How generative AI is really changing instructional design
Generative AI is disrupting practices across every industry, and instructional design is no exception. Between excessive enthusiasm and reluctance fuelled by misunderstanding or fear of change, the real question isn’t whether generative AI is transforming instructional design, but where its real impact begins and ends.
What generative AI really changes
The most immediate impact of generative AI on our profession concerns the creation of learning content. Scripts, storyboards, e-learning module structuring, scenario-based exercises, assessment questions, transforming subject-matter expertise into learner-friendly content: on these tasks, current language models deliver substantial productivity gains.
An experienced instructional designer who has mastered prompt engineering can now significantly reduce development time. Not because AI does the work for them, but because it quickly produces the raw material that the expert can then refine, restructure and contextualise.
Video, visuals and interactive interfaces are also starting to benefit from these advances, although the limitations in terms of quality and consistency are more pronounced. AI is, in particular, helping to democratise short formats such as microlearning and video-learning, thanks to the emergence of generative avatars. But the voice and image of a real presenter still carry a legitimacy that avatars have yet to earn. Vibe coding is also opening up new possibilities for prototyping, or even directly building, short interactive formats (a quiz, a simple simulation, a mini-course) without needing a full development team. An instructional designer can now test an interaction idea in a matter of hours, where it would previously have required back and forth with technical teams.
The limitations enthusiasm sometimes overlooks
The first limitation lies in the quality of the needs analysis. AI produces whatever it’s asked to produce. If the question is poorly framed, if the real need hasn’t been correctly identified, AI will produce content very quickly and very badly. The value still lies in the human expertise that defines the problem before looking for a solution.
The second limitation concerns understanding the learner context. A language model doesn’t know that your target audience is made up of time-poor managers who never have more than 20 minutes to spare. It doesn’t know that your client operates in a corporate culture where training is seen as a chore. This contextual information, which shapes every design decision, comes from experience and listening, not from an algorithm.
The third limitation concerns instructional creativity. AI produces statistically probable content, that is, content close to the average of what already exists. Drawing on widely shared datasets, it also tends to produce content that closely resembles competitors’ output: the same phrasing, the same examples, the same turns of phrase. For complex, differentiating projects, or to capture the attention of an already over-solicited learner, human added value remains decisive.
TAKOMA’s approach
After using generative AI in our production processes for over two years, our assessment is a pragmatic one.
We use it to speed up content production, generate variations on pedagogical scenarios, and quickly test different course structures. On these uses, the gain is real and measurable.
We don’t use it to replace diagnostic work, field observation or strategic design. These stages remain the core of our added value, and no tool, however capable, can substitute for them.
Our position with clients is straightforward: AI allows us to be faster and more responsive on content production, without ever reducing the level of human expertise applied to diagnosis and design. This choice is reflected in the quality of our deliverables.
Generative AI is a powerful tool in the service of instructional design, provided we don’t confuse speed of production with quality of design. At TAKOMA, striking that balance is what we aim for on every project.
Are you starting to integrate AI into your working practices? Are you looking to optimise your learning content creation without sacrificing the quality of your learning programmes?