AI in eLearning Development: Where It Helps and Where It Falls Short

That is where the catch is. Yes, it can generate all of that, but is the content it generates any good? Is the course actually designed to help someone learn? Does it speak in the client’s voice, or does it sound generic? AI is not a perfect tool, and those of us who use it often learn to recognize its weaknesses. Human knowledge, our knowledge and experience, is still what carries the work to its learning goals. In the end, AI is a bit like a self-driving car that still needs someone to tell it where to go.

In this post we share how eResponders uses AI in our workflow, what its role is and how human expertise directs the process, where AI shines and where its weaknesses show, and what to ask an eLearning partner about how they use it.

How eResponders uses AI in eLearning development

AI does real work for us at specific points in the pipeline, mainly on proposals and first-draft scripts. Give it the learning outcomes, an outline, and a theme, and it produces a solid starting point quickly. Everything after that stays with us. We revise for tone, fix errors, catch anything the AI invented, and make sure the content actually holds up.

A useful way to think about it is as a junior assistant. You can send it out on a task, and the work comes back needing feedback and another pass before it is ready. It saves time on the first mile, and the craft still lives in the revision.

On the production side, the same balance holds. Creative decisions stay on the human side, where we define the graphic line based on the client’s branding, target audience, and the tone of the course. Our software tools already include AI features that streamline the work, removing backgrounds or making quick adjustments to a character when we need to. We use AI to generate alternative text for the graphics in our Storyline courses, which supports accessibility. For visual assets, we still work mostly from stock libraries, partly because they are easy for us to edit, and partly because licensed stock comes with clear creative rights, while the licensing around AI-generated images is still unsettled. We reach for generative graphics only once a project already has a strong visual guide to hold everything together. We have used AI voice generation on a few projects to streamline the process, and we still bring in voice talent when the client requests it and the budget allows.

One part of the work never gets handed off: the client relationship. Listening to what a client actually needs, understanding what sits behind their ask, and reading the room in a meeting are not tasks you delegate to a tool. That understanding is what shapes everything the AI later helps draft.

Where AI helps in the work, and where it falls short

The clearest way to show the line is with a real example from each side.

Where it helps: when we are working within a theme and already know the learning outcome we want, AI is useful for building out an activity inside those constraints. For a course launching a standardized testing system, we used a space launch theme, and AI helped weave that narrative through the activity design in a way that would have taken much longer to draft alone. The direction was already set, so the tool could move fast inside it.

Where it falls short: A friend put it well. She asked AI for instructions to build a stool, followed them, and only at the end realized there was no plan for the legs. The result looked complete, but a core piece was just missing.

The same thing happens in learning design. AI output can look polished and structurally sound, and if you do not read it closely, you can miss that a key element is not there: an assessment tied to the objective, a scaffolding step, a check that learners can transfer the skill. It is not that the AI is wrong exactly. It is that it is confident, and confidence is not the same as complete. That gap is where the expert eye still matters, and it is the difference between a course that looks right and one that works.

That gap is not the only one worth knowing about. These are well-documented limitations that researchers and practitioners across the field have studied, and we see them in our own work too. These are the ones we keep the closest eye on:

• • • Hallucinations. AI invents data, quotes, and sources with complete fluency, and this is the hardest problem to catch, because the error looks just as polished as the fact.
    • Limited context. In long documents it loses the thread, forgets instructions given at the start, or contradicts what it said earlier.
    • A hidden cost of checking. The time it saves in production can be swallowed by review, especially for material that has to be exact.
    • Regression to the average. It leans toward the middle of what it has seen, generic aesthetics, familiar turns of phrase, conventional solutions. Originality is hard for it.
    • Specific weaknesses in images. Text inside images, hands and anatomy, keeping a single character consistent across scenes, and fine control over composition are all still unreliable.
    • Inherited bias. It reproduces and can amplify the biases in its training data, from representation to stereotypes to dominant cultural perspectives.
    • Unsettled rights and privacy. There are open questions about the copyright of training data and generated output, and a real risk of exposing sensitive information depending on the tool and its retention policy.

None of this means the tools are not worth using. It means they need a knowledgeable hand guiding them and checking the work, which is exactly the role our team plays.

What to ask a partner about how they use AI

If you are evaluating an eLearning partner, AI is a fair thing to ask about directly. The goal is not to find someone who avoids these tools, since used well they make good work faster. The goal is to understand where a partner lets AI do the work and where a human stays responsible for it.

A few questions worth asking:

• Where in your process do you use AI, and where do people stay in control of the work?
• How do you check AI-generated content for accuracy, and for the pieces that quietly go missing?
• Who reviews the learning design itself, the objectives, the assessments, the scaffolding?
• How do you handle licensing and rights for any AI-generated visuals?

A good answer will be specific and unbothered. A partner who uses AI thoughtfully can tell you exactly where it fits and exactly where their team takes over, because they have already thought it through on every project.

The Conclusion

So, can AI build your eLearning course? It can draft parts of it, and it can help a skilled team work faster. It cannot listen to your organization, understand what your learners need, or notice the missing piece that makes the difference between polished and complete. That work stays human, and it is the part worth paying for.

If you are weighing how to build your next course and want a partner who is clear about where technology helps and where craft carries the work, we would be glad to talk it through with you.

Common questions about AI in eLearning

Can AI create a full eLearning course on its own?

AI can produce first drafts of scripts and activities quickly, but it cannot reliably deliver a complete, accurate course on its own. Output often looks finished while missing key instructional elements like assessments or scaffolding, which is why an experienced team still reviews and builds on everything it produces.

Is it safe to use AI-generated images in eLearning?

The licensing and rights around AI-generated images are still unsettled, so many studios work from licensed stock or original art instead. Using generative visuals responsibly usually means having a clear visual guide and understanding the rights involved.

How should AI be used in instructional design?

AI works best as a starting point for tasks where the direction is already set, such as drafting an activity within a defined theme and learning outcome. The instructional design decisions, and the review that catches what AI leaves out, stay with people.