Subject-matter experts know the content. The course is the hard part.
I have spent years training subject-matter experts to build e-learning. The pattern is consistent enough that I can almost predict the week things break down. Week one: enthusiasm. They arrive with knowledge nobody else in their organisation has and they want to share it. Week two: a polished outline that sounds, on the surface, like the structure of a course. Week three: stuck. And the place they are stuck is never the thing they know best.
That gap — between deep subject expertise and a course that actually teaches it — is an underrated problem in workplace learning. It is also the problem that traditional authoring tools, for all their features, were never really designed to solve.
The curse of knowing too much
An expert has spent ten years building a model of their field. Everything inside that model feels obvious. The foundational concepts — the ones a learner needs in order to understand anything else — have long since disappeared into the background. Asking the expert to surface them, in the right order, with the right examples, is asking them to do something they have actively trained themselves not to do.
The result is the first lesson of nearly every first-draft course I see: the table of contents of an advanced textbook. Lesson one assumes the reader already knows what the course is for. The actually-difficult thing — the thing the learner will fail at — is buried two lessons in, because the SME forgot it was difficult.
Subject-matter experts are not bad at this. They are being asked to do something they were never trained for, which is to think like a novice.
Content dump, not instruction
A second pattern: the SME writes what they know, in the order they think about it, and calls it a lesson. Long paragraphs. Bullet points stacked on bullet points. Headings that summarise their mental model rather than the learner's path. No retrieval questions. No activities. No moments where the learner is asked to do anything except keep reading.
This is what happens when the only model of "instruction" the SME has is a lecture or a textbook. They translate their knowledge into the medium they know. The medium is wrong, but they might not even know how the right medium looks like, and the tool they are using is not pointing them toward it.
The compression problem
Everything feels important. The SME has spent years deciding what matters in their field. Asking them to cut half of it feels like a betrayal of the topic. So the course bloats. A subject that needed twelve focused minutes turns into a 90-minute marathon that learners will leave early.
I have spent many hours talking to experts, pointing at a paragraph, asking what would happen if a learner never read this. The honest answer is almost always nothing. The expert knows it. Cutting it still feels wrong.
The tool is fighting them
The other thing that consistently surprises people who have not trained SMEs is how much of their energy goes into fighting the authoring tool, not crafting the content. Articulate Storyline, Adobe Captivate, even Rise have steep learning curves. On top attractive e-learning needs professional media, including images and video which is another field of experties. An SME who is already short on time, who is doing this on top of a day job, who has never built a course before, ends up spending the first month of the project learning the interface. By the time they understand the tool, they have lost momentum on the content.
This is not a criticism of those tools. They are genuinely capable, and instructional designers who use them every day produce excellent work. But "capable tool that takes weeks to learn" is a different proposition from "tool an expert can sit down with and make something useful."
What helps, in practice
Watching the same patterns has shaped how LearnBuilder is built. The goal is not to turn SMEs into instructional designers. It is to make the gap between "I know this" and "I can teach this well" small enough that domain expertise actually reaches learners.
A few things matter specifically for the SME case.
Start from a draft, not a blank page. The hardest moment in course-building is the empty lesson. AI course generation gives the SME a complete first draft, structured around a recognised instructional approach and tuned to a stated audience experience level and learning context. The SME starts from where they are strongest, which is editing. They reject the wrong example, sharpen the awkward phrasing, add the nuance the model missed, cut the lesson that did not need to be there or add where it is needed. Their expertise comes through as a series of confident corrections rather than as a struggle to invent structure from nothing.
Pedagogy as a default, not as a prerequisite. Retrieval questions, knowledge checks, and scenario activities are generated alongside the explanatory content, interleaved at sensible intervals, scaled to the chosen instructional approach. The SME does not need to know what a desirable difficulty is. They do need to read the questions and decide whether they actually test the thing the lesson taught — and that is a judgement an SME can absolutely make from their expertise. Writing a good multiple-choice distractor from scratch is a separate skill. We should not be requiring it of them.
A block editor instead of a slide editor. The interface is built around what a lesson is, not what a slide is. A block of explanatory text. A block with an embedded question. A block with a branching scenario. A block with a video. A block with a transcript and a translation. The SME thinks about what the learner does next, not about which panel of which inspector controls which property of which timeline of which animation. The cognitive load of the tool itself stays low enough that the SME can spend their thinking on the content, where it belongs.
Forced structure, gently. The four instructional-design approaches give the SME a structural skeleton to react to rather than design. An expert who could not, on day one, design a story-driven course from scratch can absolutely tell you whether a generated story-driven course about their topic is any good, what is missing, and what to cut. The structure becomes a target the expert can refine, not a void they have to fill.
Quiet defaults that protect the learner. Brand consistency is configured once at the account level and applied everywhere — the SME is not picking fonts. Accessibility checks flag missing alt text and low-contrast colour pairs before publish. Translation across sixty languages, with AI-translated video subtitles, is one action. None of these are things the SME would have thought to do. All of them are things a learner would have noticed if they had been missed.
What this does not fix
To stay honest about the limits — the parts no tool can do for an SME.
It cannot replace the editorial judgement. If the SME does not know what the learner is supposed to be able to do at the end of the course, no amount of generation produces a course that gets the learner there. The learning objective is upstream of the tool.
It cannot make a course out of bad source material. If the document the SME uploads is poorly structured, full of jargon, and missing its conceptual through-line, the generated course will reflect that. The AI is a faithful summariser, not a magic improver.
It cannot tell the SME that their three favourite examples are the wrong three for a novice audience. The expert still has to do the work of remembering what it was like to not know — and that is the work that sits closest to the curse-of-knowledge problem this whole post started with.
And it cannot produce a course without the SME's time and attention. Total elapsed effort comes down materially — generation plus refinement is far faster than authoring from scratch — but it does not approach zero. The SME still has to sit down, read the draft, make the judgement calls, and stand behind the result. There is no version of this in which a tool produces an excellent course while the expert is on holiday.
The goal
The interesting goal here is not "anyone can build any course." It is "an expert in the subject can build a course that teaches it well, in roughly the time their actual job leaves them, without spending three months learning a tool first."
The biggest unlock — the one I keep coming back to after years of running these training sessions — is that editing is a different cognitive activity from authoring. Editing is what experts are already good at. Their judgement, their taste, their sense of what is wrong with a paragraph — these are sharper than they realise, because they have been applying them in their day job for years. The tool's job is to put the SME in front of a draft they can apply that judgement to, on a topic only they know, in an interface that does not steal their attention.
Everything else — the block editor, the instructional defaults, the accessibility checks, the translation, the brand consistency — is in service of that one move. Get the expert out of the role of author, into the role they were already going to be brilliant in, which is editor of their own field.
This is also where the related argument lives: the case for hybrid AI-assisted and hand-built workflows is essentially the case for letting the SME pick how much they want to author and how much they want to edit, on a course-by-course basis. For most SMEs in most contexts, mostly-edit is the right setting. For a few high-craft courses, mostly-build. The dial is the point.
If you train subject-matter experts to build e-learning — or you are one yourself, looking at a long document and a Friday deadline — the free trial is the place to start.