needs training. Someone drops source material into an AI tool. Seconds later, they have a clean module structure and a set of knowledge-check questions. The language looks professional. The content looks accurate. It looks finished. So the course goes live.
Six months later, the behavior the training was meant to change hasn’t changed.
Part of the problem is structural. Large language models generate text by predicting what language should come next, not by verifying truth or applicability. That means they can produce content that is fluent, coherent, and very wrong. Researchers refer to this phenomenon as hallucination, where models produce plausible but incorrect information. Even with modern systems, fabricated or misleading information still appears in a measurable share of complex reasoning tasks.
Accuracy isn’t the only risk. AI can also reproduce the shape of good instructional design without the thinking behind it: assessments that test recall instead of application, scenarios that sound realistic but miss the real decision pressure learners face, and modules that feel complete while leaving the original performance gap untouched.
Closing that gap still requires human judgment. It requires someone who can look at the scenario and ask: Is this what the learner will actually face on the job? It requires someone who can recognize when a knowledge check measures whether the learner can do the work differently, as opposed to whether or not they just read the slide.
That skill lives in people: the expert in the loop.
From Creator to Curator
AI isn’t changing the need for expertise, but rather where that expertise shows up.
For decades, instructional designers were primarily content creators—writing scripts, building slides, and producing assets. Now, AI can generate those drafts, variations, images, and translations almost instantly. That frees up time that used to go into production and redirects it toward work that produces outcomes.
What it produces, however, is closer to raw material than finished learning. The designer’s job then becomes selecting, refining, and reshaping those outputs into something that aligns with how people actually learn: sequencing ideas, building retrieval opportunities, embedding practice, and ensuring the experience maps to real work.
In other words, the work moves upstream, from producing content to shaping learning systems. This “creator to curator” shift was a recurring theme in industry discussions last year, particularly as organizations realized that the value of AI outputs depends almost entirely on the expertise guiding them.
The tool generates; the expert decides what matters.
The Pipeline Problem
Here’s an uncomfortable question behind the excitement: if early-career roles become primarily AI-assisted production, where does the next generation of that expertise come from?
The people using AI most effectively right now are the ones who already know what “good” looks like. A seasoned instructional designer can spot a weak scenario immediately: the stakes feel artificial, the distractors are obvious, or maybe the assessment misses the real decision point. That expertise is a product of deliberate practice—doing the work, getting feedback, getting it wrong, and doing it again.
Expertise develops slowly.
Research on skill acquisition consistently shows that expert judgment develops through direct engagement with complex problems and iterative feedback, not simply through reviewing outputs. If those developmental experiences shrink, organizations may eventually face a shortage of people qualified to oversee the very systems they depend on.
The developmental phase doesn’t disappear when you accelerate past it. It defers. And when it surfaces, it tends to surface in the work.
Why Life Sciences Raises the Stakes
In most industries, a training program that doesn’t transfer knowledge is a disappointment and a sunk cost. In life sciences, the consequences are often more tangible.
Professionals in this field aren’t rehearsing abstract frameworks. They’re preparing for conversations about clinical evidence, patient access, treatment pathways, and payer dynamics. The quality of their preparation shows up quickly in real interactions. That’s why the expert in the loop matters so much.
At CMR Institute, that loop involves more than a single reviewer; it’s a system: physician-founded expertise, subject-matter specialists validating clinical accuracy, and instructional designers translating complex science into credible learning experiences.
As AI expands the range of possible content—and the variability of its quality—that kind of expert-driven pipeline becomes more important, not less.
What the Tools Are Working With
The principles that govern how adults learn predate every learning technology and will outlast this one. Motivation and autonomy. Relevance to real work. Retrieval practice. Application in context. Feedback loops that support transfer. These are the mechanisms through which training becomes capability, and they operate the same way regardless of how fast the content was produced.
The organizations finding their footing in this moment are the ones where experienced practitioners stay in the decision-making seat: defining the real problem before selecting a solution, designing the learning architecture before generating the content, and verifying that learners left with what they came to get.
AI may write the first draft, but the expertise that turns information into capability is still very much human.
Shelly Cline
Shelly Cline is a Senior eLearning Developer & Instructional Designer at CMR Institute, with over a decade of experience creating engaging, learner-centered training programs. She specializes in translating complex concepts into accessible, impactful learning experiences and has led large-scale initiatives across industries. Shelly is particularly interested in the role of AI in learning and development, exploring how emerging technologies can enhance engagement and design efficiency. She is committed to continuous improvement and building inclusive learning experiences.
