Perspective
AI Can Build the Course. That Was Never the Hardest Part.
AI is rapidly collapsing the distance between an idea and a finished learning artifact. But faster production does not resolve the harder question: Are we building the right solution to the right problem?
We have reached the point where AI can produce nearly every visible component of an eLearning course.
Give it a topic and it can develop the outline. Give it source material and it can write the storyboard, generate the visuals, create the narration, build the interactions, write the assessment, and package everything for an LMS.
What once required several specialized tools and weeks of production can increasingly be accomplished through a collection of prompts, AI platforms, and automated workflows.
That is a remarkable technical achievement.
It is also not the breakthrough learning and development needs most.
We are shortening the wrong distance
Much of the conversation about AI in our profession is focused on how quickly we can move from “I know what I want to teach” to “The course is live.”
The space between those two statements is becoming dramatically smaller. Storyboards that once took days can be produced in minutes. Interactions that required development expertise can be generated from a description. Voiceovers no longer require a recording session. Even the technical packaging and deployment of a course can be automated.
From a production perspective, this is progress.
But the most important distance in learning design is not the distance between having content and publishing a course.
It is the distance between “We have a problem” and “We understand what is actually causing it.”
The hardest and most consequential work happens before the first prompt asks AI to create anything.
The workflow begins with an untested assumption
“I know what I want to teach” sounds like a reasonable starting point. In many organizations, it is the starting point.
A stakeholder identifies information people need to know. A learning request is submitted. The instructional designer organizes the content, writes objectives, creates activities, builds an assessment, and publishes the final course.
The process appears to work because it produces something tangible.
But starting with what we want to teach means several important decisions may have already been made without enough examination. We have assumed that teaching is necessary. We have assumed that a lack of knowledge or skill is causing the problem. We have assumed that a course is the right intervention. We have assumed that the requested content is relevant to the work. We have assumed that publishing something in the LMS will help people perform differently.
Those assumptions existed long before generative AI. AI simply allows us to turn them into polished learning artifacts much faster.
The missing question
Before asking AI to build the course, we should be asking whether the performance problem requires a course at all.
The artifact is not the outcome
Learning and development has spent decades describing its work through deliverables.
We build courses, videos, job aids, simulations, workshops, and assessments. We track enrollment, completion, satisfaction, and scores. Our platforms are organized around content, and our processes are often designed to move that content toward publication.
As a result, it is easy to mistake the production of a learning artifact for the solution of a performance problem.
A course can be beautifully designed and still be unnecessary. A branching scenario can be interactive and still practice the wrong decisions. A knowledge check can produce perfect scores and still tell us nothing about performance on the job. A SCORM package can function exactly as intended and still fail to change a single meaningful behavior.
None of this makes the artifacts inherently bad. Their value simply depends on the quality of the decisions that shaped them.
Before we begin building, we need to understand what people are expected to do, what they are doing today, and what is preventing them from closing that gap.
Is the problem really a lack of knowledge? Or are employees working within a broken process? Are expectations unclear? Are the necessary tools difficult to use? Do incentives reward the wrong behavior? Are managers reinforcing a different priority? Is critical information unavailable at the moment of need?
A course cannot solve every performance barrier.
Generating the course faster does not change that.
Effortless production can amplify weak decisions
AI is changing the economics of content production. Storyboards, scripts, illustrations, voiceovers, interactions, assessments, and working code can now be generated in a fraction of the time.
That creates enormous opportunities for learning teams. It also creates a new risk.
When courses become easier to produce, organizations may create more of them without becoming more selective about which ones should exist. Every policy update can become a module. Every stakeholder request can become a course. Every performance concern can be routed toward the LMS because the cost of saying yes appears to have fallen.
But the true cost of an ineffective course was never limited to the hours required to build it.
There is the learner’s time. There is the operational time taken away from the work. There is the attention consumed by content that may not be relevant. There is the maintenance burden created by another asset. There is the false confidence that comes from high completion rates. There is the opportunity cost of not addressing the real barrier.
AI may reduce the cost of production, but it does not eliminate the consequences of solving the wrong problem.
It may actually make those consequences easier to multiply.
The opportunity is bigger than automated course production
The most exciting role for AI in learning design may not be generating the final artifact.
It may be helping us improve the decisions that come before it.
AI should help us interrogate a request before we fulfill it. It should help us examine the evidence behind a perceived performance gap. It should help us distinguish between a knowledge problem, a skill problem, an environmental problem, and a motivation problem. It should help us connect business needs to observable behaviors and those behaviors to meaningful practice.
It should challenge objectives that cannot be observed. It should identify assessments that measure the wrong thing. It should reveal when a scenario does not resemble the pressure, ambiguity, or consequences of the actual work. It should help us determine whether learners need instruction, practice, feedback, performance support, workflow changes, or some combination of them.
Most importantly, it should help us recognize when a course is not needed.
That version of AI does not merely make instructional designers faster. It helps them become more rigorous.
Better tools require stronger designers
As AI takes on more production work, the role of the learning designer becomes more strategic, not less important.
Designers will need to become better problem framers. They will need to recognize when a learning request is built on an assumption. They will need to connect organizational goals with specific behaviors and credible evidence. They will need to design practice that reflects real decisions rather than decorate content with interactions.
They will also need to think beyond the boundaries of a course and consider the complete system surrounding performance.
If we continue to position instructional design primarily as the ability to convert content into courses, AI will understandably appear to replace much of that work.
If we define learning design as the practice of diagnosing performance needs, designing meaningful practice, supporting transfer, and measuring change, then AI becomes something different.
It becomes a powerful production partner inside a much larger design discipline.
AI can draft the storyboard. It can generate the interaction. It can write the code. It can create the voiceover. It can package the course.
But it cannot rescue a solution built around the wrong problem.
We should raise the benchmark
A functioning course generated through an AI workflow is an impressive demonstration of where the technology is heading.
We should learn from it. We should experiment with it. We should consider how it might make our work faster and more accessible.
But “It’s live in the LMS” cannot remain our definition of success.
The future of learning design should not be measured by how quickly we can move from content to course. It should be measured by the quality of the decisions we make before anything is built.
Did we identify the real performance problem? Did we determine whether learning was necessary? Did we define the behavior that needed to change? Did we create meaningful opportunities to practice? Did we address the conditions surrounding the work? Did we provide support at the moment of need? Did anything improve after the experience?
AI can help us answer those questions, but only if we ask it to do more than build.
The future of learning design will not be determined by who can generate the most courses or who can move content into an LMS the fastest. It will be determined by who can use increasingly powerful tools to make better decisions about what people actually need.
AI can build the course. That was never the hardest part.
Learning Rewired Lab
Use AI to improve the thinking, not just accelerate the build.
The Lab is a global learning community for people solving real workplace learning problems. Explore perspectives, practical tools, and systems-minded resources built to help learning professionals make stronger decisions before another course gets created.
Challenge the request, examine the system, and design for performance before production begins.
- A global learning community
- Real-world design conversations
- Product-minded learning tools
- AI-assisted design workflows
- Practical workplace application